2008-01-28 22:20 gb * [r8334] ChangeLog: Regenerated 2008-01-28 21:20 gb * [r8333] ChangeLog: Regenerated 2008-01-28 20:20 gb * [r8332] ChangeLog: Regenerated 2008-01-28 19:20 gb * [r8331] ChangeLog: Regenerated 2008-01-28 18:20 gb * [r8330] ChangeLog: Regenerated 2008-01-28 17:20 gb * [r8329] ChangeLog: Regenerated 2008-01-28 16:20 gb * [r8328] ChangeLog: Regenerated 2008-01-28 15:20 gb * [r8327] ChangeLog: Regenerated 2008-01-28 14:20 gb * [r8326] ChangeLog: Regenerated 2008-01-28 13:20 gb * [r8325] ChangeLog: Regenerated 2008-01-28 12:20 gb * [r8324] ChangeLog: Regenerated 2008-01-28 11:20 gb * [r8323] ChangeLog: Regenerated 2008-01-28 10:20 gb * [r8322] ChangeLog: Regenerated 2008-01-28 09:20 gb * [r8321] ChangeLog: Regenerated 2008-01-28 08:20 gb * [r8320] ChangeLog: Regenerated 2008-01-28 07:20 gb * [r8319] ChangeLog: Regenerated 2008-01-28 06:20 gb * [r8318] ChangeLog: Regenerated 2008-01-28 05:20 gb * [r8317] ChangeLog: Regenerated 2008-01-28 04:20 gb * [r8316] ChangeLog: Regenerated 2008-01-28 03:20 gb * [r8315] ChangeLog: Regenerated 2008-01-28 02:20 gb * [r8314] ChangeLog: Regenerated 2008-01-28 01:20 gb * [r8313] ChangeLog: Regenerated 2008-01-28 00:20 gb * [r8312] ChangeLog: Regenerated 2008-01-27 23:20 gb * [r8311] ChangeLog: Regenerated 2008-01-27 22:20 gb * [r8310] ChangeLog: Regenerated 2008-01-27 21:20 gb * [r8309] ChangeLog: Regenerated 2008-01-27 20:20 gb * [r8308] ChangeLog: Regenerated 2008-01-27 19:20 gb * [r8307] ChangeLog: Regenerated 2008-01-27 18:20 gb * [r8306] ChangeLog: Regenerated 2008-01-27 17:20 gb * [r8305] ChangeLog: Regenerated 2008-01-27 16:20 gb * [r8303] ChangeLog: Regenerated 2008-01-27 15:20 gb * [r8302] ChangeLog: Regenerated 2008-01-27 14:20 gb * [r8301] ChangeLog: Regenerated 2008-01-27 13:20 gb * [r8300] ChangeLog: Regenerated 2008-01-27 12:20 gb * [r8299] ChangeLog: Regenerated 2008-01-27 11:20 gb * [r8298] ChangeLog: Regenerated 2008-01-27 10:20 gb * [r8297] ChangeLog: Regenerated 2008-01-27 09:20 gb * [r8296] ChangeLog: Regenerated 2008-01-27 08:20 gb * [r8295] ChangeLog: Regenerated 2008-01-27 07:20 gb * [r8294] ChangeLog: Regenerated 2008-01-27 06:20 gb * [r8293] ChangeLog: Regenerated 2008-01-27 05:35 gb * [r8292] ChangeLog: Regenerated 2008-01-27 05:20 gb * [r8291] ChangeLog: Regenerated 2008-01-27 04:20 gb * [r8290] ChangeLog: Regenerated 2008-01-27 03:20 gb * [r8289] ChangeLog: Regenerated 2008-01-27 02:23 gb * [r8288] ChangeLog: Regenerated 2008-01-27 02:20 gb * [r8287] ChangeLog: Regenerated 2008-01-23 02:08 gb * [r8236] lib/format.lisp: Daniel Dickinson's patch to ticket:230 (format ~F). 2008-01-20 00:02 gb * [r8228] lisp-kernel/pmcl-kernel.c, lisp-kernel/ppc-gc.c: Fix typo in #-DARWIN sample_paging_info. Collect gc paging info on ppc, too. 2008-01-19 23:31 gb * [r8227] lisp-kernel/gc.h, lisp-kernel/pmcl-kernel.c, lisp-kernel/x86-gc.c: Report paging stats when gc verbose. 2008-01-16 09:27 gb * [r8157] .: copy 2008-01-11 15:49 gz * [r8052] : For ticket #228: reset thread state whenever reset the tcr (new-tcr-for-thread, init-thread-from-tcr). 2008-01-08 07:00 gb * [r8015] : Fix typo in linuxx8664 version of xp-mxcsr. 2008-01-08 05:43 gb * [r8014] : Add a STATUS slot to ARITHMETIC-ERROR, with reader CCL::ARITHMETIC-ERROR-STATUS. Define xp-mxcsr for all current x86-64 platforms; use it to access the mxcsr on an exception which generates an ARITHMETIC-ERROR, and initialize that ARITHMETIC-ERROR's status slot to the mxcsr value. 2008-01-07 21:21 mikel * [r8013] : added final text and pictures to Tiger version of IB instructions for Currency Converter HOWTO 2008-01-05 15:53 gb * [r8002] : Try to ensure that a feature naming the OS is on *features* if a corresponding :[OS[-TARGET was present at build time. 2008-01-04 05:48 rme * [r7996] : Use imm2 to hold pointers, rather than assuming/hoping that the addresses will be 64-bit aligned, and hence fixnum-tagged. 2008-01-03 03:30 rme * [r7991] : Update %aset2, %aref3, and %aset3 to take account of the return address on the stack. It's not easy for these functions to get called, but they shouldn't just sit around being wrong. 2008-01-03 02:51 rme * [r7990] : fpzero (not fp0) is the register with zero in it. 2007-12-31 23:05 gz * [r7968] : Fix ticket #106: allow toplevel-read to return a set of bindings to be used and updated by toplevel-eval. Arrange for selection-input-streams to use this to set up *package* and have it persist over multiple forms in a single selection. 2007-12-21 23:02 mikel * [r7930] : added more steps to creating UIs with Tiger 2007-12-21 16:49 gz * [r7927] : allow non-echoing input streams for the repl 2007-12-21 16:26 gz * [r7926] : split off housekeeping-loop so the gui can use it 2007-12-19 01:42 gb * [r7921] : Learn to read disassembled code; look for _sigtramp context in r8 on all Darwin versions. 2007-12-18 18:19 mikel * [r7920] : added more steps in the Tiger IB instructions 2007-12-13 23:30 gb * [r7909] : Save major OS version on startup, use it to determine where _sigtramp keeps its contenxt argument (which differs between 10.4 and 10.5.) 2007-12-13 17:59 gz * [r7905] : Use defloadvar for *quitting* (needed because of changeset #7699) 2007-12-12 20:41 gb * [r7902] : When entering the kernel debugger on X8664/Darwin, try to note if the unhandled exception occurred at _sigtramp and find and use the signal context passed to _sigtramp if so. 2007-12-12 04:08 rme * [r7885] : Correct type specifier for 3d arrays in %TYPE-ERROR-TYPESPECS% 2007-12-12 01:13 gz * [r7884] : Make Indent With Tabs be a simple boolean variable, default false (ticket #187); also make non-lisp mode indent obey it as well. 2007-12-07 21:15 gz * [r7845] : Print the 'defining new Objc message' note if either *compile-print* or (the new variable) *objc-verbose* is true. Unimportantly default *objc-verbose* to false 2007-12-06 19:52 gb * [r7834] : Change order of clauses in FASL-DUMP-DISPATCH, so that the IVECTOR case precedes the VECTOR case. (This allows FIXNUM vectors - which don't yet have highly specialized support - to be dumped as IVECTORs, which is a whole lot better than %SVREFing around in them ...) 2007-12-06 19:28 rme * [r7832] : Set table view data source to nil before autoreleasing window controller so that the table view doesn't get any ideas about redrawing itself as things are getting torn down. 2007-12-06 04:25 rme * [r7829] : Remove stray backquote. 2007-12-05 14:14 gb * [r7824] : Use commas when printing bytes allocated/freed in GC messages; try to make fields in before/after GC messages line up. 2007-12-03 11:30 gb * [r7809] : Use WITHOUT-INTERRUPTS in FD-STREAM-FORCE-OUTPUT. 2007-12-03 11:30 gb * [r7808] : Use *LISTENER-PROMPT-FORMAT*.in PRINT-LISTENER-PROMPT. 2007-12-03 11:29 gb * [r7807] : Export *LISTENER-PROMPT-FORMAT*. 2007-12-03 11:28 gb * [r7806] : Complain if lisp stack pointer isn't in lisp stack when calling out for process-interrupt. 2007-12-03 11:14 gb * [r7805] : Fix INFER-LINE-TERMINATION termination. Wasn't this checked in a few days ago ? 2007-12-01 22:43 mikel * [r7800] : various corrections to HOWTO; added section on IB 2.x 2007-12-01 22:28 mikel * [r7799] : various corrections to HOWTO; added section on IB 2.x 2007-11-30 00:01 mikel * [r7786] : tested build-application with CurrencyConverter; added CurrencyConverter.nib 2007-11-29 23:45 mikel * [r7785] : tidied up build-application; changed default app class to GUI::COCOA-APPLICATION 2007-11-29 23:00 mikel * [r7784] : debugging build-application 2007-11-29 13:27 gb * [r7779] : lisp_Debugger() takes an extra Boolean "in foreign context" arg, rather than trying to set a bit in the exception code (which might be negative.) Add a (T)hread info command to kernel debugger. 2007-11-29 13:22 gb * [r7778] : Use "uname -m" to determin architecture on Linux, since some distros don't include /bin/arch. 2007-11-29 09:46 gb * [r7777] : Add a comment or two to Y-OR-N-P (and it was ticket:190 that described the incorrect behavior.) 2007-11-29 09:41 gb * [r7776] : Fix Y-OR-N-P: don't assume that PEEK-CHAR does a READ-CHAR, but if we see a buffered #\Newline, read it. This should fix ticket:190, which should have been called "Y-OR-N-P makes invalid assumptions about how PEEK-CHAR is implemented." (All of this just has to do with consuming a trailing #\Newlime so that it isn't echoed.) 2007-11-28 18:22 af * [r7767] : Add easygui:cell-count for forms. 2007-11-28 13:37 rme * [r7766] : %natural- and %natural--c vinsns take only two result/argument specs. 2007-11-24 19:18 mikel * [r7714] : replaced IB 2.5 nibfile with IB 3.0 nibfile 2007-11-24 19:17 mikel * [r7713] : replaced IB 2.5 nibfile with IB 3.0 nibfile 2007-11-24 19:15 mikel * [r7712] : added explanation of how to add custom classes and connections in IB 3.0 2007-11-23 02:12 mikel * [r7711] : added instructions for adding custom classes 2007-11-23 01:20 mikel * [r7710] : removed obsolete IDE-building instructions 2007-11-23 01:19 mikel * [r7709] : added several images for the HOWTO 2007-11-21 20:08 gz * [r7706] : couple cases of missing package prefix 2007-11-21 01:02 rme * [r7701] : Respect the "use screen fonts" preference in editor/listener echo areas. 2007-11-20 16:26 gz * [r7700] : Delete unused file, already archived in unused/archive/srccom.lisp 2007-11-20 16:21 gz * [r7699] : prepare-to-quit: leave *quitting* set to T if really quitting 2007-11-20 15:14 gz * [r7698] : A new package and a reorg: I put all the cocoa-ide files (except for a greatly stripped-down cocoa.lisp and cocoa-application.lisp) in a new package named "GUI". The package is defined in defsystem.lisp, which also defines a function to load all the files explicitly, putting the fasls in cocoa-ide;fasls; I stripped out all pretense that the files can or should be loaded individually. Also, it is no longer necessary or appropriate to compile hemlock separately, as it now compiles as needed as part of the normal loading sequence. (Over time I am hoping to get hemlock more and more integrated into the IDE, and having to maintain it as if it still were a separate package is an unnecessary burden). Updated the README file appropriately. 2007-11-20 00:40 rme * [r7684] : Add preference option so that bitmap screen font substitution can be disabled. 2007-11-19 21:48 gz * [r7683] : Move hemlock/src/archive up to hemlock/unused/archive, for easier grepping 2007-11-19 13:12 gb * [r7682] : (from branches, not in trunk before): try to ensure that rsp/rbp are lisp's when asserting TCR_STATE_LISP: we don't want to run an interrupt callback on the foreign stack. 2007-11-15 02:33 rme * [r7643] : Enhance apropos window. 2007-11-15 02:32 rme * [r7642] : More Cocoa memory management. (Bring on the GC, please.) 2007-11-13 14:53 af * [r7641] : Improve easygui view hierarchy handling. From Ron Garret's list of things: implement remove-subviews. * Add an example, example/view-hierarchy.lisp. * Move the other examples into example/ as well. * Implement a few memory management helpers: functions retain-object, release-object, and macro retaining-objects. * Allow box-views to have a title. 2007-11-13 02:56 rme * [r7640] : Create obsolete/ directory and move CocoaBridgeDoc.txt to it. Add an explanatory README referencing the release notes. 2007-11-11 22:41 gb * [r7635] : Spell charset name(s) correctly in STREAM-EXTERNAL-FORMAT method on CHARACTER-STREAM. 2007-11-11 22:40 gb * [r7634] : Propagage Gary Palter's fix to %GET-IP-INTERFACES to trunk. 2007-11-11 20:39 gb * [r7633] : In X862-MVCALL, account for effect of returning to label pushed earlier. 2007-11-09 22:37 rme * [r7621] : Fix some Cocoa object leaks. 2007-11-09 19:28 rme * [r7618] : Make Show Callers command display a scrolling list of the callers of the symbol at point. Double-click a row in the list to go to the definition of the caller. 2007-11-09 18:12 rme * [r7617] : Implement very naive Show Callers command. 2007-11-09 17:58 rme * [r7616] : Only send Hemlock a no-op if the left mouse button is clicked with no modifier keys. (This fixes ticket:93) 2007-11-09 10:37 gb * [r7615] : In PPC2-ASET2, we need to use 4 targeted regs when need-memoization is true. 2007-11-08 18:43 rme * [r7613] : Simplify definition of hemlock-document-controller class. 2007-11-08 18:20 rme * [r7612] : Get rid of commented-out code that will never be used. 2007-11-08 06:52 gb * [r7610] : Try (again) to work around a bug having to do with nanosleep returning EINTR and also exceeding the original sleep time. (AFAICT, this was fixed a long time ago in Leopard betas, a workaround that seemed to negatively affect Linux was removed a few weeks ago, and the bug is back in in the final version of Leopard, at least on x86-64.) 2007-11-08 03:44 gb * [r7609] : Add a new reader macro (#>). #>foo -> :FOO, and #>Foo -> :OO. (e.g, reduce the need to type as many angle brackets in foreign type/ accessors names.) 2007-11-08 03:42 gb * [r7608] : Add READ-SYMBOL-TOKEN, use it in the sharp-colon reader macro function. 2007-11-08 02:59 rme * [r7607] : Remove "Save" from the Default Modes list. (The implementation of Save mode was moved to the archive/ directory.) 2007-11-08 00:48 rme * [r7606] : Remove debugging code accidentally left in. 2007-11-07 20:05 rme * [r7604] : Implement "wrap lines to window" editor preference. (still need to get rid of horizontal scroller when this preference is selected.) 2007-11-07 17:31 rme * [r7603] : Ensure that text container width tracking and text view horizontal resizing are not both enabled at the same time. 2007-11-06 07:30 gb * [r7598] : more ObjC frameworkds 2007-11-06 02:25 gb * [r7597] : Add some more frameworks. Update the CFLAGS in Cocoa-based frameworks' populate.sh scripts to include CFLAGS appropriate for gcc-5465. 2007-11-05 20:41 rme * [r7596] : Use proper bold font. 2007-11-05 06:39 gz * [r7595] : A bit of cleanup just so I don't have to merge as much later: Get rid of gap symbol-macros, get rid of hi::*buffer-gap-context*. Change editor source code to use functions to look up info in hi::*current-buffer*. Code that wants to do quick-and-dirty editor stuff out of context needs to bind hi::*current-buffer*. Note that really you should setf (current-buffer) but currently that's a little heavy duty. Don't export *editor-input* (it will go away eventually, for now I want warnings to see where it's used). Get rid of some #+clx stuff. Make m-x grep be case-insensitive by default. See *grep-ignore-case*. List all the ide source files explicitly (in cocoa.lisp), so I don't have to follow REQUIRE trails to see what's going on. A fix in hi::buffer-note-insertion for temporary marks. 2007-11-05 06:11 gz * [r7594] : Make %show-stack-frame and %show-args-and-locals catch errors. This makes it possible to see the rest of the backtrace even if a particular frame craps out. 2007-11-05 06:09 gz * [r7593] : make print-object catch any errors and just print #. This can be disabled (e.g. if you're debugging print methods) by setting *print-catch-errors* to nil 2007-11-05 05:46 gz * [r7592] : Move elisp/, spell/, tty/, and wire/ into archive, as they're not used 2007-11-03 21:01 gb * [r7591] : Hack wouldn't work on FreeBSD; try conditionalized hack. 2007-11-03 19:27 gb * [r7590] : crude hack to enable running on kernel 2.6.23 2007-11-02 19:55 rme * [r7589] : Make the apropos window work on Tiger. The main thing is to disable automatic KVO notifications, and send them manually (with willSetValueForKey/didSetValueForKey). 2007-11-02 19:53 rme * [r7588] : Set the text fields cell to allow rich text. 2007-11-02 17:31 rme * [r7587] : Use IB 2.5.6 to save the nib, explicity making AproposWindowController a subclass of NSWindowController. 2007-11-02 03:46 rme * [r7586] : Inspect the right thing when the user re-sorts the table view. 2007-11-01 23:45 rme * [r7585] : Comment out some code not in use yet. 2007-11-01 23:42 rme * [r7584] : New feature: basic apropos window. This needs to be enhanced, but it's a place to start. 2007-11-01 23:40 rme * [r7583] : Indentation tweak for rlet 2007-11-01 17:26 rme * [r7580] : Let user select different fonts for listener input and output. 2007-10-31 19:49 rme * [r7579] : De-kludge defining font defaults a little bit. 2007-10-31 19:06 rme * [r7578] : Add dead-macptr-p. 2007-10-31 17:35 rme * [r7577] : Leave out debugging code, duh. 2007-10-31 17:33 rme * [r7576] : Different kludge to deal with NSFont objects and def-cocoa-default. 2007-10-31 16:46 rme * [r7575] : Initialize the defaults that are NSFont objects in +[LispApplicationDelegate initialize]. 2007-10-31 06:07 gb * [r7574] : Use documented foreign function calls (only) in ENABLE-FOREGROUND. Need to double-check to ensure that that all works on PPC/Tiger as well. The old mechanism - CPSEnableForeground - was generating deprecation warnings in /var/log/system.log. (Along with all of the usual messages about Xcode memory leaks ...) 2007-10-31 05:52 rme * [r7573] : Make sure toolbar button doesn't show up on the prefs panel. (For some reason, setting it in IB3 doesn't seem to do the trick on Tiger/PPC32.) 2007-10-31 05:39 rme * [r7572] : Use image for General prefs item 2007-10-31 05:36 rme * [r7571] : New file for benefit of non-Leopard systems. 2007-10-31 04:52 rme * [r7570] : Update prefs panel title according to selected view 2007-10-31 04:44 rme * [r7569] : Revert change to treat CCL as a special Objective-C word. 2007-10-31 03:18 rme * [r7568] : Correct 32/64 bit portability error. 2007-10-31 03:04 rme * [r7567] : Remove old prefs panel 2007-10-31 03:03 rme * [r7566] : Remove Leopard-ism. 2007-10-31 02:58 rme * [r7565] : CGFloat screw, again. 2007-10-31 02:56 rme * [r7564] : Grr. 2007-10-31 02:53 rme * [r7563] : New user default editorFont, an actual NSFont instance. New user default wrapLinesToWindow, not used yet. Local variants of create-paragraph-style and create-text-attributes, in preparation for further refactoring. Remove defaults for modeline font. Don't enable HyperSpec menu item if HyperSpec lookup is disabled. Don't use old preference panel. Use hemlock-document-controller instead of ns:ns-document-controller in a couple of places. 2007-10-31 02:46 rme * [r7562] : Define two new listener font defaults. Not used yet. 2007-10-31 02:45 rme * [r7561] : Retarget Preferences... menu item action. 2007-10-31 02:44 rme * [r7560] : Comment out old showPreferences: action. Correct a couple of Cocoa memory leaks. 2007-10-31 02:42 rme * [r7559] : Use new user defaults key hyperspecURLString. 2007-10-31 02:39 rme * [r7558] : Add :font type. 2007-10-31 02:37 rme * [r7557] : New preferences controller, with updated nib and some toolbar item icons. 2007-10-31 02:34 rme * [r7556] : Move lisp-application-delegate class from cocoa.lisp into new file app-delegate.lisp. 2007-10-31 02:32 rme * [r7555] : Use a better toolbar identifier; release the toolbar after adding it to the window. 2007-10-30 20:12 gb * [r7554] : Show function docstring when describing functions. 2007-10-30 17:41 jaj * [r7552] : Add support for redo. 2007-10-30 15:01 jaj * [r7551] : Make control-w in incremental search correctly echo the search string. Fixes reopened bug#46. 2007-10-30 09:59 gb * [r7550] : Rename #/insertText: to #/insertString:, since the text system calls #/insertText: on key events even if the text view isn't editable (and gets confused if out #/insertText: makes it editable on the fly.) See ticket:121 . 2007-10-29 00:36 gz * [r7541] : Make meta-. more forgiving: if symbol not found in current package, look in other packages 2007-10-29 00:08 gz * [r7540] : Implement symbol completion, and bind it to meta-/ (as in Emacs and Lispworks) and meta-` (as in MCL). (defcommand "Expand Dynamic Abbreviation" (p) "Treats the symbol before point as an abbreviation and expands it. It checks the following in order until a suitable expansion is found: - last accepted expansion for this abbreviation, if any - symbols in current buffer before point - symbols in current buffer after point - symbols in all other editor windows, front to back - symbols visible in the current package, fbound/bound symbols first - symbols in all other packages (in no particular order) If called repeatedly from the same position, replaces the previous expansion with the next possible one. A symbol is a suitable expansion for an abbreviation if the abbreviation is a proper prefix of the symbol, or the abbreviation consists of the initials of the individual words within the symbol (e.g. mvb => multiple-value-bind). " (declare (ignore p)) 2007-10-28 23:39 gz * [r7539] : Add comment 2007-10-28 16:12 gz * [r7538] : Indentation tweaks for if, make-instance 2007-10-28 15:16 gz * [r7537] : Add buffer-package, returns package object or nil 2007-10-28 11:52 gb * [r7536] : New. 2007-10-28 03:12 rme * [r7535] : Remove redundant form. 2007-10-28 02:59 rme * [r7534] : Make echo area and window resize handle area white. 2007-10-28 02:03 rme * [r7533] : Beautify modeline. 2007-10-27 19:59 mikel * [r7532] : added implementations for the Compile Buffer and Compile And Load Buffer menu items 2007-10-27 19:01 mikel * [r7531] : implemented the Load Buffer menu item 2007-10-26 19:02 jaj * [r7530] : Fix crashing bug introduced by wrapping incremental search. For the record, when calling push-buffer-mark, you should be careful to make a copy of the mark. 2007-10-26 14:25 af * [r7529] : Implement slider-view. * Introduce a value-mixin (and subclasses for string and numeric values). * Use it in view-text-via-stringvalue-mixin. * Put the demos into the easygui-demo package, export their windows. 2007-10-26 02:09 gz * [r7528] : Got tired of restarting my lisp all the time, made various hemlock printing functions bind the right magic context variables. (This is way ugly but temporary, as the whole threading scheme needs to be replaced by something less fragile anyway, and this can be cleaned up then). I'm checking in Matt's patch for ticket #135, because it's in my sources, it works, and whatever else might be wrong with it is better than the original bug. add find-not-attribute and reverse-find-not-attribute so I don't have to remember which way the magic arg to find-attribute goes. make region-to-string take an optional output-string argument, return length as second value. Implement hi:window-buffer, also add ordered-hemlock-windows (returns a front-to-back list of hemlock windows). 2007-10-25 16:47 rme * [r7520] : Add CCL special prefix. 2007-10-25 14:22 mikel * [r7519] : added menu items for Compile Buffer, Load Buffer, and Compile and Load Buffer. Implementation of the actions is pending 2007-10-25 14:21 mikel * [r7518] : added a definition for ccl::recursive-delete-directory 2007-10-25 14:20 mikel * [r7517] : added a definition for ccl::%rmdir 2007-10-24 20:29 gz * [r7510] : Make the function-doc argument (i.e. the "second doc string") to DEFCOMMAND be optional, because I hate fascist software. 2007-10-24 19:33 jaj * [r7509] : Implement wrapping in incremental search. 2007-10-24 18:17 rme * [r7508] : Print out a message saying where the generated application bundle is being saved. 2007-10-24 06:06 rme * [r7507] : Argh. Get the name the finder shows for .lisp files right. 2007-10-24 04:37 rme * [r7506] : Replace Credits.rtf with Credits.html (for ease of including links). Include a link to the OpenMCL Trac site to encourage users to go there and report bugs and enhancements. 2007-10-24 04:21 rme * [r7505] : Put the value of *openmcl-svn-revision* into Info.plist as the value of CFBundleVersion. (This shows up in the about box.) 2007-10-24 03:52 rme * [r7504] : Use the right value for CFBundleTypeName. 2007-10-24 03:28 rme * [r7503] : Address ticket:136 (make lisp source files show up in quick look) Add UTI information for lisp files to Info.plist-proto, and also add another document type for txt/text/* documents, removing the * from the lisp type. In cocoa-editor.lisp, tell the save panel to keep the whining down and to let the user save files with any extension. 2007-10-23 21:28 rme * [r7502] : Resolve ticket:111 (shouldn't be able to plain save in listener) hemlock/src/bindings.lisp: In Listener mode, rebind c-x c-s to "Illegal" and bind c-x c-w to "Save To File". hemlock/src/filecoms.lisp: New command "Save To File". Added Save To... item to File menu. cocoa-listener.lisp: Arrange for Save and Save As... menu items to be disabled when a listener window is main. cocoa-editor.lisp: Add method hi::save-hemlock-document-to 2007-10-23 17:30 rme * [r7501] : Include omitted format arg. 2007-10-23 14:36 af * [r7500] : Make the easygui currency converter demo more robust. * No longer break on pushing "Convert" if non-numbers are in Amount and Exchange rate, and * make the result field non-editable. 2007-10-23 14:28 af * [r7499] : Implement mouse event methods for drawing views. * Print method for points * redisplay method * Update tiny.lisp to provide mouse handler example: drag up to increase Number of points, down to decrease it. 2007-10-23 00:01 gz * [r7498] : Don't need the MacOS subdir any more 2007-10-22 23:49 gz * [r7497] : Got rid of ccl:cocoa-ide;Clozure CL.app and moved its contents to ccl:cocoa-ide;ide-contents; Modified the IDE loading scheme to create bundles by copying from cocoa-ide;ide-contents; as follows: (require'cocoa) creates 'ccl:temp bundle.app', a fake bundle that is only there to enable using cocoa immediately after loading it. This is not something that users are expected to do in normal circumstances, but is just there for us to bootstrap. (Users who use cocoa will ordinary use the IDE, which will have cocoa preloaded). (require'cocoa-application) creates 'ccl:Clozure CL.app', the real IDE bundle that is ready to use and/or redistribute. By default it includes the interface headers for the target architecture, so it can be moved out of the ccl directory if so desired. I also made it load easygui unconditionally. 2007-10-22 21:15 rme * [r7495] : I'm going to go ahead and commit this, even though I'm not happy with it. I removed an old workarond for what appears to have been a Carbon bug of some sort, so if something blows up, I guess that's a prime candidate for examination. I'm almost convinced that it would be advantageous to make nib files for editor and listener windows, just so that it's easier to get all the details right. 2007-10-22 19:59 gz * [r7494] : Moved some non-cocoa pathname utilities out of cocoa-ide to core lisp: ensure-directory-pathname converts a pathname to be directory-pathname-p recursive-copy-directory copies directory and all subdirectories Tweaked them some to make them more general, in particular made them work on logical pathnames, made the latter accept a :test arg and a limited set of :if-exists values. The default is now :if-exists :error. 2007-10-22 17:37 rme * [r7493] : Implement "Execute Selection" item on the Lisp menu, with shortcut command-E. It's at least semi-standard for command-E to be used for "Use Selection for Find", but in a nod to MCL muscle memory, we'll usurp it. cocoa.lisp: added #/ensureListener: action (and minor update to #/showListener:). cocoa-listener.lisp: call #/ensureListener: on the main thread from ui-object-choose-listener-for-selection so that we always have a listener to send data to. cocoa-editor.lisp: add menu item validation for #/evalSelection:. (As a bonus, changed OpenMCL Help... in the Help menu to Clozure CL Help...) 2007-10-22 04:28 rme * [r7489] : Revise behavior of command-L shortcut. When no listener is open, command-L will create one. If a listener window is not key, command-L will bring the closest listener (in window z-order) to the front and make it key. If a listener window is already key, command-L will cycle between other open listeners. If there are no other open listeners, it will just beep. 2007-10-21 11:53 gz * [r7487] : Add :clozure-common-lisp to *features*. Allow the init-file arg to startup-ccl to be a list of filenames - the first one that exists is used. Make the lisp-development-system application-init-file be the list of ccl-init and openmcl-init, but warn if openmcl-init is the one that's actually used. Delete cocoa-ide/OpenMCL.app to minimize confusion (this means removing :clozure-common-lisp from *features* no longer works for the IDE, although it still works for the command line build). 2007-10-20 01:01 rme * [r7483] : Use DEFLOADVAR, not DEFVAR, to define *processes-window-controller*. 2007-10-19 20:03 rme * [r7481] : www.lisp.org URL doesn't work; use www.lispworks.com instead. 2007-10-19 19:46 rme * [r7480] : Call function HYPERSPEC-ROOT-URL instead of referencing the corresponding special variable. Make #/newDisplayDocumentWithTitle:content: return an :id instead of nil, thereby avoiding a huge backtrace. 2007-10-19 18:52 rme * [r7479] : Fix typo. 2007-10-19 17:11 rme * [r7478] : Update nib files for processes window in Clozure CL.app, too. 2007-10-19 15:48 rme * [r7477] : Add a Tools menu to the IDE, and put a Processes... menu item in it. This menu item displays a utility window listing all the processes (threads) running in the lisp. Double-click on a process to inspect it, or use the toolbar items. 2007-10-19 04:22 gb * [r7476] : make modeline font pref work 2007-10-18 21:17 jaj * [r7475] : Implement c-w for incremental search. Bind del to Delete Next Character. Add a log-debug utility that makes it easier to call NSLog. 2007-10-18 16:27 gz * [r7474] : Fix fencepost 2007-10-18 15:27 gz * [r7473] : In the standalone cocoa ide, check if there are header directories inside the bundle, and if so, use them. This makes it possible (though not required) to distribute a standalone IDE that works e.g. from the desktop and still offers basic Cocoa support. Note that this isn't implemented by making "CCL:" point inside the bundle, instead, it works by making the header reader not use "CCL:". Change the default *ccl-directory* preference to be the empty string, and make it mean the directory containing the application bundle. If there is nothing else of interest there, certain things (e.g. meta-. on preloaded stuff, require) will not work unless other arrangements are made (e.g. explicitly adding other locations to the require path, etc). 2007-10-18 15:14 rme * [r7471] : When generating ff-calls to objc methods that have formal parameters of type CGFloat, coerce the actual parameter to the appropriate kind of float. This coercion is not performed for variadic methods 2007-10-18 15:07 gz * [r7470] : Allow the possibility that headers can be stored somewhere other than "ccl:": A new variable, *interfaces-root*, default "ccl:", can be set to change the root headers directory. 2007-10-18 15:01 gz * [r7469] : make ccl-directory compute the heap image name rather than relying on *heap-image-name*, which might still have a stale value. This fixes a bug in initializing "ccl:" for the command line version. 2007-10-17 23:15 gb * [r7465] : don't trust assembler not to relax branch in CONS, %ALLOCATE-UVECTOR 2007-10-17 17:56 rme * [r7464] : Don't use a border on scroll views in editor/listener windows. 2007-10-17 17:53 gb * [r7463] : bogons in EGC-CONFIGURATION, CONFIGURE-EGC 2007-10-17 14:15 jaj * [r7461] : First pass at implementing undo. There is still no redo or grouping. 2007-10-16 22:52 gz * [r7460] : Conditionally (#+easygui) preload EasyGUI into cocoa-application 2007-10-16 22:00 gz * [r7459] : add easygui 2007-10-16 17:34 gb * [r7457] : try to detect exception in foreign context early 2007-10-16 15:14 gz * [r7455] : Make find-module-pathnames handle wildcards, and put ccl:examples 2007-10-15 18:44 gb * [r7451] : In %CSTR-POINTER, either do the simple-string case or call %CSTR-SEGMENT-POINTER to handle the general case. Don'd do both ... 2007-10-15 18:43 gb * [r7450] : When reviving ObjC classes, zero out the class/metaclass pointers of each class if they were "dead macptrs". We don't want to revive a pointer at its old address, since that may accidentally conflict with some other class's current address, causing grief and dispair and much gnashing of teeth in GDB. 2007-10-15 03:26 gb * [r7449] : ignorable declaration in COLLECT 2007-10-15 03:04 gb * [r7448] : Needed more parens. 2007-10-15 02:52 gb * [r7447] : Most changes are just indentation ... Precompile send functions, send-super functions for each method signature. 2007-10-15 02:51 gb * [r7446] : New "logical" string for FAKE-CFBUNDLE-PATH to replace; affects the name of the preferences file. FAKE-CFBUNDLE-PATH wants an argument (reversed domain thing) to prepend to the (conditional) bundle name. 2007-10-15 02:45 gb * [r7445] : Pass another arg (reversed domain URL) to FAKE-CFBUNDLE-PATH. 2007-10-15 02:44 gb * [r7444] : STREAM-SURROUNDING-CHARACTERS specialized to T. 2007-10-15 01:05 gb * [r7443] : too much zeal 2007-10-14 08:57 gb * [r7440] : Try to ensure that the directory namestring returned from #/selectCCLdirectory has a trailing slash (see ticket:113). 2007-10-14 08:55 gb * [r7439] : Try to detect the case where declared classes aren't yet available (loaded ?) in the runtime environment when we're ready to save an image. 2007-10-14 08:53 gb * [r7438] : Conditionalize out an _NSLog call in #/replaceCharactersInRange:withString: 2007-10-14 00:03 gb * [r7436] : Restore something that had been fixed in changeset:6737 but got clobbered (comparisons when both args fixnum constants.) 2007-10-13 22:12 mikel * [r7435] : added currency converter example and HOWTO 2007-10-13 16:00 gb * [r7434] : Still more bogus syscall fixes. 2007-10-13 01:16 gb * [r7433] : Simple case of INDENT: only move mark if it would otherwise wind up in initial whitespace (ticket:107). 2007-10-12 18:03 gb * [r7425] : Um, check in the right file. 2007-10-12 13:12 gb * [r7420] : Fix bogus definitions for getsockname and getpeername. 2007-10-11 22:27 gz * [r7377] : Different toy icon, just until we get some real ones 2007-10-11 03:31 gb * [r7376] : Don't automatically generate DYNAMIC-EXTENT declarations for simple pointer-typed callback arguments. 2007-10-11 02:51 gb * [r7375] : In BIT-OFFSET-TO-LOCATION, don't get confused by records/structs (whose alignment rarely matches their size.) 2007-10-09 04:21 gb * [r7372] : Regenerated 2007-10-09 03:25 gb * [r7371] : Regenerated 2007-10-09 02:53 gb * [r7370] : Try to make sure that the reader treats #\No-Break_Space as whitespace. 2007-10-09 02:50 gb * [r7369] : Treat non STANDARD-CHARs which obviously have "case" as if they had case; this affects UPPER-CASE-P/LOWER-CASE-P, CHAR-UPCASE/-DOWNCASE, ALPHA-CHAR-P and BOTH-CASE-P, [N]STRING-UPCASE/DOWNCASE, etc. This stuff is currently only defined for characters C for which either: (char-upcase C) is distinct from C, and downcasing the uppercase version returns C, or (char-downcase C) is distinct from C, and uppercasing the downcase version returns C. 2007-10-08 21:38 wws * [r7368] : Make the new fifth arg to arguments-and-locals optional. This isn't the correct fix. What really needs to happen is for FRAME-SUPPLIED-ARGS and (METHOD INSPECTOR::COMPUTE-FRAME-INFO (INSPECTOR::ERROR-FRAME T)) to be updated to use the new "unavailable" arg to arguments-and-locals, but I don't have time for that at present. I just want my slime backtrace to work. 2007-10-08 02:54 gb * [r7367] : (Hopefully) straighten out some backtrace stuff. 2007-10-07 22:21 gb * [r7366] : CALL .SPspecref, don't jump to it. Sheesh. 2007-10-05 20:42 gz * [r7365] : Support for creating Clozure CL.app rather than OpenMCL.app when :clozure-common-lisp is on *features* list 2007-10-05 20:23 gz * [r7364] : conditionalize lisp-implementation-type, application-init-file and add home:ccl;modules; to module search path 2007-10-05 17:10 gz * [r7363] : add assume-cocoa-thread and assume-not-editing fns, sprinkle calls around, by default these are noops but one can turn them on to try to debug ticket #41. 2007-10-05 16:06 gz * [r7355] : Account for grep's line numbers being 1-based. Add Hemlock variable "Grep Search Comments" to allow ignoring comments, but for now the only comments that are recognized are whole-line semi-colon comments. Pathetic but more useful than nothing at all... 2007-10-05 14:48 gb * [r7354] : Remove brain-damage from INIT-RANDOM-STATE-SEEDS. 2007-10-04 19:10 mikel * [r7352] : moved load-nibfile to builder-utilities 2007-10-04 18:11 mikel * [r7351] : fixed pathname bug in builder-utilities 2007-10-04 17:11 gb * [r7350] : INIT-RANDOM-STATE-SEEDS: use (%current-tcr), (primary-ip-interface-address) to generate more uniqueness. 2007-10-04 15:04 gz * [r7348] : Add a "read only listener" preference (defaultly true) to allow listener buffers to be editable. It currently only works on listeners created after you change the preference. It would be nice to have it toggle existing listeners when you change the preference, but I don't know how to intercept the changing of the value. 2007-10-03 23:21 gb * [r7345] : Regenerated 2007-10-03 22:55 gb * [r7344] : regenerated 2007-10-03 19:26 gb * [r7343] : %VALID-REMAINING-TIMESPEC-TIME-P was a buggy workaround for a post-10.4 bug that's since been fixed; if we can assume that #_nanosleep works, trust the remaining time it returns when interrupted. 2007-10-03 19:16 rme * [r7342] : Use NTH-IMMEDIATE to access closed-over values, enabling the x86-64 port to display the values correctly. 2007-10-02 19:30 palter * [r7341] : For :IF-EXISTS :SUPERSEDE, be sure to convert the target pathname to a native namestring before calling UNIX-RENAME. 2007-10-01 18:13 wws * [r7330] : (join-process (p process) &key default) Waits for process p to quit. Returns the values returned by it initial function if it quits normally. If killed, returns default if specified, or signals an error otherwise. Patterned after SBCL's join-thread function. 2007-09-30 06:21 gb * [r7329] : Regenerated 2007-09-30 05:58 gb * [r7328] : Regenerated 2007-09-30 05:51 gb * [r7327] : Regenerated 2007-09-29 22:21 gb * [r7323] : 2007-09-29 22:08 gz * [r7322] : Make c-x c-x activate region (first part of Ticket #86) 2007-09-29 04:22 gb * [r7320] : 2007-09-29 03:52 wws * [r7319] : Fix an obscure method combination bug in l1-dcode.lisp that's been there since Gail and I first implemented CLOS for MCL in 1990 or thereabouts. compute-method-list takes a list of possibly applicable methods, separates it up into before, after, around, and primary methods, and trims the tail of the around/primary methods at the point where there's no more call-next-method. Thing is, unless we're computing the methods for the last specialized argument, those methods are only POSSIBLY applicable. Some of them might not be. Hence, except for the final specialized argument, when sub-dispatch? is false, only if no around method does call-next-method will it be impossible to reach the primary methods, and we can't trim the primary methods at all. This fixes Gary Palter's no-applicable-method bug in the application conversion he's working on. 2007-09-28 20:21 gb * [r7317] : 2007-09-28 19:14 gb * [r7316] : Set ioblock encode/decode "literal limits" in SETUP-IOBLOCK-OUTPUT/INPUT, so that they're set if the encoding is changed on-the-fly. 2007-09-28 17:21 gb * [r7315] : 2007-09-28 16:56 gb * [r7314] : Try to make (SETF STREAM-EXTERNAL-FORMAT) work. Whether there should be a SETF inverse on a CL-package symbol that's not supposed to be SETFable is another question, as is whether/how well this works. 2007-09-28 01:21 gb * [r7313] : 2007-09-28 00:44 gb * [r7312] : When unlocking a read-write lock on x86-64, clear the owner field before incrementing the value field if the value field is currently -1. (The PPC was already doing so; doing it in the order that x86-64 had been doing it caused the lock to become free, and zeroing the owner after the incf could cause the lock to be locked for writing with no owner set ... 2007-09-26 20:21 gb * [r7310] : 2007-09-26 19:21 gb * [r7309] : 2007-09-26 19:14 wws * [r7308] : run-program and run-external-process take an additional "env" parameter, a list of pairs. For each pair the environment variable (string (car pair)) is set to (cdr pair) in the external program's environment before it is exec'd. 2007-09-26 19:13 palter * [r7307] : Allow CCL::DECODE-STRING-FROM-OCTETS to work on non-simple vectors 2007-09-26 18:44 palter * [r7306] : Fix vector-decode-function for :US-ASCII encoding 2007-09-26 04:21 gb * [r7302] : 2007-09-26 03:46 gb * [r7301] : Try to be more paranoid about null pointers (even if they have their type asserted.) ENCODE-OBJC-TYPE: try to get record field types right (being careful about recursion); need to do this to help NSInvocation deal with record types on some platforms that'll remain nameless. 2007-09-26 03:43 gb * [r7300] : ignore property set 2007-09-25 19:21 gb * [r7292] : 2007-09-25 18:44 jaj * [r7291] : Add function windows that returns an ordered list of the windows that are open. 2007-09-25 06:21 gb * [r7290] : 2007-09-25 05:58 gb * [r7289] : In .SPdestructuring_bind, don't assume that a "movzbl" sets the flags. (It doesn't, and this meant that there was a 50-50 chance of required arguments not being processed correctly.) 2007-09-25 05:55 gb * [r7288] : Reserve lcells when reserving outgoing frame, in case it matters someday. 2007-09-24 12:21 gb * [r7285] : 2007-09-24 12:04 gb * [r7284] : Missing #ifdef DARWIN in arbstack_signal_handler; not used on Darwin. 2007-09-24 03:21 gb * [r7283] : 2007-09-24 02:47 gb * [r7282] : Don't use sigaltstack on any x86-64 platform: even if it works, it makes it harder to deal with exceptions in foreign code (we've often gotten segfaults from running out of space on the alt stack, the mechanism isn't reentrant, etc.) Try to report cases where the kernel debugger is entered due to an exception in foreign code. Todo: make it less tempting to use (L) in that case, maybe try to make backtrace find saved_rbp from tcr on x86-64, etc. 2007-09-24 02:42 gb * [r7281] : APPLICATION-VERSION-STRING (which probably was used to support "get info", back in the day) is just LISP-IMPLEMENTATION-VERSION. 2007-09-23 15:21 gb * [r7280] : 2007-09-23 14:28 gb * [r7279] : Handle saving svn revision as string or fixnum. 2007-09-23 14:21 gb * [r7278] : Try to use "svnversion" to get svn version info in LOCAL-SVN-REVISION. 2007-09-23 14:10 gb * [r7277] : ignore property 2007-09-23 14:09 gb * [r7276] : ignore property 2007-09-23 05:21 gb * [r7275] : 2007-09-23 04:51 gb * [r7274] : Set *openmcl-svn-revision* to the value of (LOCAL-SVN-REVISION) when writing the bootstrapping image; this will show up in the Welcome banner and LISP-IMPLEMENTATION-VERSION. May want to tweak this some more, but it may make it easier to determine what you're running relative to what's in SVN/CVS. 2007-09-23 04:47 gb * [r7273] : Replace *OPENMCL-SUFFIX* (the thing which had the date in it) with *OPENMCL-SVN-REVISION*, which is defvar'ed to NIL here. (It may be set early in the cold load or via the xloader.) 2007-09-23 04:44 gb * [r7272] : Treat NIL and :UNSPECIFIC name/type components in the pattern consistently in %FILE*=. 2007-09-23 04:43 gb * [r7271] : LOCAL-SVN-REVISION tries to determine the (local) svn revision, by looking for a .svnrev file (used in the svn->cvs synching process) or by parsing "svn info" output. 2007-09-21 18:20 gb * [r7270] : 2007-09-21 18:01 jaj * [r7269] : Rename "cache" to "mirror" in hemlock-text-storage. This clarifies the purpose of the slot and avoids confusion with buffer caches, which are completely different. 2007-09-21 03:20 gb * [r7266] : 2007-09-21 02:12 palter * [r7265] : Make the pipe streams created by RUN-PROGRAM be basic streams 2007-09-19 21:20 gb * [r7264] : 2007-09-19 20:45 wws * [r7263] : Prevent unused gensym warnings in with-accessors. Same fix that Gary Palter did to with-slots. 2007-09-19 08:21 gb * [r7256] : 2007-09-19 07:52 gb * [r7255] : FRAME-REQUIRED-ARGS: try to make better. This is only used by SLIME; it was once used for frame restarting. Try to make the result reasonable in the SLIME case. 2007-09-19 07:44 gb * [r7254] : ARGLIST-FROM-MAP: don't be afraid to say &LEXPR. ARG-NAMES-FROM-MAP: &rest in map before &keys now. 2007-09-19 03:20 gb * [r7253] : 2007-09-19 02:19 palter * [r7252] : In WITH-SLOTS, if there are bindings, declare the gensym'ed instance symbol IGNORABLE just in case none of the symbol macros are actuall used. 2007-09-18 11:20 gb * [r7250] : 2007-09-18 10:21 gb * [r7249] : Typo in constant again. 2007-09-18 03:20 gb * [r7248] : 2007-09-18 02:17 palter * [r7247] : Return all three values from CCL::PRINT-DEFERRED-WARNINGS 2007-09-17 12:21 gb * [r7238] : 2007-09-17 11:53 gb * [r7237] : Simply clearing the top bit of the intermediate result from %NEXT-RANDOM-PAIR had the effect of using doing (mod result (expt 2 31)), not (mod result (1- (expt 2 31))). Do a real MOD to obtain the next item in the series and update the state with that value. (This means that the high bit of the high half of the result is always clear; or in the sign bit from the low half.) 2007-09-17 00:21 gb * [r7233] : 2007-09-16 23:21 gb * [r7232] : 2007-09-16 23:14 gb * [r7231] : X862-BIND-LAMBDA: bind &rest before &key ... 2007-09-16 23:04 gb * [r7230] : PPC2-BIND-LAMBDA: rest arg is in scope before key arg(s), so assign it an address before processing &key initforms. 2007-09-15 23:20 gb * [r7226] : 2007-09-15 23:02 gb * [r7225] : Observe *BACKTRACE-PRINT-LEVEL*, -LENGTH* when printing break messages. 2007-09-15 09:38 gb * [r7224] : Revert to older version. 2007-09-14 02:21 gb * [r7213] : 2007-09-14 01:32 palter * [r7211] : Add support for :element-type to ccl:run-program to allow for binary streams 2007-09-13 13:21 gb * [r7209] : 2007-09-13 12:40 gb * [r7208] : er, a mishap 2007-09-13 01:59 gb * [r7207] : (Botched last commit message. Just add comment; the following applies to r7206). Introduce X862-VPUSH-LABEL, which wraps stack memoization around VPUSH-LABEL vinsn. Replace (! start-mv-call) with calls to X862-VPUSH-LABEL. 2007-09-13 01:49 gb * [r7206] : Replace START-MV-CALL with VPUSH-LABEL; set vinsn attributes on new vinsn. 2007-09-11 08:20 gb * [r7188] : added 2007-09-11 08:18 gb * [r7187] : added 2007-09-09 08:35 mikel * [r7186] : removed currency converter example until OSX version legalities are resolved 2007-09-09 03:04 mikel * [r7185] : added more text and images to HOWTO 2007-09-09 02:20 mikel * [r7184] : updated links to use new name of HOWTO_files 2007-09-09 02:17 mikel * [r7183] : moved HOWTO files to HOWTO_files to unconfuse cvs 2007-09-08 01:45 mikel * [r7182] : added more instructions to the HOWTO, on how to create the IB project 2007-09-08 01:12 mikel * [r7181] : removed unneeded local directory 2007-09-08 01:10 mikel * [r7180] : added currency converter example to repo 2007-09-06 20:21 gb * [r7173] : 2007-09-06 19:45 mikel * [r7172] : added key 'CFBundleName' to write-info-plist 2007-09-05 05:21 gb * [r7143] : 2007-09-05 04:40 gb * [r7142] : New! Improved! The code which arranges for editing changes to be processed in the main thread now does so via an NSInvocation (ObjC's notion of a closure, sort of); the methods which run in the main thread handle the job of synchronizing the underlying the Hemlock string and the "real" NSMutableAttributedString as well as informing the text system of pending editing changes via #/edited:range:changeInLength:. The code which processes deletions does the change processing before modifying the string (other methods do it after the change); this seems to be necessary to avoid some things that could cause ticket:50. (I'm not sure if this fixes ticket:50, but it does seem to fix a reproducible case.) The #/replaceCharactersInRange:withString: method on hemlock-text-storage (which handles buffer changes originating from non-Hemlock sources) is careful to use a :right-inserting temporary mark when modifying the buffer, to help ensure that insertions/deletions take place at the correct absolute position. (This may at least partially address bugs related to things getting confused/out-of-bounds errors when doing lots of cutting and pasting, though it's hard to know without reproducible cases.) The macros NSSTRING-ENCODING-TO-NSINTEGER and NSINTEGER-TO-NSSTRING-ENCODING are defined before they're used (at least one of them wasn't.) The handling of text encoding for editor documents has changed in a few ways. If "automatic" encoding is used in an "Open ..." dialog and is successfully able to infer a text file's encoding, the encoding that was inferred becomes the document's encoding (used in subsequent "save" operations.) On some OSX versions, the automatic detection mechanism can only recognize UTF-16 files that contain a leading BOM (and the editor's code for dealing with the automatic mechanism didn't expect this to ever work.) On other systems, the automatic mechanism can return something correct (I've seen UTF-8) but possibly suboptimal. The Open... dialog now defaults to something derived from the user's setting of CCL:*DEFAULT-FILE-CHARACTER-ENCODING*; because "automatic" is a little unpredictable, the new default is probably better. The NSDocument mechanism for creating backup files (with names of the form "name~.ext") is not used; when backup files are created, we try to copy the original to a file whose name conforms to the (simple, non-versioned) Emacs scheme ("name.ext~"). More things seem to work better when the Emacs scheme is used (though I suppose some people might prefer the NSDocument scheme and we could make this a preference.) 2007-09-05 04:21 gb * [r7141] : 2007-09-05 03:54 gb * [r7140] : Don't call #/setFileName: when creating a Listener, since (a) that method's deprecated in favor of #/setFileURL: and (b) a newly-created Listener doesn't really have an associated filename, and so shouldn't get the icon and popup in its titlebar. Do arrange that (unless/until a listener document is saved to a file) its window title is based on the underlying Hemlock buffer's name. (If one ever attempts to save a listener document, the proposed file name is the internal NSDocument name; we might or might not want to change that.) 2007-09-05 03:44 gb * [r7139] : Suppress warnings about *MODELINE-FONT-NAME*, -SIZE*. 2007-09-05 01:21 gb * [r7138] : 2007-09-05 01:08 gb * [r7137] : Lose "HeapHighWaterMark", which was supposed to allow us to track the high end of a range of pages that'd been written to and not freed (so we'd zero them lazily when they were reallocated.) Such a scheme would really need to track both ends of such a range, and the old scheme wound up being overly zealous and often zeroed pages that were already zeroed. Zero the range between the old free pointer and the new one after each GC, instead. At least partly address ticket:101, by doing GROW-DYNAMIC-AREA more carefully. 2007-09-05 00:58 gb * [r7136] : Fix to 64-bit %NEXT-RANDOM-SEED, random algorithm comments from Bernd Beuster. 2007-09-05 00:37 gb * [r7135] : Don't coerce single-float results to double-floats in DEFCALLBACK-BODY. Do do this in the PPC backends. (This needs testing ...) 2007-09-05 00:36 gb * [r7134] : Don't assume that :single-float callback results have been coerced to :double-float on x8664. 2007-09-04 00:21 gb * [r7130] : 2007-09-03 23:55 gz * [r7129] : Make meta-. use either the selection or the symbol around point, regardless of global parsing state (Ticket:95). While in there, make meta-. with a numarg prompt for symbol to look up. 2007-08-31 02:21 gb * [r7128] : 2007-08-31 01:21 gb * [r7127] : 2007-08-31 01:20 gz * [r7126] : Add a restart for unbound variable/undefined function errors -- if there is a unique bound/fbound symbol of the same name in another package, offer to use it instead. 2007-08-31 00:21 gb * [r7125] : 2007-08-31 00:11 gb * [r7124] : Tell grep not to complain about inaccesible files; it might return an exit status of 2 if it finds such files, so treat that as "success" for now (ticket:98). 2007-08-31 00:09 gb * [r7123] : Fix format call in Show Variable command output (ticket:97). 2007-08-31 00:08 gb * [r7122] : EVAL-REGION uses VARIABLE-VALUE to access buffer's package (ticket:82). 2007-08-30 21:20 gb * [r7121] : 2007-08-30 21:03 mikel * [r7120] : added code to copy user-supplied nibfiles into a saved application bundle 2007-08-30 20:21 gb * [r7119] : 2007-08-30 19:33 gz * [r7118] : Make sure all callers of current-point-for-deletion handle the nil return value case. This fixes ticket:85 2007-08-30 02:21 gb * [r7117] : 2007-08-30 01:52 gz * [r7116] : umm, load cocoa-grep where it will work... 2007-08-30 01:45 gz * [r7113] : load cocoa-grep 2007-08-29 19:21 gb * [r7111] : 2007-08-29 18:27 gz * [r7110] : Implement m-x grep 2007-08-29 18:20 gb * [r7109] : 2007-08-29 17:47 gb * [r7108] : Indent command: if no prefix arg (simple case), move point to the location of the temporary mark after calling the indent function. 2007-08-29 17:46 gb * [r7107] : COUNT-LEADING-WHITESPACE returns 2 values: the "column" (with tabs expanded) of the first non-whitespace character on the line and that character's position. ENSURE-LISP-INDENTATION moves the (temporary) mark to that position if no changes are necessary. 2007-08-29 17:43 gb * [r7106] : In HEMLOCK-TEXT-VIEW #/paste: method, make an NS-MUTABLE-STRING even if the string from the clipboard already appears to be an NS-MUTABLE-STRING. (NSCFString is apparently a subclass of NS-MUTABLE-STRING, but isn't really ... mutable.) For ticket:84. 2007-08-29 05:20 gb * [r7105] : 2007-08-29 04:12 gz * [r7104] : Make hi::scroll-window actually scroll, not just move insertion point around. Fixes ticket:88. 2007-08-26 19:21 gb * [r7099] : 2007-08-26 18:43 mikel * [r7098] : new keyword argument to build-application: main-nib-name; names the nigfile that is to be the main nib of the built application 2007-08-26 18:20 gb * [r7097] : 2007-08-26 17:44 mikel * [r7096] : added application bulder to trunk 2007-08-25 08:20 gb * [r7095] : 2007-08-25 07:55 gb * [r7094] : HI::SEND-STRING-TO-LISTENER-PROCESS quotes #\^d, sends unquoted #\^d after sending the string. (This is just a workaround for ticket:82; the trailing #\^d is just treated as whitespace and terminates calls to READ. We really want some way of recognizing "end of selection" and treating that like a transient EOF in some cases.) 2007-08-25 07:43 gb * [r7093] : In COCOA-EDIT, invoke #/openDocumentWithContentsOfURL:display:error:, since calling #/openDocumentWithContentsOfURL:display: complains that #/dataRepresentationOfType: isn't implemented in 32-bit versions of Cocoa. (Note that we'd been calling #/openDocumentWithContentsOfURL:display:error: as part of the m-. implementation without incident.) 2007-08-25 07:20 gb * [r7092] : "Capitalize Word" command notes buffer modification. 2007-08-25 07:19 gb * [r7091] : FILTER-REGION notes buffer modification. 2007-08-23 23:20 gb * [r7088] : 2007-08-23 22:26 gb * [r7087] : Remove a duplicate #/continue: method. 2007-08-23 22:26 gb * [r7086] : Remove a duplicate #/editingInProgress method. Add some more (conditionalized) debug info. If ticket:50 bug is encountered, try to force DBG on Hemlock command thread as well as in Cocoa thread. Make sure that Hemlock command thread has bindings of command-processing data structures, prompt data structures. Do not bind *LAST-SEARCH-STRING* or *LAST-SEARCH-PATTERN*; note that *LAST-SEARCH-STRING* is usually only set by isearch when it exits. 2007-08-23 22:20 gb * [r7085] : 2007-08-23 22:20 gb * [r7084] : Use new defaults-based backtrace font name and size; initially, Monaco 9. (Not yet changeable through preferences panel.) Adjust outline view's row height based on font height. Set the data cell to "non-editable" (could have done this in IB). Provide handlers for some menu commands (continue, restarts, exitBreak); no need to validate menu items for them (those commands should be available whenever backtrace dialog is active.) 2007-08-23 22:14 gb * [r7083] : Use RECURSIVE-GET-KEY-EVENT in isearch, query-replace. 2007-08-23 22:13 gb * [r7082] : Export RECURSIVE-GET-KEY-EVENT. 2007-08-23 22:13 gb * [r7081] : Ignore erorrs when trying to read symbol for arglist. (Should probably be more careful elsewhere, too.) 2007-08-23 22:12 gb * [r7080] : Keep track of *COMMAND-KEY-EVENT-BUFFER* in %COMMAND-LOOP, to enable the RECURSIVE-GET-KEY-EVENT redisplay mechanism. 2007-08-23 22:10 gb * [r7079] : Don't call DOCUMENT-SET-POINT-POSITION in CLEAR-ECHO-AREA: editing changes may still be pending on the echo area (how ?), and DOCUMENT-SET-POINT-POSITION may need to do layout/display updates. Use RECURSIVE-GET-KEY-EVENT when prompting in the echo area. 2007-08-23 22:07 gb * [r7078] : Define RECURSIVE-GET-KEY-EVENT, to allow redisplay on the buffer designated as *COMMAND-KEY-EVENT-BUFFER* in while getting key events in some modal loop in a command. (The idea is that toplevel editing commands inhibit redisplay in the main buffer while they're active; commands which recursively process key events and which may make changes to that buffer should allow redisplay while they're waiting in RECURSIVE-GET-KEY-EVENT.) 2007-08-23 22:02 gb * [r7077] : Use the NSDocument-based revert. 2007-08-19 01:21 gb * [r7070] : 2007-08-19 00:31 gb * [r7069] : Define and use %GF-DISPATCH-TABLE-STORE-CONDITIONAL; this is intended to avoid race conditions where two threads try to update a distpatch table index with different {wrapper, combined-method} pairs at the same time. (As it's written, if the store-conditional fails because a slot that we thought was free is now in use, we simply don't cache the pair and get a cache miss the next time; that may or may not be more work than repeating the whole process would be.) The store-conditional is used on attempt to update the combined-method slot in the pair; the wrapper slot is only updated if the store-conditional succeeds. Code which probes the wrapper slots shouldn't be confused by a half-updated pair (should never see a non-null wrapper slot and a null combined-method slot.) 2007-08-19 00:29 gb * [r7068] : Define and use %GF-DISPATCH-TABLE-STORE-CONDITIONAL; this is intended to avoid race conditions where two threads try to update a distpatch table index with different {wrapper, combined-method} pairs at the same time. (As it's written, if the store-conditional fails because a slot that we thought was free is now in use, we simply don't cache the pair and get a cache miss the next time; that may or may not be more work than repeating the whole process would be.) The store-conditional is used on attempt to update the combined-method slot in the pair; the wrapper slot is only updated if the store-conditional succeeds. Code which probes the wrapper slots shouldn't be confused by a half-updated pair (should never see a non-null wrapper slot and a null combined-method slot.) 2007-08-18 23:21 gb * [r7067] : 2007-08-18 22:21 gb * [r7066] : 2007-08-18 22:11 gb * [r7065] : Don't treat (:* :void) as (:* nil) in %DECODE-TYPE. 2007-08-18 22:05 gb * [r7064] : Use #/instancesRespondToSelector: to determine if selection highlighting can be used, since #_class_respondsToSelector requires ObjC-2.0) 2007-08-16 15:21 gb * [r7060] : 2007-08-16 14:22 gb * [r7059] : 2007-08-16 13:36 gb * [r7058] : Hemlock text storage can use extended selection highlighting if it's supported. Clear the shift bit from key events associated with standard-chars. 2007-08-16 13:32 gb * [r7057] : Shifted and unshifted keys bound to movement commands run different commands. Left-mouse has the :mouse-exit logical attribute. 2007-08-16 13:31 gb * [r7056] : CURRENT-POINT-COLLAPSING-SELECTION and CURRENT-POINT-EXTENDING-SELECTION replace CURRENT-POINT-FOR-MOVEMENT. 2007-08-16 13:29 gb * [r7055] : Simplify key-event translation. Define NOTE-SELECTION-SET-BY-SEARCH, which can cause special selection highlighting under Leopard. 2007-08-16 13:28 gb * [r7054] : Movement commands come in "shifted" and "unshifted" variants. 2007-08-16 13:27 gb * [r7053] : :mouse-exit is a new logical key attribute. 2007-08-16 13:26 gb * [r7052] : GET-KEY-EVENT*: clear shift bit for standard-char (and upper-case.) This means that #k"ctrl-shift-a" and #k"ctrl-A" both refer to the same key event (in which the shift modifier isn't explicitly set) and that #k"rightarrow" and #k"shift-rightarrow" are distinct (and the shift modifier is set in the latter.) 2007-08-16 13:23 gb * [r7051] : When indenting, don't change the buffer if the line's already indented properly. Movement commands come in two variants: those that collapse and those that extend the selection. 2007-08-16 13:21 gb * [r7050] : Export CURRENT-POINT-EXTENDING-SELECTION and CURRENT-POINT-COLLAPSING-SELECTION. Don't export CURRENT-POINT-FOR-MOVEMENT. 2007-08-16 13:21 gb * [r7049] : 2007-08-16 13:21 gb * [r7048] : Hilite the selection more empatically (under Leopard) after successful (non incremental) search. Incremental search also hilites the selection, but doesn't yet manage the selection on failure. 2007-08-16 12:42 gb * [r7047] : activate-hemlock-view: implemenent by invoking #/activateHemlockView on main thread; call deactivate-hemlock-view on the peer. Not sure if edit-count fuss in deactivate-hemlock-view is correct, but I'm fairly sure that this stuff needs to happen on the main thread. 2007-08-09 22:21 gb * [r7020] : 2007-08-09 21:19 gb * [r7019] : Use PRINC-TO-STRING (not PRIN1-TO-STRING) in ARGLIST-STRING, to avoid prinding package qualfiers (ticket:30). 2007-08-07 21:21 gb * [r7014] : 2007-08-07 20:40 jaj * [r7013] : Bind home and end to beginning of buffer and end of buffer. It's Mac standard, and I use it a lot. Home was bound to "Help". Make clicking the mouse while incremental searching exit instead of abort, leaving the insertion point where you click instead of where the isearch was started. 2007-08-07 05:21 gb * [r7008] : 2007-08-07 04:44 gb * [r7007] : When generating a hemlock key event, set the hemlock shift bit if the the Cocoa shift bit was set. In HI::GET-KEY-EVENT, return an event with the shift bit cleared and the keysym downcased (if necessary). Leave HI::*LAST-KEY-EVENT-TYPED* as the unmodified event (which may have shift bits set.) Define HI:CURRENT-POINT-FOR-MOVEMENT, which deactivates the region ("collapses the selection") if the shift bit was clear in HI::*LAST-KEY-EVENT-TYPED* and ensures that a region is active otherwise. Use HI:CURRENT-POINT-FOR-MOVEMENT in (most, hopefully all) movement commands, to address ticket:36 . Define a "Do Nothing" hemlock command, bind #k"leftdown" (left mouse down) to it to address ticket:44 . 2007-08-07 01:21 gb * [r7006] : 2007-08-07 00:17 gb * [r7005] : careful with db stuff 2007-08-04 19:21 gb * [r7004] : 2007-08-04 18:21 gb * [r7003] : 2007-08-04 18:12 gb * [r7002] : Conditionalize an #_NSLog call. Post a #k"leftdown" to Hemlock on mouseDown: in the main text view. (Should maybe do this on mousedown in the echo area, too.) Some indentation changes. 2007-08-04 17:59 gb * [r7001] : Make #k"leftdown" a logical :abort key event. 2007-08-04 17:59 gb * [r7000] : Fix format string in window title for show lossage command. 2007-08-04 17:58 gb * [r6999] : Define the keysym for left mouse down, so that we can use #k"leftdown". 2007-08-04 17:58 gb * [r6998] : Revive mouse-event stuff. 2007-08-03 21:21 gb * [r6997] : 2007-08-03 20:33 rme * [r6996] : Correct spelling errors. 2007-08-03 16:21 gb * [r6994] : 2007-08-03 15:21 gb * [r6993] : 2007-08-03 15:10 gb * [r6992] : Use list-to-vector, not (apply #'vector ...), since the latter can run into problems with CALL-ARGUMENTS-LIMIT. 2007-08-03 15:09 gb * [r6991] : Define list-to-vector. 2007-08-02 13:21 gb * [r6984] : 2007-08-02 12:55 gb * [r6983] : Don't make font panel broadcast #/changeFont:; use targeted messages. Ensure that font panel picks up the right font. 2007-08-02 12:54 gb * [r6982] : Don't enable "metal" style by default. 2007-08-02 12:53 gb * [r6981] : Make "peer" text view (main text view, echo area) unselectable when activating via Hemlock command. Text views don't use font panel by default; enable/disable font panel when setting text color. 2007-08-02 12:50 gb * [r6980] : New. 2007-08-02 12:47 gb * [r6979] : Define and export CREATE-INTERFACES. 2007-08-02 12:46 gb * [r6978] : Don't need carbon interfaces. 2007-08-02 12:46 gb * [r6977] : Don't be verbose about explict struct returns. 2007-08-02 12:45 gb * [r6976] : Don't use carbon interfaces anymore. 2007-08-01 01:20 gb * [r6974] : 2007-08-01 00:12 gb * [r6973] : Added. 2007-07-31 13:06 gb * [r6972] : 2007-07-31 12:13 gb * [r6971] : Default the :MODE option to SAVE-APPLICATION to #o644 (not #o666); make sure that it's actually used when creating the image (even if not prepending a kernel to the image.) 2007-07-31 12:10 gb * [r6970] : When processing --version option, don't write to *terminal-io* (which doesn't exist yet.) 2007-07-29 23:27 gb * [r6969] : set executable property on 2007-07-23 05:39 gb * [r6962] : 2007-07-23 05:25 gb * [r6961] : Add more stuff. 2007-07-23 05:07 gb * [r6960] : 2007-07-23 04:24 gb * [r6959] : Bump. 2007-07-23 04:23 gb * [r6958] : Update for 1.1-pre-070722. 2007-07-23 03:07 gb * [r6957] : 2007-07-23 02:23 gb * [r6956] : Export WITH-ENCODED-CSTRS. 2007-07-23 02:22 gb * [r6955] : New file. 2007-07-23 01:00 gb * [r6954] : wasn't used 2007-07-23 00:39 gb * [r6953] : 2007-07-21 22:57 gb * [r6952] : Hook name is in inspector package. 2007-07-21 22:52 gb * [r6950] : Put special args/locals in symbol map. 2007-07-21 22:52 gb * [r6949] : Bump. 2007-07-21 22:51 gb * [r6948] : Allocate callback trampolines via mmap on ppc, too. 2007-07-21 22:50 gb * [r6947] : Don't expose the PROGRAM arg to RUN-PROGRAM to merge-pathnames. 2007-07-21 22:49 gb * [r6946] : Don't define BEEP. 2007-07-21 22:49 gb * [r6945] : utf-8 length-of-memory-encoding function was counting chars (nchars) twice. 2007-07-21 22:48 gb * [r6944] : selection-stream changes. 2007-07-21 22:47 gb * [r6943] : :raw breakloop command. Put info about restarts in backtrace context. 2007-07-21 22:46 gb * [r6942] : PROCESS-DEBUG-CONDITION, so that it can be overridden. 2007-07-21 22:45 gb * [r6941] : Add objc-bridge, cocoa-ide to *MODULE-SEARCH-PATH*. 2007-07-21 22:45 gb * [r6940] : Handling of * format directive when dealing with circularity screws up other things that expect *FORMAT-ARGUMENTS* to share structure with *FORMAT-ORIGINAL-ARGUMENTS*. 2007-07-21 22:43 gb * [r6939] : NATIVE-UNTRANSLATED-NAMESTRING, ENSURE-DIRECTORY-NAMESTRING. 2007-07-21 22:42 gb * [r6938] : Handle xcf's on x86 when processing an interactive break. 2007-07-21 22:41 gb * [r6937] : Stream error context stuff. 2007-07-21 22:40 gb * [r6936] : Try to recognize foreign exceptions (e.g., as NSExceptions.) 2007-07-21 22:39 gb * [r6935] : Use COLLECT macro in CLASS-SLOT-INITARGS. 2007-07-21 22:38 gb * [r6934] : Move the x86-64 callback trampoline allocation stuff (which uses mmap to ensure that the callback is in executable memory) here, since it'll be needed on ppc64 leopard, at least. 2007-07-21 22:37 gb * [r6933] : Argument handling and interactive stream initialization happens in a :before method on (toplevel-function application). 2007-07-21 22:34 gb * [r6932] : Symbol-address recording stuff: record special args/locals, too. 2007-07-21 22:33 gb * [r6931] : Support for disassemble-list, for backtrace/inspect someday. 2007-07-21 22:32 gb * [r6930] : Globals for ObjC2 exception stuff. 2007-07-21 22:30 gb * [r6929] : Callbacks that try to do magic foreign error stuff use HANDLER-BIND, not HANDLER-CASE. Other callback-changes to reduce incidental consing sometimes. 2007-07-21 22:28 gb * [r6928] : %show-args-and-locals, %stack-frames-in-context: changes for new backtrace presentation. 2007-07-21 22:26 gb * [r6927] : Remove some duplicate definitions. 2007-07-21 22:23 gb * [r6926] : Some (signigicant) stack-frame changes. 2007-07-21 22:21 gb * [r6925] : Tighten some things up a bit. 2007-07-21 22:20 gb * [r6924] : Provide -some- arglist info for special operators. ARG-NAMES-FROM-MAP. 2007-07-21 22:19 gb * [r6923] : Enter the 1990s (eval-when ..). 2007-07-21 22:19 gb * [r6922] : Comment formatting. 2007-07-21 22:18 gb * [r6921] : Use hash tables in #=, etc. Expect reader errors to provide context. 2007-07-21 22:17 gb * [r6920] : Tighten up some frame accessors; don't get confused by null register values, etc. 2007-07-21 22:15 gb * [r6919] : GET-TIMEZONE: conditionalize on target word size, not ppc32/ppc64. 2007-07-21 22:13 gb * [r6918] : Try to defer GC locking in GETHASH. Some 64-bit key-scrambling stuff. 2007-07-21 22:13 gb * [r6917] : Less zeal in SAFE-GET-PTR. 2007-07-21 22:12 gb * [r6916] : s32->u32, u32->s32. 2007-07-21 22:12 gb * [r6915] : foreign exception error stuff. 2007-07-21 22:11 gb * [r6914] : Sockets created with :format :binary don't have element-type of character, though this crap is all pretty nebulous. 2007-07-21 22:03 gb * [r6913] : $XFOREIGNEXCEPTION. 2007-07-21 22:03 gb * [r6912] : set-pty-raw. 2007-07-21 22:03 gb * [r6911] : Moved elsewhere. 2007-07-21 21:58 gb * [r6910] : Basically, all frames on the vstack are lisp frames. 2007-07-21 21:57 gb * [r6909] : Add ObjC2 exception metainfomation. 2007-07-21 21:56 gb * [r6908] : Clear and restore the tcr foreign exception bit on interrupt. 2007-07-21 21:55 gb * [r6907] : tcr.flags is a full word. define tcr flag bits, including new one for foreign exceptions. 2007-07-21 21:54 gb * [r6906] : Globals for ObjC2 exceptions. 2007-07-21 21:53 gb * [r6905] : New tcr flag bit for foreign exceptions. 2007-07-21 21:53 gb * [r6904] : ensure_stack_limit() when creatings stack for pthread. Foreign exception callbacks: put some 0s on the stack for callback. 2007-07-21 21:52 gb * [r6903] : Restore save regs on callbacks. 2007-07-21 21:51 gb * [r6902] : usage_exit() via _exit(). ensure_stack_limit(): if stack size is constrained by limits, don't exceed those limits. 2007-07-21 21:50 gb * [r6901] : globals for ObjC2 exceptions. 2007-07-21 21:49 gb * [r6900] : prototype for ensure_stack_limit(). 2007-07-21 21:49 gb * [r6899] : XFOREIGNEXCEPTION error constant. 2007-07-21 21:48 gb * [r6898] : New directory, with bridge stuff from examples. 2007-07-21 21:47 gb * [r6897] : RECORD-SOURCE-FILE for compiler-macros, too. 2007-07-21 21:46 gb * [r6896] : moved elsewhere, and changed a lot 2007-07-21 21:45 gb * [r6895] : Moved elsewhere, and changed a lot. 2007-07-21 21:44 gb * [r6894] : New directory. 2007-07-21 21:43 gb * [r6893] : moved elsewhere, and changed a lot 2007-06-16 01:11 gb * [r6740] : 2007-06-16 00:57 gb * [r6739] : use hash tables in #=, ##. Sometimes 2007-06-15 20:11 gb * [r6738] : 2007-06-15 19:20 gb * [r6737] : x862-numcmp: if both args fixnums ... x862-inline-numcmp: don't reverse cr-bit if both args fixnums (shouldn't get here if we're catching that upstream, but ...) 2007-06-01 14:27 gb * [r6640] : 2007-06-01 14:13 gb * [r6639] : 2007-06-01 00:43 gb * [r6638] : Do ~-expansion in more places. Hopefully caught most places. 2007-06-01 00:42 gb * [r6637] : %TOKEN-PACKAGE generates a READER-ERROR if it can't find the package, accepts a stream argument in order to signal that type of error. (Note that this has never been recoverable in general, especially in the case where the input stream is *terminal-io* or shares its input stream.) 2007-06-01 00:40 gb * [r6636] : READER-ERORS use STREAM-ERROR-CONTEXT when reporting themselves. 2007-06-01 00:39 gb * [r6635] : Define and use STREAM-ERROR-CONTEXT. 2007-06-01 00:09 gb * [r6634] : Define UID-FROM-NAME, TILDE-EXPAND. Use TILDE-EXPAND in %REALPATH, but any leading tildes should probably have been caught earlier. 2007-06-01 00:06 gb * [r6633] : Try to provide some context - "surrounding characters" - for reader/stream errors on character input streams. 2007-05-20 08:05 gb * [r6576] : Get number of args right in call to STRING-ENCODED-LENGTH-IN-BYTES from CSTRING-ENCODED-LENGTH-IN-BYTES. Sorry. 2007-05-19 20:53 gb * [r6568] : %%get-unsigned-longlong, %%get-signed-longlong: args in UNBOX-FIXNUM macro were in the wrong order. Both functions are usually open-coded (and the inlined versions seem to be OK), but the ObjC bridge funcalls these functions when getting/setting foreign slots of the appropriate type. 2007-05-16 08:33 gb * [r6560] : Add STRING-ENCODED-LENGTH-IN-BYTES. Make it, CSTRING-ENCODED-LENGTH-IN-BYTES, and ENCODE-STRING-FROM-MEMORY deal with non-simple strings. 2007-05-16 07:31 gb * [r6559] : Try to deal with the "SIMPLE-BASE-STRING, all of it" case in %CSTR-POINTER; deal with all other cases in %CSTR-SEGMENT-POINTER. 2007-05-16 07:25 gb * [r6558] : xp-argument-list: stack args are under return address, exception callback frame (xcf) on stack. handle-udf-call: don't try to fix up the stack if we try to continue; let kernel deal with this after the callback. 2007-05-16 07:23 gb * [r6557] : finish_function_entry: there's a return address on the top of the stack when this is entered (#-of-arguments traps). Put it in the right place when building the frame. create_exception_callback_frame: look for tra on top of lisp stack, not in register (usually.) handle_error(): if we return after calling out for int #xc7 (call to undefined function), fix up the stack and other registers so that we can return the value(s) returned by the lisp-side handler. (It's hard for the lisp-side handler to do the stack manipulation, because of the xcf on the stack and because of the need to deal with the return address on the stack.) 2007-05-16 07:16 gb * [r6556] : uuos in subrprims called via lea/jmp: push ra0 on stack before trap, for the benefit of create_exception_callback_frame(). 2007-05-16 07:11 gb * [r6555] : Eric Marsden's fixes to MISMATCH 2007-05-15 00:49 gb * [r6554] : :utf-32 memory-encode function: don't write a BOM here. 2007-05-14 18:37 gb * [r6553] : %RETURN-ADDRESS-OFFSET has been wrong since the switch to the new tra scheme. 2007-05-14 18:37 gb * [r6552] : WITH-ENCODED-CSTR: accept arguments as produced by WITH-ENCODED-CSTRS. Handle 0-termination a little more sanely. 2007-05-13 00:22 gb * [r6548] : 2007-05-12 23:02 gb * [r6547] : Update for 070512. 2007-05-12 21:20 gb * [r6546] : Fix typo in new_heap_segment(). 2007-05-12 21:18 gb * [r6545] : Fix typo in ppc2-%i- 2007-05-12 21:16 gb * [r6544] : Bump. 2007-05-12 20:26 gb * [r6543] : new_heap_segment(): don't extend heap by more than what we need if GC is obviously disabled. 2007-05-12 20:25 gb * [r6542] : Use *ppc2-target-half-fixnum-type* to suppress overflow checking in some cases. 2007-05-12 20:24 gb * [r6541] : Fix indentation. 2007-05-12 12:56 gb * [r6540] : new_heap_segment(): don't extend beyond segment if GC (obviously) disabled. 2007-05-09 12:00 gb * [r6539] : Fix typo. 2007-05-09 11:58 gb * [r6538] : Keep bootsrapping separate character encode/decode limits. 2007-05-09 11:39 gb * [r6537] : Start to bootstrap separate encode/decode limits. 2007-05-09 11:17 gb * [r6536] : Hide Darwin/Leopard exception context nonsense when referencing mxcsr in debug_show_fpu(). 2007-05-09 11:11 gb * [r6535] : Need alias for __r6 in Tiger and earlier. Sheesh. 2007-05-09 09:10 gb * [r6534] : Fix bad call to %pname-hash. 2007-05-09 09:07 gb * [r6533] : Bump, for now. 2007-05-09 09:04 gb * [r6532] : Bump fasl versions, image ABI version. 2007-05-09 09:03 gb * [r6531] : nmkunwind. New calling sequence; new TRA interpretation. 2007-05-09 09:02 gb * [r6530] : nmkunwind. 2007-05-09 09:02 gb * [r6529] : Make_Catch uses %imm2, not %temp2. nMake_Catch handles return addresses on stack. repret, for use when a one-byte ret instruction is a branch target or is preceded by a conditional branch. 2007-05-09 09:00 gb * [r6528] : New TRA handling; new node regs in exception contexts. 2007-05-09 09:00 gb * [r6527] : New stuff for finding FP state in exception context; new TRA handling. 2007-05-09 08:59 gb * [r6526] : New tra handling. new_heap_segment: try to avoid requesting too much when user sets lisp_heap_gc_threshold to inappropriate value. Pass old_valence down to fault handler, don't call out to lisp on faults in foreign code. Handle USE_MACH_EXCEPTION_LOCK conditionalization. 2007-05-09 08:55 gb * [r6525] : New register assignments. 2007-05-09 08:55 gb * [r6524] : New register assignments. 2007-05-09 08:50 gb * [r6523] : New tra handling, new calling sequence. 2007-05-09 08:50 gb * [r6522] : New tra handling. 2007-05-09 08:49 gb * [r6521] : New tra handling. 2007-05-09 08:49 gb * [r6520] : Use repret for RET instructions that're branch targets or that follow conditional branches. 2007-05-09 08:48 gb * [r6519] : Conditionalize for USE_MACH_EXCEPTION_LOCK. 2007-05-09 08:47 gb * [r6518] : .SPnmkunwind. 2007-05-09 08:47 gb * [r6517] : .SPnmkunwind. Old catch/unwind: do explicit blr. 2007-05-09 08:46 gb * [r6516] : Use -Wno-deprecated-declarations. Gee, I hope that works. 2007-05-09 08:45 gb * [r6515] : Make_Catch macro: don't include a blr. 2007-05-09 08:42 gb * [r6514] : Only declare mach_exception_lock if using the mechanism. 2007-05-09 08:41 gb * [r6513] : new_heap_segment: if we can't extend with (possibly very large) lisp_heap_gc_threshold, try smaller values. Pass "old_valence" to exception handlers; don't call out to lisp on memory faults from foreign code. 2007-05-09 08:39 gb * [r6512] : xmm register alias nonsense. 2007-05-09 08:39 gb * [r6511] : Change some function prototypes. Don't use a MACH_EXCEPTION_LOCK on Darwin, since it doesn't seem to be correct/necessary. 2007-05-09 08:37 gb * [r6510] : lisp register display changes for x86-64. Get FPU display working on x86-64. 2007-05-09 08:36 gb * [r6509] : gtra, just in time for the calling-sequence change. 2007-05-09 08:35 gb * [r6508] : New ABI version. 2007-05-09 08:35 gb * [r6507] : Ignore SIGSYS. If only I could; if only I could ... 2007-05-09 08:34 gb * [r6506] : More imm temps! (one more ...). 2007-05-09 08:34 gb * [r6505] : Bump current fasl version. 2007-05-09 08:33 gb * [r6504] : RLET, MAKE-RECORD, etc: don't set pointer type in cross-compiled code. 2007-05-09 08:33 gb * [r6503] : Don't use #. to initialize *host-ftd* interface dir, because of cross-compilation issues. 2007-05-09 08:32 gb * [r6502] : Support "queries" in FFI reader macros: #_?foo, #$?foo, and #&?foo return T if foo is defined, NIL otherwise. 2007-05-09 08:30 gb * [r6501] : Lose some stuff leftover from 68k->PPC MCL transition. 2007-05-09 08:30 gb * [r6500] : Mapped-vector stuff. 2007-05-09 08:29 gb * [r6499] : Add a comment. 2007-05-09 08:28 gb * [r6498] : Very small lap change. 2007-05-09 08:28 gb * [r6497] : Move %FIXNUM-REF-MACPTR and %FIXNUM-SET-MACPTR elsewhere. 2007-05-09 08:27 gb * [r6496] : File-mapped vectors. 2007-05-09 07:57 gb * [r6495] : OPEN : :if-exists defaults to :new-version if pathname-version of filename is :NEWEST. 2007-05-09 07:56 gb * [r6494] : %BREAK-IN-FRAME might be called with a CONDITION object. 2007-05-09 07:56 gb * [r6493] : *HOST-PAGE-SIZE* initialized earlier. 2007-05-09 07:55 gb * [r6492] : If there's no host component in a namestring argument, LOGICAL-PATHNAME signals a TYPE-ERROR. 2007-05-09 07:54 gb * [r6491] : Initial-lisp-thread, initial listener stack sizes now static. 2007-05-09 07:54 gb * [r6490] : Pathname arcana. 2007-05-09 07:53 gb * [r6489] : Optionally, invoke debugger-hook on SIGINT. 2007-05-09 07:52 gb * [r6488] : Clear dispatch table cache if GF APO changes. 2007-05-09 07:51 gb * [r6487] : Don't cons when clearing spin lock. Commented-out (#+bad-idea) GC-lock-via-exception-lock idea. 2007-05-09 07:49 gb * [r6486] : Use %pname-hash to hash ... pnames. 2007-05-09 07:45 gb * [r6485] : %FIXNUM-REF-MACPTR, %FIXNUM-SET-MACPTR: move here. 2007-05-09 07:45 gb * [r6484] : Remove an old declaration. 2007-05-09 07:44 gb * [r6483] : New calling sequence. 2007-05-09 07:44 gb * [r6482] : New calling sequence. Use a CMOV in EQL. 2007-05-09 07:43 gb * [r6481] : New calling sequence. No need to be careful with %rcx (now %imm2) in shift instructions. 2007-05-09 07:42 gb * [r6480] : New calling sequence. 2007-05-09 07:42 gb * [r6479] : New calling sequence. 2007-05-09 07:42 gb * [r6478] : Handle new-style tras, other new-calling-sequence changes. 2007-05-09 07:41 gb * [r6477] : New calling sequence. %rcx is %imm2 now, not %temp2, and no need to be so careful with it. 2007-05-09 07:41 gb * [r6476] : New calling sequence. %rcx is %imm2 now, not %temp2, and no need to be so careful with it. 2007-05-09 07:40 gb * [r6475] : New calling sequence. %rcx is %imm2 now, not %temp2, and no need to be so careful with it. 2007-05-09 07:39 gb * [r6474] : global label-valued temporaries. 2007-05-09 07:38 gb * [r6473] : FLOAT with no protototype argument. 2007-05-09 07:38 gb * [r6472] : Try to do a little type inference for more flavors of AREF. 2007-05-09 07:37 gb * [r6471] : FUNCTION-INFORMATION might return an afunc (or equivalent) as a non-nil second value. 2007-05-09 07:36 gb * [r6470] : Support the new (call/ret) calling sequence, new tra/talign scheme. One more imm reg, so 3d aset is less nasty. 2007-05-09 07:35 gb * [r6469] : Support the new (call/ret) calling sequence, new tra/talign scheme. 2007-05-09 07:34 gb * [r6468] : Support :talign, (@ (:^ label) (% rip)). 2007-05-09 07:34 gb * [r6467] : Recognize CALL label operands, rip-relative label refs. 2007-05-09 07:33 gb * [r6466] : Allow label-valued temporaries, support :talign directive in vinsn templates. 2007-05-09 07:32 gb * [r6465] : Add call instruction; note that %rip is a register ... 2007-05-09 07:31 gb * [r6464] : Use "tail-aligned" calls for most things, except calls to subprims that manipulate the stack. 2007-05-09 07:30 gb * [r6463] : Try to make cross-compilation with x86-64 targets a little easier, by using a cross-ftd interface directory. 2007-05-09 07:29 gb * [r6462] : move away from lea/jmp, towards call/ret. rcx is now an imm reg; ra0 = temp2. new tra scheme. 2007-05-09 07:27 gb * [r6461] : %immediate-set-xxx: get sign right. unwind-protect uses nmkunwind (more happens out-of-line.) 2007-05-09 07:26 gb * [r6460] : nmkunwind. 2007-05-09 07:26 gb * [r6459] : getu32/gets32, nmkunwind. 2007-05-09 07:25 gb * [r6458] : New .SPnmkunwind. 2007-05-09 07:24 gb * [r6457] : Make %GET-KERNEL-GLOBAL-PTR actually work. 2007-05-09 07:24 gb * [r6456] : Make %GET-KERNEL-GLOBAL-PTR actually work. 2007-05-08 01:15 gb * [r6438] : 2007-05-08 00:28 gb * [r6432] : ELT, SET-ELT: don't assume that NTHCDR returns a CONS, signal the right flavor of TYPE-ERROR if it doesn't. 2007-05-06 00:15 gb * [r6429] : 2007-05-05 23:24 gb * [r6428] : If PUTHASH decides that it wants to grow the hash table, it tries to enable the GC before doing so. 2007-05-05 04:15 gb * [r6427] : 2007-05-05 03:45 gb * [r6426] : FUNCTION-INFORMATION returns what it finds in the environment, not just non-nil. 2007-05-02 00:15 gb * [r6394] : 2007-05-01 23:56 gb * [r6393] : Fix the "bytes allocated" message and computation for GC-VERBOSE entry. 2007-05-01 12:15 gb * [r6392] : 2007-05-01 11:31 gb * [r6391] : Set literal-code-limit to #xd800 for utf-16 variants. Fix the typo (found by Takekiko Abe) in the utf-8 stream-decode function. 2007-05-01 11:27 gb * [r6390] : Mis-align some fields in the darwin ppc64 thread state structure, just like Apple does. Can I blame them for getting this wrong, please ? Use the right argument names in ppc64 cmain callback; can't blame Apple. 2007-05-01 05:15 gb * [r6389] : 2007-05-01 04:18 gb * [r6388] : running_under_rosetta only exists on Darwin, and should probably be removed completely. Now that it's explicitly extern, don't try to reference it at runtime on linuxppc. 2007-05-01 04:16 gb * [r6387] : use_mach_exception_handling is referenced at runtime and explicitly extern, so make sure that it's initialized on all platforms. 2007-05-01 04:15 gb * [r6386] : running_under_rosetta - which is kind of an oxymoron - is specific to Darwin. 2007-04-21 06:16 gb * [r6320] : 2007-04-21 05:52 gb * [r6319] : In SAVE-APPLICATION: check for containing directory after translating logical pathnames; it's generally meaningless to do PROBE-FILE on the directory component of a logical pathname. 2007-04-19 21:16 gb * [r6308] : 2007-04-19 20:50 gb * [r6307] : In x862-aset3, maybe push/pop the val reg if it conflicts with an immediate temporary (we'll need all the imm temps we have on x86-64.) 2007-04-18 02:17 gb * [r6288] : 2007-04-18 02:02 gb * [r6287] : Actually remember to check in the PPC version of recent aset2/aset3 register-conflict changes. 2007-04-16 21:17 gb * [r6286] : 2007-04-16 20:17 gb * [r6285] : 2007-04-16 20:08 gb * [r6284] : Don't try to access ARRAY-CTYPE-ELEMENT-TYPE of things that aren't ARRAY-CTYPEs in ACODE-FORM-TYPE. 2007-04-16 19:52 gb * [r6283] : :signed-doubleword, not :signed-doublewod. 2007-04-16 19:17 gb * [r6282] : 2007-04-16 18:15 gb * [r6281] : Memory-fault errors on PPC, too. 2007-04-16 18:15 gb * [r6280] : Memory-fault errors on PPC, too. 2007-04-16 11:17 gb * [r6279] : 2007-04-16 10:29 gb * [r6278] : Define __err, to work around Leopard renaming nonsense. 2007-04-16 08:17 gb * [r6277] : 2007-04-16 07:17 gb * [r6276] : 2007-04-16 07:17 gb * [r6275] : is_write_fault for FreeBSD. 2007-04-16 07:08 gb * [r6274] : is_write_vault for x86-64 Linux. 2007-04-16 07:08 gb * [r6273] : Change wording of INVALID-MEMORY-ACCESS error message, a little. 2007-04-16 06:38 gb * [r6272] : Call out to lisp on unhandled memory faults. 2007-04-16 06:37 gb * [r6271] : "advance PC" command is PPC-only, for now. 2007-04-16 06:37 gb * [r6270] : Map SIGBUS signal to INVALID-MEMORY-ACCESS condition. 2007-04-16 06:36 gb * [r6269] : INVALID-MEMORY-ACCESS condition. 2007-04-16 01:17 gb * [r6268] : 2007-04-16 00:53 gb * [r6267] : Use HARD-REGSPEC-VALUE in recent aset2/aset3 changes. 2007-04-16 00:48 gb * [r6266] : in aset2/aset3: don't clobber val-reg if it's an imm-reg; we need another imm-reg to do some index calculations. 2007-04-15 00:17 gb * [r6265] : 2007-04-14 23:17 gb * [r6264] : raise_pending_interrupt(): unscramble what it means for an interrupt to be pending on x86-64. 2007-04-14 06:18 gb * [r6263] : 2007-04-14 05:38 gb * [r6262] : I have no idea why SIGSYS is generated when it is in ... unmentionable OS releases. 2007-04-14 05:37 gb * [r6261] : Say "extern" a few times, to work around laughable Apple ld64 gibberish. 2007-04-14 05:37 gb * [r6260] : Say "extern" a few times, to work around laughable Apple ld64 gibberish. 2007-04-12 08:16 gb * [r6259] : 2007-04-12 07:20 gb * [r6258] : (SETF (AREF ...) ...) of two-dimensional arrays: need to target specific registers if array has element type T. 2007-04-12 07:19 gb * [r6257] : Add features :x86-64 and :x86_64 if appropriate, not :x86-64 twice. 2007-04-11 11:16 gb * [r6256] : 2007-04-11 10:11 gb * [r6255] : Set IOBLOCK-ELEMENT-TYPE to CHARACTER for bivalent streams. STREAM-ELEMENT-TYPE of BASIC-STREAM, BUFFERED-STREAM-MIXIN: don't maintain the element-type separately, just look in ioblock. (STREAM-ELEMENT-TYPE isn't required to work on closed streams.) 2007-04-11 04:16 gb * [r6254] : 2007-04-11 03:51 gb * [r6253] : Branch at C(egc_store_node_conditional_success_test) should bracnch backwards, not forwards. (This was leading to spurious " is not a proper list" errors, because the next label named "0" is in middle of the the _SPprogvsage code that tests for a proper list.) 2007-04-10 06:16 gb * [r6252] : 2007-04-10 05:42 gb * [r6251] : Don't bother using RLET or equivalent to avoid making GCable pointers for ObjC dispatch function args; trust the compiler to do that for us. 2007-04-10 05:41 gb * [r6250] : If any method defined on a message passes a structure by value, install a compiler macro on the dispatch function. In that compiler, look for apparent calls to know structure constructors; replace those with calls to the corresponding initialization function, and stack-allocate a structure instance around the call. (Suggested by Kevin Reid, and something that SEND/SLET did - sort of - that the new approach didn't.) Keep track of info about common foreign structs (initializer, constructor, etc.) to make doing stuff like this easier. 2007-04-09 22:16 gb * [r6249] : 2007-04-09 21:08 gb * [r6248] : ALLOCATE-INSTANCE of OBJC:OBJC:CLASS: if the result of sending the init message is a null pointer, return a null pointer. 2007-04-09 06:16 gb * [r6247] : 2007-04-09 05:58 gb * [r6246] : Update for 070408. (Yes, it's 070409 here already ...) 2007-04-08 21:16 gb * [r6245] : 2007-04-08 20:27 gb * [r6244] : Bump image version. 2007-04-08 20:16 gb * [r6243] : 2007-04-08 20:08 gb * [r6242] : Bump min fasl version. 2007-04-08 19:58 gb * [r6241] : Org to the next page boundary at end of jump table, to work around (still more) Apple ld64 nonsense. 2007-04-08 19:16 gb * [r6240] : 2007-04-08 19:04 gb * [r6239] : Bump current fasl version. 2007-04-08 18:49 gb * [r6238] : Export CCL:FREE. 2007-04-08 18:46 gb * [r6237] : Bump max fasl version. 2007-04-08 18:39 gb * [r6236] : Pass -arch ppc64 to C, so that people who try to build this kernel on an x86 mac will take longer to realize that that's usually a bad idea. 2007-04-08 18:16 gb * [r6235] : 2007-04-08 17:07 gb * [r6234] : Use new syntax: OBJC:DEFMETHOD, #/, MAKE-INSTANCE. Some changes in font, nsstring stuff to avoid using deprecated featurs. 2007-04-08 17:06 gb * [r6233] : Use new syntax: OBJC:DEFMETHOD, #/, MAKE-INSTANCE. Use OBJC:LOAD-FRAMEWORK. 2007-04-08 17:05 gb * [r6232] : Use new syntax: OBJC:DEFMETHOD, #/, MAKE-INSTANCE. Use OBJC:LOAD-FRAMEWORK. 2007-04-08 17:05 gb * [r6231] : Use new syntax: OBJC:DEFMETHOD, #/, MAKE-INSTANCE. 2007-04-08 17:04 gb * [r6230] : Use new syntax: OBJC:DEFMETHOD, #/, MAKE-INSTANCE. Handle CGFLOAT issues in mouse event handlers. 2007-04-08 17:02 gb * [r6229] : Define CGFLOAT, ObjC-2 foreign types here. Define "foreign struct classes" for several common structure types (rectangle, point, size); generate inlined accessors for their slots, PRINT-OBJECT methods, initialization functions, WITH-xxx macros. #/ reader macro: reads a string containing "constituent" characters (including #\:), preserving case. Does a little bit of sanity-checking on it, then interns it in NEXTSTEP-FUNCTIONS package. Try to deal with cases where foreign type info may be parsed or unparsed. Introduce OBJC-DISPATCH-FUNCTIONs, which are funcallable instances. SHARED-INITIALIZE :AFTER method on OBJC-DISPATCH-FUNCTION looks at message info, tries to determine ambiguity, calls signature function (possibly after trying to resolve ambiguity.) %CALL-NEXT-OBJC-METHOD does some of the same stuff at runtime, calling a signature function that knows how to dispatch to SUPER method. (Likewise, %CALL-NEXT-OBJC-CLASS-METHOD for class methods.) POSTPROCESS-OBJC-MESSAGE-INFO (re-) initializes the associated OBJC-DISPATCH-FUNCTION. Handle PROTOCOLs differently in ambiguous SEND. 2007-04-08 16:46 gb * [r6228] : Compile message send functions per signature. Handle vararg message sends, partly at runtime. INIT messages just have their signature info precomputed. Defining a named OBJC-CLASS makes the class name a static variable whose value is the class (and likewise for the metaclass.) %DECLARE-OBJC-MESSAGE notes method redefinition, cerrors. OBJC:DEFMETHOD and support for it. 2007-04-08 16:32 gb * [r6227] : Use new syntax. OBJC:LOAD-FRAMEWORK. Avoid some deprecated C-string stuff (but wimp out and assume ASCII.) Keep track of objc-class-count, provide MAYBE-MAP-OBJC-CLASSES to map new classes iff class count changes. Handle OBJC-PROTOCOLs a bit differently. Move CGFLOAT definitions, etc. elsewhere. 2007-04-08 16:23 gb * [r6226] : RECOGNIZE-OBJC-OBJECT: maybe try mapping classes if first try fails. ALLOCATE-INSTANCE of objc-object: don't canonicalize instance if no lisp slots (so MAKE-INSTANCE is about as lightweight as MAKE-OBJC-INSTANCE.) 2007-04-08 16:20 gb * [r6225] : COMPUTE-OBJC-TO-LISP-FUNCTION-NAME was once going to do something non-trivial, now just interns in NEXSTEP-FUNCTIONS. 2007-04-08 16:18 gb * [r6224] : New file; export more stuff. Import some of it back into CCL. Provide a NEXTSTEP-FUNCTIONS package for pseudo-GFs. 2007-04-08 16:17 gb * [r6223] : Foreign-type-ordinals, some of which are canonical. Straighten out the auxiliary-foreign-types mess a bit. Intern integer types up to 64 bits. 2007-04-08 16:13 gb * [r6222] : DEFGLOBAL = DEFSTATIC. DEFLOADVAR makes its argument static. WITH-MACPTRS is saner. WITH-ENCODED-CSTR writes NUL octet(s) to the end of the encoded string. MAKE-GCABLE-RECORD, COPY-RECORD. 2007-04-08 04:57 gb * [r6221] : Slight changes (preserve typedef name) in function/objc method arglists. UNESCAPE-FOREIGN-NAME: handle missing brackets. 2007-04-08 04:56 gb * [r6220] : Handle signals on the altstack if not Darwin. In Darwin, evactuate off of the lisp stack while interrupts are still disabled. 2007-04-08 04:49 gb * [r6219] : USE_SIGALTSTACK, but not on Darwin (where Apple hasn't yet invented the notion that it needs to be thread-specific.) 2007-04-08 04:48 gb * [r6218] : Clear pending interrupt flag when destroying TCR. 2007-04-08 04:47 gb * [r6217] : Lose (commented-out) code which worked around an old Darwin bug. 2007-04-08 04:47 gb * [r6216] : Don't define USE_SIGALTSTACK here. 2007-04-08 04:46 gb * [r6215] : Don't nuke null pointers. 2007-04-08 04:45 gb * [r6214] : Add gtra. 2007-04-08 04:45 gb * [r6213] : zeroed stack-block zeroes all bits of macptr.domain, macptr.type. 2007-04-08 04:44 gb * [r6212] : Deal with pkg.intern-hook. 2007-04-08 04:43 gb * [r6211] : add pkg.intern-hook to package. 2007-04-08 04:41 gb * [r6210] : Start trying to conditionalize some of the stack-frame inspector stuff. 2007-04-08 04:41 gb * [r6209] : Structures are basically passed as 64-bit word components, with FP fields passed in registers. This requires some special handling ("hybrid" parameters) in the compiler and %FF-CALL. 2007-04-08 04:39 gb * [r6208] : %GET-BITFIELD and SETF thereof: deal with endianness. 2007-04-08 04:38 gb * [r6207] : FASL-SCAN: null-pointers as constants are OK. 2007-04-08 04:37 gb * [r6206] : From Takehiko Abe: handle EQL specializers in %SOURCE-FILES-LIKE-EM. 2007-04-08 04:37 gb * [r6205] : DRIBBLE via PROCESS-DRIBBLE. 2007-04-08 04:36 gb * [r6204] : From Takehiko Abe: handle EQL specializers in MAKE-XREF-ENTRY. 2007-04-08 04:34 gb * [r6203] : Export DEFSTATIC, PAREF, +NULL-PTR+. 2007-04-08 04:33 gb * [r6202] : Bump; new snapshots tomorrow. 2007-04-08 04:32 gb * [r6201] : DEFGLOBAL->DEFSTATIC 2007-04-08 04:32 gb * [r6200] : Very long timeout for WAIT-ON-SEMAPHORE, don't use PROCESS-WAIT, manipulate the whostate if we have to wait at all. Same idea for TIMED-WAIT-ON-SEMAPHORE. Make PIPE code slightly easier to read. 2007-04-08 04:29 gb * [r6199] : DEFGLOBAL->DEFSTATIC. 2007-04-08 04:29 gb * [r6198] : Add NUL-ENCODING field to CHARACTER-ENCODING, use it to determine number/sequence of octets used to encode NUL. 2007-04-08 04:26 gb * [r6197] : DEFGLOBAL->DEFSTATIC. READ-TOPLEVEL-FORM on ECHO-STREAM has do I/O on the echo stream (so DRIBBLE works); on TWO-WAY-STREAM, should read from the input side so that echoing looks more natural. 2007-04-08 04:21 gb * [r6196] : Extra slot (for intern hook) in PACKAGE object. 2007-04-08 04:12 gb * [r6195] : Some changes, then reverted those changes manually and most of the effects were cosmetic. TOPLEVEL-PRINT takes an output stream arg. 2007-04-08 03:58 gb * [r6194] : Per-process DRIBBLE-ing. 2007-04-08 03:58 gb * [r6193] : %TCR-INTERRUPT: don't signal if interrupt is obviously pending. 2007-04-08 03:57 gb * [r6192] : Change pointer printing. 2007-04-08 03:56 gb * [r6191] : +NULL-PTR+. 2007-04-08 03:56 gb * [r6190] : DEFSTATIC is now the preferred name for DEFGLOBAL. 2007-04-08 03:55 gb * [r6189] : DEFSTATIC is now the preferred name for DEFGLOBAL. 2007-04-08 03:55 gb * [r6188] : DEFSTATIC is now the preferred name for DEFGLOBAL. Mechanism for associating a (built-in) class with a foreign type ordinal. 2007-04-08 03:18 gb * [r6187] : DEFSTATIC is now the preferred name for DEFGLOBAL. %CALLBACK-FUNCTION accepts a symbol (as well as a pointer that's likely the global value of some symbol.) 2007-04-08 03:09 gb * [r6186] : DEFSTATIC is now the preferred name for DEFGLOBAL. Standard CL streams now bound per-thread. Install standard foreign types here, not every time that FOREIGN-TYPES is loaded. Canonicalize foreign type ordinals as soon as we can. (This is part of a scheme to give foreign types "ordinals' that can be used to assert pointer types; we need some of those ordinal numbers to be pre-assigned, since we can't reference foreign types early in the cold load.) 2007-04-08 03:06 gb * [r6185] : DEFSTATIC is now the preferred name for DEFGLOBAL. GCABLE-PTR-P. 2007-04-08 03:04 gb * [r6184] : %INSERT-SYMBOL might call pkg.intern-hook. Handle maintenance of auto-export packages differently. $fasl-timm just needs an :int, not a :long. 2007-04-08 03:02 gb * [r6183] : SET-TYPE-PREDICATE returns the "functional", not random symbol guts. 2007-04-08 03:01 gb * [r6182] : Use %PTR-STORE-FIXNUM-CONDITIONAL in %GET-SPIN-LOCK. Locking code: spin lock must be released before interrupts are re-enabled. 2007-04-08 02:58 gb * [r6181] : FREE tries to handle gcable pointers. 2007-04-08 02:58 gb * [r6180] : %PTR-STORE-FIXNUM-CONDITIONAL. 2007-04-08 02:58 gb * [r6179] : %PTR-STORE-FIXNUM-CONDITIONAL. 2007-04-08 02:57 gb * [r6178] : Handle "hybrid" integer/float arguments in DarwinPPC64 #'%FF-CALL. 2007-04-08 02:56 gb * [r6177] : When trying to infer result type of MAKE-ARRAY call, unquote quoted type specifier. Strength-reduce FLOAT when second argument is of known type. 2007-04-08 02:54 gb * [r6176] : Operators of %SINGLE-FLOAT, %DOUBLE-FLOAT. 2007-04-08 02:54 gb * [r6175] : Allow extra "hybrid" ff-call argspecs (for DarwinPPC64.) Cut-and-paste bug re: nx1-%fixnum-to-single/double. Parse %short-float, %double-float when no "result" arg; can at least try to constant-fold late and/or do some cases if we have more type info. 2007-04-08 02:51 gb * [r6174] : Lose some whitespace in the "static var can't be bound" error message. 2007-04-08 02:50 gb * [r6173] : Same inline/lambda-bind/dynamic-extent bug as on PPC. %SINGLE-FLOAT, %DOUBLE-FLOAT here, too. 2007-04-08 02:49 gb * [r6172] : In EXPAND-FF-CALL, adjust remaning GPR/FPR counts even if arg is "simple" and passes straight through to %FF-CALL. 2007-04-08 02:48 gb * [r6171] : Fix a long-standing bug involving LAMBDA-BIND operator, dynamic-extent &REST, and stack memoization. Handle DarwinPPC64 ABI conventions, where SINGLE-FLOATs may be passed in the same doubleword as GPRs. %DOUBLE-FLOAT, %SINGLE-FLOAT stuff: do inlining late, catch some other cases. 2007-04-08 02:39 gb * [r6170] : RELOAD-SINGLE-C-ARG-HIGH, for Darwin structure-passing ABI. 2007-04-08 02:38 gb * [r6169] : Convert .nib to use utf-8 representation, not binary. 2007-04-08 02:36 gb * [r6168] : Convert .nib to use utf-8 representation, not binary. 2007-04-07 04:16 gb * [r6143] : 2007-04-07 03:18 gb * [r6142] : Patch from Andi Kleen: set vmemoryuse (RLIMIT_AS) current resource limit to max limit, since some distros limit it. 2007-04-03 09:17 gb * [r6116] : 2007-04-03 08:30 gb * [r6115] : Takehiko Abe's fix to %IOBLOCK-WRITE-U16-ENCODED-CHAR. 2007-04-03 04:17 gb * [r6114] : 2007-04-03 03:30 gb * [r6113] : Takehiko Abe's fixes (missing "stream" arg in utf-16 encode, missing #x in little-endian bom-marker definition.) 2007-03-15 05:20 gb * [r6040] : 2007-03-15 04:52 gb * [r6039] : Remove a debugging fprintf. 2007-03-14 11:20 gb * [r6037] : 2007-03-14 10:54 gb * [r6036] : Fix typo (extra quote) in generate-callback-bindings. 2007-03-13 12:21 gb * [r6035] : 2007-03-13 12:05 gb * [r6034] : (re-)enable -O2. 2007-03-13 10:21 gb * [r6033] : 2007-03-13 08:27 gb * [r6032] : 2007-03-12 11:55 gb * [r6031] : Fix fenceposts in CLOSURE-CLOSED-OVER-VALUES. 2007-03-12 11:05 gb * [r6030] : Simplify interrupt_handler() a little. 2007-03-12 11:02 gb * [r6029] : Fix "extra arguments" error message in %EXTERNAL-CALL-EXPANDER. 2007-03-12 10:59 gb * [r6028] : class-wrapper-random-state: use encapsulation, and a lock. 2007-03-12 10:53 gb * [r6027] : *RANDOM-STATE* is thread-local. 2007-03-12 10:53 gb * [r6026] : *RANDOM-STATE* is thread-local. 2007-03-12 08:08 gb * [r6025] : in catch_exception_raise(), try to deal with variations in Mach behavior: "exception" may be EXC_BAD_ACCESS when an illegal instruction causes a "general protection fault". 2007-03-09 22:54 gb * [r6023] : Bill's fix to (MAKE-PATHNAME :DIRECTORY NIL ...) 2007-03-09 07:20 gb * [r6022] : *** empty log message *** 2007-03-09 07:01 gb * [r6021] : Use CLEAR-OPEN-FILE-STREAMS. 2007-03-09 07:00 gb * [r6020] : Accessors of open-file-streams; no more *OPEN-FILE-STREAMS*. 2007-03-09 07:00 gb * [r6019] : Encapsulation for open-file-streams. 2007-03-09 06:59 gb * [r6018] : OPEN-FILE-STREAMS is a function, returns a copy of a list of open file streams in PREPARE-TO-QUIT. 2007-03-09 06:58 gb * [r6017] : Lose *OPEN-FILE-STREAMS* and the lock associated with it. 2007-03-09 06:57 gb * [r6016] : Bind *GENSYM-COUNTER* per-thread. 2007-03-09 06:47 gb * [r6015] : When REMHASH deletes the last entry in a hash-table, it fills the hash-table vector with "free" markers. (There may already be lots of "deleted" markers in the table.) 2007-03-09 06:20 gb * [r6014] : *** empty log message *** 2007-03-09 05:26 gb * [r6013] : *spin-lock-tries* with a trailing *. 2007-03-09 05:24 gb * [r6012] : printing random-state: don't shift seeds. 2007-03-09 05:24 gb * [r6011] : MAKE-RANDOM-STATE: don't shift. 2007-03-09 05:24 gb * [r6010] : Do %next-random-seed more portably. 2007-03-09 05:23 gb * [r6009] : Typo in comment. 2007-03-09 05:23 gb * [r6008] : Just a 32-bit version of %next-random-pair. 2007-03-09 05:22 gb * [r6007] : And, while we're at it: do the next-random-seed permutation in %next-random-pair, which accepts and returns a pair of 16-bit integers. That's in lap on ppc32 and in lisp on 64-bit platforms. 2007-03-09 05:20 gb * [r6006] : Don't shift random seeds. 2007-03-08 11:20 gb * [r6005] : *** empty log message *** 2007-03-08 10:30 gb * [r6004] : Check for trap opcode >= #x70 rather than >= #x80, because of the new way that uuo-error-slot-unbound is encoded. 2007-03-08 10:28 gb * [r6003] : Recognize uuo-error-slot-unbound. 2007-03-08 10:28 gb * [r6002] : Use (int #x7_) for uuo-error-slot-unbound, since some (int #x8_) traps confuse Mach. 2007-03-07 16:20 gb * [r6001] : *** empty log message *** 2007-03-07 15:29 gb * [r6000] : Don't shadow the definition of PPC-DIV2. 2007-03-07 15:20 gb * [r5999] : *** empty log message *** 2007-03-07 14:45 gb * [r5998] : For #+linux-target, try to use #_sysconf in CPU-COUNT, and fall back to parsing /proc/cpuinfo only if that fails. (CPU-COUNT is called early in the cold load, before we can make the file-stream necessary to do the parsing.) 2007-03-07 14:33 gb * [r5997] : Handle new GNU ld "hash-style" linker options. 2007-03-07 14:25 gb * [r5996] : Darwin signal return: logior in the el-bizarro syscall mask, pass the magic number 0x1e as a second arg, just in case. 2007-03-07 14:24 gb * [r5995] : OPEN, CLOSE lock *OPEN-FILE-STREAMS-LOCK* before accessing *OPEN-FILE-STREAMS*. 2007-03-07 14:23 gb * [r5994] : Add comment about *OPEN-FILE-STREAMS-LOCK*. 2007-03-07 14:23 gb * [r5993] : Add *OPEN-FILE-STREAMS-LOCK*. 2007-03-07 14:21 gb * [r5992] : Bind *FCOMP-PREVIOUS-POSITION* in FCOMP-READ-LOOP (e.g., don't just hammer on the static binding.) 2007-03-07 14:20 gb * [r5991] : *** empty log message *** 2007-03-07 14:05 gb * [r5990] : Bump current image version. 2007-03-07 14:02 gb * [r5989] : Some (currently unused) CPUID "calls" want an argument in ecx. Set that arg to 0 for now. 2007-03-07 14:01 gb * [r5988] : Use spinlocks to guard updates to "real" locks; those updates can be simplified. lock_recursive_lock: no (ignored) timeout arg. 2007-03-07 13:59 gb * [r5987] : Bump current, max abi versions; continue to support the old one. 2007-03-07 13:59 gb * [r5986] : #include #, for sched_yield() prototype. Change declaration of LOCK macro: no null timespec required. RECURSIVE_LOCKs have spinlock fields. 2007-03-07 13:56 gb * [r5985] : bump, indicate work-in-progress development version. 2007-03-07 13:56 gb * [r5984] : Determing *spin-lock-tries* at launch time, based on cpu-count. YIELD moved here, does #_sched_yield rather than something lower-level. 2007-03-07 13:53 gb * [r5983] : Locks use spinlocks, try *spin-lock-tries* to obtain them before yielding. 2007-03-07 13:52 gb * [r5982] : Move YIELD elsewhere. 2007-03-07 13:51 gb * [r5981] : Add spinlock field to lockptr storage layout. 2007-03-02 12:20 gb * [r5980] : *** empty log message *** 2007-03-02 12:10 gb * [r5979] : Toplevel function: initial-listener class is 'TTY-LISTENER. 2007-03-02 12:09 gb * [r5978] : Print n's value, not n in error message. 2007-03-02 12:09 gb * [r5977] : New process class: TTY-LISTENER. EXIT-INTERACTIVE-PROCESS: for most processes, just PROCESS-KILL. For TTY-LISTENER, call QUIT. 2007-03-02 12:08 gb * [r5976] : Don't call QUIT on EOF, call EXIT-INTERACTIVE-PROCESS instead. Look at *QUIT-ON-EOF* when deciding what to do in response to EOF. 2007-03-02 12:06 gb * [r5975] : Export *quit-on-eof*. 2007-03-02 12:06 gb * [r5974] : WITH-MACPTRS: move initialization/modification of null pointers into body. 2007-03-02 12:05 gb * [r5973] : Try to handle two more cases of consus-interruptus. 2007-03-02 12:05 gb * [r5972] : (the ) transforms to if constant is of type .) Do we need an ignore-errors here ? Or does need to be defined ? 2007-03-02 12:03 gb * [r5971] : Do the / * thing. 2007-03-02 12:02 gb * [r5970] : things that do calls need :call attribute. 2007-03-01 11:20 gb * [r5969] : *** empty log message *** 2007-03-01 11:04 gb * [r5968] : binding primitives: set value after new binding frame 2007-02-28 14:20 gb * [r5967] : *** empty log message *** 2007-02-28 14:04 gb * [r5966] : In %fixnum-truncate, use -8(sp) rather than %rbp; trust callback_for_interrupt() to preserve it. 2007-02-28 14:03 gb * [r5965] : Try to preserve -8(sp) across callback_for_interrupt(). 2007-02-28 12:20 gb * [r5964] : *** empty log message *** 2007-02-28 12:18 gb * [r5963] : check_node: ignore tiny things (shift counts) that may wind up on the vstack. Since those tiny things might be tagged as headers, walk the vstack with new "headerless" functions that ignore headers. 2007-02-28 12:16 gb * [r5962] : In pc_luser_xp: if allocating and before the alloc trap and interrupting, back out of the modification to save_allocptr (and allocptr register) and set *interrupt_displacement. 2007-02-23 17:20 gb * [r5961] : *** empty log message *** 2007-02-23 17:13 gb * [r5960] : Changes to GUESS-ALIGNMENT. Again, this only makes sense for cases where we have to guess; a bitfield that's 8 bits wide still has an alignment of 1. Factor in the alignment of the containing int when bitfields are involved. When parsing a record type, set the "bits" slot of each field. (Not too many things look at that slot, but we want the effect of loading a record from the database to match the effect of parsing a definition as exactly as possible, and to be able to use EQUALP to compare the results.) 2007-02-23 17:06 gb * [r5959] : Some extra fields in objc-message-info/objc-method-info structures. When calculating overall alignment of a structure, treat bitfields as if they were :unsigned-int (since there's a "containing" unsigned int that affects alignment. N.B: this should be imposed for anything that has an alignment of 1, not just a size of 1.) When loading integer types, try to use the precomputed types rather than consing a new one. This (a) reduces consing a little (b) is more likely to get the alignment correct. For (b), we really need to distinguish between bitfields and addressable integers. 2007-02-23 16:30 gb * [r5958] : Nonsense in x862-set-bit. 2007-02-22 10:20 gb * [r5957] : *** empty log message *** 2007-02-22 09:33 gb * [r5956] : Tweak ROOM output a little. 2007-02-22 09:28 gb * [r5955] : Fix up the verbose GC herald a bit (get heap size right ...). 2007-02-22 09:27 gb * [r5954] : Tell GNU ld to use the traditional (sysv) style of hash maps, if it understands that option. 2007-02-21 13:20 gb * [r5953] : *** empty log message *** 2007-02-21 13:01 gb * [r5952] : Handle the case where nothing has been consed since last GC. (There are some fenceposts there.) verbose GC herald tries to indicate how much is allocated in the area being GCed. 2007-02-21 12:58 gb * [r5951] : If full GC is deferred, ensure that a full GC happens after WITHOUT-GCING exits. 2007-02-21 12:57 gb * [r5950] : Continue disassembly after uuo-gc-trap. 2007-02-19 12:20 gb * [r5949] : *** empty log message *** 2007-02-19 11:50 gb * [r5948] : Use keywords more consistently. (Note: there seems to be a bug which is keeping the cdb data structures from getting defined correctly at load time; the workaround is to compile foreign-types.lisp - which defines them - in the same session as and before db-io - which uses them - is compiled.) 2007-02-19 11:48 gb * [r5947] : UNSET-FIN-TRAMPOLINE: don't tail-call error. 2007-02-19 11:42 gb * [r5946] : UNSET-FIN-TRAMPOLINE: don't tail-call the error function. 2007-02-19 11:40 gb * [r5945] : MAYBE-UPDATE-OBSOLETE-INSTANCE: handle funcallable objects that aren't STANDARD-GENERIC-FUNCTIONs. 2007-02-19 11:40 gb * [r5944] : Move GENERIC-FUNCTION-NAME to FUNCALLABLE-STANDARD-OBJECT. %ALLOCATE-GF-INSTANCE: don't create a dispatch table if not creating a GF. 2007-02-19 11:36 gb * [r5943] : STANDARD-OBJECT-LINE-N: handle funcallable-stanard-objects better. 2007-02-19 11:34 gb * [r5942] : Export FUNCALLABLE-STANDARD-OBJECT from CCL, OPENMCL-MOP. 2007-02-19 09:20 gb * [r5941] : *** empty log message *** 2007-02-19 09:09 gb * [r5940] : FUNCALLABLE-STANDARD-OBJECT is a FUNCALLABLE-STANDARD-CLASS. 2007-02-19 09:07 gb * [r5939] : MAKE-OPTIMIZED-SEND: missing "msg" arg in error message. 2007-02-15 10:20 gb * [r5938] : *** empty log message *** 2007-02-15 09:46 gb * [r5937] : Call REGISTER-OBJC-INIT-MESSAGES after MAP-OBJC-CLASSES in LOAD-OBJC-EXTENSION-FRAMEWORK. 2007-02-15 09:45 gb * [r5936] : UPDATE-OBJC-METHOD-INFO: don't process init methods yet. 2007-02-14 10:20 gb * [r5935] : *** empty log message *** 2007-02-14 09:39 gb * [r5934] : Fix a typo. 2007-02-14 09:35 gb * [r5933] : Update for 070214 snapshots. 2007-02-14 07:20 gb * [r5932] : *** empty log message *** 2007-02-14 06:36 gb * [r5931] : Bump. 2007-02-14 06:36 gb * [r5930] : Update image-abi-version. 2007-02-14 06:29 gb * [r5929] : Finish bumping fasl version. 2007-02-14 06:20 gb * [r5928] : *** empty log message *** 2007-02-14 06:15 gb * [r5927] : Bump current fasl version. 2007-02-14 06:02 gb * [r5926] : rm -rf ./usr 2007-02-14 05:45 gb * [r5925] : Start bumping fasl-version. 2007-02-14 05:42 gb * [r5924] : rm -rf ./usr 2007-02-14 04:20 gb * [r5923] : *** empty log message *** 2007-02-14 04:16 gb * [r5922] : Avoid processing some files that contain incompatible definitions of some math library stuff ("log" either has to do with logarithms or logging ...). 2007-02-14 02:20 gb * [r5921] : *** empty log message *** 2007-02-14 01:42 gb * [r5920] : Change the way that TIME is reported by default: mention the number of available CPU cores, don't assume that there's that much of a relationship between CPU time and wall-clock time. 2007-02-14 01:41 gb * [r5919] : Don't do whatever el-bizarro Mach thing once seemed to be necessary in %%RUSAGE. Require syscall stuff when :execute, too. 2007-02-14 01:35 gb * [r5918] : It's %%get-unsigned-longlong, and the offset arg is required. 2007-02-14 01:20 gb * [r5917] : *** empty log message *** 2007-02-14 01:12 gb * [r5916] : Add rdtsc (returns 60 bit non-negative fixnum) and rdtsc64 (returns 64-bit non-negative integer.) 2007-02-14 01:10 gb * [r5915] : Simpler GET-INTERNAL-RUN-TIME (tries harder not to cons.) 2007-02-14 01:10 gb * [r5914] : A little bit of cgfloat stuff; there are more/other float-size issues deeper down in the code. 2007-02-14 01:09 gb * [r5913] : Fixup the x86-64 error-return callback trampiline code (was missing a byte of machine code.) 2007-02-14 01:08 gb * [r5912] : Set the mxcsr/fpctrl fields of jmp-buf to safe/sane values. 2007-02-14 01:07 gb * [r5911] : Don't change-class *cocoa-event-process* until we start to run the event loop. Don't do _deallocHardCore: explicitly. 2007-02-14 00:35 gb * [r5910] : poweropen-foreign-args: on ppc64, a DOUBLE-FLOAT aliases just one gpr. 2007-02-14 00:34 gb * [r5909] : Select the SDK directory. 2007-02-14 00:34 gb * [r5908] : return-value pointers in OPEN-PTY-PAIR are just 32-bit ints, not "long". 2007-02-14 00:33 gb * [r5907] : Callback arguments: update offset/delta for each scalar, not each arg. Fix some backquoting/indentation. 2007-02-14 00:31 gb * [r5906] : Fix typo (^get-signed-long). 2007-02-13 00:20 gb * [r5905] : *** empty log message *** 2007-02-12 23:23 gb * [r5904] : Select an SDK. 2007-02-12 04:20 gb * [r5903] : *** empty log message *** 2007-02-12 03:22 gb * [r5902] : method-typestring needs to be conditionalized. How has this been working ? 2007-02-12 03:22 gb * [r5901] : remove some debugging code. 2007-02-12 03:21 gb * [r5900] : %mlist-containing is #-apple-objc-2.0. 2007-02-12 03:21 gb * [r5899] : Lose a quote mark, fix another typo. 2007-02-11 13:20 gb * [r5898] : *** empty log message *** 2007-02-11 12:35 gb * [r5897] : Bump. 2007-02-11 12:28 gb * [r5896] : Make the SDK prefix an optional argument; default to the Tiger SDK. 2007-02-11 12:26 gb * [r5895] : Ignore errors; it's an error to ignore an errror. 2007-02-11 12:26 gb * [r5894] : Conditionalize for gratuitous Leopard renaming. Suffer. 2007-02-11 12:25 gb * [r5893] : MAKE-MCL-LISTENER-PROCESS will preset an existing process (so that the listener process appears to be "persistent".) 2007-02-11 12:24 gb * [r5892] : LISP-DEVELOPMENT-SYSTEM tries to make the initial-listener object persistent acrosss SAVE-APPLICATION. 2007-02-11 07:20 gb * [r5891] : *** empty log message *** 2007-02-11 07:01 gb * [r5890] : Deftype for CF-FLOAT. Handle callback-error-returns on x86-64. 2007-02-11 07:00 gb * [r5889] : Pointer comparisons in splay trees use %ptr<, to hide word-size differences. __NSConstantStringClassReference isn't what it used to be, so look up NSConstantStringClass the hard way. Handle ns-exception/lisp-error stuff on x86-64. Do arg/result coercions differently. Does any code depend on COERCE-TO-ADDRESS behavior ? (It can cause extra consing of MACPTRs). Use the new FFI interface for message-sends, callbacks to try to hide platform-specific ABI issues. The _deallocate-hook mechanism - which never really worked - probably can't work on Objc-2.0. 2007-02-11 06:54 gb * [r5888] : Make structure return-type explicit in objc-message-send calls in send/stret and variants. 2007-02-11 06:20 gb * [r5887] : *** empty log message *** 2007-02-11 05:34 gb * [r5886] : Some indentation changes and slight reorganization of the code; shouldn't be any functional/semantic changes. 2007-02-11 05:30 gb * [r5885] : Need to ensure that floats are CG-FLOATs when appropriate; other changes for 64-bit platforms, Objc-2.0. 2007-02-11 05:29 gb * [r5884] : Need to ensure that floats are CG-FLOATs when appropriate. 2007-02-11 05:28 gb * [r5883] : Ignore fasls for all platforms. 2007-02-11 05:28 gb * [r5882] : VALUES uses :ARGLIST pseudo-op. 2007-02-11 05:27 gb * [r5881] : EXPAND-FF-CALL: ignore-errors when trying to parse the last arglist form as a foreign type. 2007-02-11 05:26 gb * [r5880] : Pretty much commited to the new FF-CALL expansion ... 2007-02-11 05:26 gb * [r5879] : Preserve %rax when doing ff-call; use %rdx and %r11 to hold target address. 2007-02-11 05:23 gb * [r5878] : Unicode stuff when accessing control char array; conditionalize for #+darwin-target, not just #+darwinppc-target. 2007-02-11 05:22 gb * [r5877] : Fix a comment (defcallback isn't ppc-specific; the comment dated from the days of 68K MCL.) 2007-02-11 05:21 gb * [r5876] : Fix some typos; do struct return 32 bits at a time, to avoid consing little bignums. 2007-02-11 05:20 gb * [r5875] : *** empty log message *** 2007-02-11 05:15 gb * [r5874] : Support :arglist directive. 2007-02-11 05:15 gb * [r5873] : alloc-c-frame vinsn takes "extra" word count, not total byte count. Fix handling of $undo-x86-c-frame in x862-nlexit. (Need to check the PPC case, as well.) 2007-02-11 05:11 gb * [r5872] : There doesn't need to be a "zero-extend-s8" vinsn, but there does need to be a "zero-extend-u8". 2007-02-10 13:20 gb * [r5871] : *** empty log message *** 2007-02-10 12:38 gb * [r5870] : Use new :ARGLIST directive in #'VALUES. 2007-02-10 12:38 gb * [r5869] : Support an :ARGLIST directive, so that ARGLIST can say -something- about lap functions that don't take a fixed number of arguments. 2007-02-07 10:20 gb * [r5868] : *** empty log message *** 2007-02-07 09:54 gb * [r5867] : They spell "valu" with an "e". 2007-02-02 10:20 gb * [r5866] : *** empty log message *** 2007-02-02 09:32 gb * [r5865] : Start to integrate with FFI changes. No more %SEND, process and precompile "init" messages when interfaces are accessed. (Note: should try to handle varargs init messages, even if we have to call %FF-CALL at runtime.) 2007-02-02 09:31 gb * [r5864] : Change the initialization code a bit. 2007-02-02 09:27 gb * [r5863] : Try to fix another case involving "auxiliary types". I think that the concept of auxiliary types might make sense in a context like: (def-foreign-type nil (:struct :foo (:a :int) (:b (:struct :bar (:x :int) (:y :int))))) In that case, :struct :bar is local to the definition. In the similar case: (def-foreign-type nil (:struct :foo (:a :int) (:b (:struct :bar)))) :BAR is presumably defined (globally) elsewhere. 2007-02-02 09:20 gb * [r5862] : *** empty log message *** 2007-02-02 09:17 gb * [r5861] : Specify the SDK to use. (Setting "SDK=" will hopefully work with older version of xcode.) 2007-02-02 09:15 gb * [r5860] : Ignore ./Developer. 2007-02-01 07:20 gb * [r5859] : *** empty log message *** 2007-02-01 06:26 gb * [r5858] : Don't treat :registers arg as a real argument in ppc2-poweropen-foreign-args. 2007-02-01 06:20 gb * [r5857] : *** empty log message *** 2007-02-01 05:47 gb * [r5856] : Needed ,@ to return assign structure components from registers. 2007-02-01 05:45 gb * [r5855] : UNESCAPE-FOREIGN-NAME: strings with no angle brackets are returned verbatim, not downcased. 2007-02-01 05:44 gb * [r5854] : Runtime version of 64-bit %ff-call accepts :registers. 2007-02-01 00:20 gb * [r5853] : *** empty log message *** 2007-01-31 23:36 gb * [r5852] : New (and, so far, useless) files. 2007-01-31 23:35 gb * [r5851] : Add a comment; CHUD.Framework still seems to be 32-bit only. 2007-01-31 23:34 gb * [r5850] : Ignore ./Developer. 2007-01-31 23:34 gb * [r5849] : Use SDK; add dlfcn.h. 2007-01-31 23:33 gb * [r5848] : Tweak the warning given if CDB-OPEN fails: missing cdb file: dog bytes man; missing parent directory is more newsworthy. 2007-01-31 23:20 gb * [r5847] : *** empty log message *** 2007-01-31 22:38 gb * [r5846] : Tweak a little, so that it might work on darwinppc64 (and on darwinx8664 when purify does something there.) 2007-01-31 22:20 gb * [r5845] : *** empty log message *** 2007-01-31 22:06 gb * [r5844] : Add dlfcn.h. 2007-01-31 10:20 gb * [r5843] : *** empty log message *** 2007-01-31 09:49 gb * [r5842] : BIGNUM-NEGATE-TO-PTR. 2007-01-31 06:20 gb * [r5841] : *** empty log message *** 2007-01-31 05:53 gb * [r5840] : new files. 2007-01-31 05:21 gb * [r5839] : new files. 2007-01-31 05:20 gb * [r5838] : *** empty log message *** 2007-01-31 04:58 gb * [r5837] : BIGNUM-NEGATE-TO-POINTER. 2007-01-31 04:58 gb * [r5836] : new file 2007-01-31 04:49 gb * [r5835] : Learn to conditionalize! It's easy and fun! 2007-01-31 04:40 gb * [r5834] : Conditionalize out the event-polling task. 2007-01-31 04:39 gb * [r5833] : new file 2007-01-31 04:35 gb * [r5832] : process glu.h 2007-01-31 01:20 gb * [r5831] : *** empty log message *** 2007-01-31 00:47 gb * [r5830] : Always encode explicit structure-typed function return values. Note when we see one, for now. 2007-01-31 00:45 gb * [r5829] : %EXTERNAL-CALL-EXPANDER allows/requires a leading argument to be paired with a structure-typed return-type. 2007-01-31 00:43 gb * [r5828] : linuxppc32 always returns structures via the first-arg hack. 2007-01-31 00:42 gb * [r5827] : Bump *interface-abi-version*, but continue to support the old version. Don't check the :struct-by-value flag when encoding function args/results. (That decision's made above our pay grade now.) When reading a :struct/:union definition, try to update an existing defintion rather than consing up a new one (this matters if the record is anonymous.) 2007-01-31 00:39 gb * [r5826] : Try to ensure that the :FORMAT argument gets passed to MAKE-TCP-STREAM-SOCKET and MAKE-FILE-STREAM-SOCKET correctly. Fix c_socket for #+linuxppc. 2007-01-30 16:20 gb * [r5825] : *** empty log message *** 2007-01-30 15:20 gb * [r5824] : Bump. 2007-01-30 15:20 gb * [r5823] : Use make -j to build kernel. 2007-01-30 14:20 gb * [r5822] : *** empty log message *** 2007-01-30 14:02 gb * [r5821] : .SPcallback returns rax/rdx, xmm0/xmm1. 2007-01-30 14:01 gb * [r5820] : Use new callback interface unconditionally. (A day's work turned into 2 weeks, somehow.) 2007-01-30 14:00 gb * [r5819] : New callback stuff for x8664 (all platfroms appear to use the same ABI.) 2007-01-30 11:20 gb * [r5818] : *** empty log message *** 2007-01-30 10:26 gb * [r5817] : Don't use boxed "disp" for displacement in %ATOMIC-INCF-NODE. 2007-01-29 15:20 gb * [r5816] : *** empty log message *** 2007-01-29 14:27 gb * [r5815] : Enable new callback stuff on #+ppc-target. 2007-01-29 14:27 gb * [r5814] : new callback stuff. 2007-01-29 14:27 gb * [r5813] : typos-r-us. 2007-01-29 12:20 gb * [r5812] : *** empty log message *** 2007-01-29 11:37 gb * [r5811] : Flesh out darwinppc32 callback stuff. 2007-01-29 11:37 gb * [r5810] : Fix some typos/bugs; more likely remain. 2007-01-29 11:36 gb * [r5809] : REQUIRE-FOREIGN-TYPE-BITS. 2007-01-29 11:36 gb * [r5808] : Use new callback scheme for darwinppc32, too. 2007-01-29 11:36 gb * [r5807] : No more (old) callback stuff here. 2007-01-29 10:20 gb * [r5806] : *** empty log message *** 2007-01-29 09:33 gb * [r5805] : Flesh out the new callback stuff. 2007-01-29 05:20 gb * [r5804] : *** empty log message *** 2007-01-29 04:47 gb * [r5803] : Flesh things out a bit more, change the API, get things working ... still a moving target. 2007-01-29 04:46 gb * [r5802] : For linuxppc, use the new callback mechanism. (Other platforms still need more bootstrapping.) 2007-01-29 04:44 gb * [r5801] : Lose the old eabi-callback stuff; linuxppc 32/64-bit callbacks now use the new scheme. 2007-01-28 11:20 gb * [r5800] : *** empty log message *** 2007-01-28 10:21 gb * [r5799] : Remove some typos. Need to pass dynamic-extent, other decls separately. 2007-01-28 10:20 gb * [r5798] : Fix some of this stuff; use the new (ftd-based) scheme. 2007-01-28 09:20 gb * [r5797] : *** empty log message *** 2007-01-28 08:43 gb * [r5796] : Fix some typos. (This code's getting replaced, but fix 'em anyway.) 2007-01-28 08:42 gb * [r5795] : If a .cdb file can't be opened, show the translated pathname and mention the fact that this might indicate a problem with "ccl:"'s logical translations. 2007-01-25 12:20 gb * [r5794] : *** empty log message *** 2007-01-25 11:37 gb * [r5793] : Never fully understood the concept "auxiliary" foreign types, but try to canonicalize intermediate struct/union/enum references, so that (EQ (PARSE-FOREIGN-TYPE '(:STRUCT :FOO)) (PARSE-FOREIGN-TYPE '(:STRUCT :FOO))) is more likely to return T (e.g., something registered in the FTD. It seems clear that this is the right thing when the structure reference is in global scope; I'm not sure what it'd mean for that not to be the case. 2007-01-25 11:32 gb * [r5792] : More plausible callback-bindings callback. 2007-01-25 07:20 gb * [r5791] : *** empty log message *** 2007-01-25 06:41 gb * [r5790] : First cut at callback-bindings thing; may not even compile. 2007-01-24 10:20 gb * [r5789] : *** empty log message *** 2007-01-24 09:39 gb * [r5788] : Stubs for new FTD functions. 2007-01-24 09:28 gb * [r5787] : A few more fields (for callbacks) in ftd defstruct. 2007-01-23 23:20 gb * [r5786] : *** empty log message *** 2007-01-23 22:28 gb * [r5785] : Get FPRs from where subprim saved them. Return FPRs in the same place. (Will need to re-work all of this for platform-dependent structure stuff.) 2007-01-23 22:26 gb * [r5784] : poweropen-callback, callbackX; save fp args here. Return all (arg) GPRs and FPRs (needs cooperation from the lisp side.) 2007-01-23 22:21 gb * [r5783] : Add room in c_reg_save for saved fp args (f1-f13). 2007-01-23 22:20 gb * [r5782] : *** empty log message *** 2007-01-23 22:14 gb * [r5781] : FF-CALL implemented in terms of the new ftd hook. 2007-01-23 14:20 gb * [r5780] : *** empty log message *** 2007-01-23 14:19 gb * [r5779] : ppc32, ppc64 also had trouble with (CLASS-OF non-simple-string). 2007-01-23 14:02 gb * [r5778] : x8664 class-of: typo in non-simple array case, needed new string definition. 2007-01-23 14:01 gb * [r5777] : Define ffi stuff in terms of common things in compiler/X86/X8664/x8664-backend. 2007-01-23 13:59 gb * [r5776] : Define (common) x8664 ffi stuff here. 2007-01-21 16:20 gb * [r5775] : *** empty log message *** 2007-01-21 15:32 gb * [r5774] : Name vinsn ff-call-return-registers (not "-returning-"). Typecheck the :registers arg. 2007-01-21 15:32 gb * [r5773] : ff-call-return-registers. 2007-01-21 15:31 gb * [r5772] : WITH-FFCALL-RESULTS: zero the stack-block. 2007-01-21 15:31 gb * [r5771] : Don't use save3 in .SPffcall_return_registers (save3 = r11); use save2 (= r13) instead. 2007-01-21 14:20 gb * [r5770] : *** empty log message *** 2007-01-21 13:46 gb * [r5769] : WITH-FFCALL-RESULTS needs space for 13 float registers, not 31. 2007-01-21 13:45 gb * [r5768] : WITH-FFCALL-RESULTS for x86-64. 2007-01-21 13:38 gb * [r5767] : ff-call: handle n-word struct-by-value, register return without typos. 2007-01-21 13:20 gb * [r5766] : *** empty log message *** 2007-01-21 13:15 gb * [r5765] : ff-call: handle n-word struct-by-value, register return. 2007-01-21 13:14 gb * [r5764] : Don't assume that passing a structure by value in N words is Darwin-specific. (LinuxPPC32 is probably the only platform that doesn't support the convention.) 2007-01-21 13:12 gb * [r5763] : .SPffcall-return-registers replaces .SPunused-4. 2007-01-21 13:10 gb * [r5762] : .SPffcall_return_registers (just returns %rax/rdx, %xmm0/%xmm1). (replaces .SPunused_4). 2007-01-21 07:20 gb * [r5761] : *** empty log message *** 2007-01-21 07:05 gb * [r5760] : Use argument/result coercion functions in os::expand-ff-call. 2007-01-21 06:53 gb * [r5759] : Coerce arg, results when expanding an ff-call. 2007-01-21 06:52 gb * [r5758] : Default (no-op) functions for coercing foreign args, results. 2007-01-21 06:20 gb * [r5757] : *** empty log message *** 2007-01-21 05:31 gb * [r5756] : 4 down. Whether 2 or 3 to go depends on your perspective (and on how close x86-64 Linux and x86-64 FreeBSD are to each other. I'd guess that they're pretty close, and that x86-64 Darwin is random.) 2007-01-21 05:20 gb * [r5755] : *** empty log message *** 2007-01-21 04:55 gb * [r5754] : Pretty simple. 3 down, 4 to go. 2007-01-21 03:20 gb * [r5753] : *** empty log message *** 2007-01-21 02:34 gb * [r5752] : Package prefixes. D'oh! 2007-01-21 02:25 gb * [r5751] : Add references to the right os::ff-call-expander. 2007-01-21 02:24 gb * [r5750] : Remove some wip code. 2007-01-21 02:24 gb * [r5749] : Use os::expand-ff-call, for each value of OS. 2007-01-21 02:23 gb * [r5748] : Saw an opportunity to use PAREF; probably several more. 2007-01-21 02:22 gb * [r5747] : Two down, 5 to go. 2007-01-21 02:21 gb * [r5746] : Looks mostly right; might need to be extended to handle coercions at the Cocoa level, and probably needs some cooperation from %EXTERNAL-CALL-EXPANDER (e.g., what's installed by #_). 2007-01-21 02:20 gb * [r5745] : *** empty log message *** 2007-01-21 02:18 gb * [r5744] : Encode, decode union types in arglists. 2007-01-21 02:17 gb * [r5743] : ftd's :ff-call-expand-function is os::expand-ff-call. Canonicalize the :void type a bit better. 2007-01-21 02:16 gb * [r5742] : Remove the debugging FORMAT call. 2007-01-19 06:20 gb * [r5741] : *** empty log message *** 2007-01-19 06:15 gb * [r5740] : Load with the new ffi-backend fasls. 2007-01-19 06:15 gb * [r5739] : Deal with the new ffi-backend sources. 2007-01-19 06:15 gb * [r5738] : Add the new ffi-backend sources. 2007-01-19 06:14 gb * [r5737] : New (so far, empty) files. 2007-01-18 12:20 gb * [r5736] : *** empty log message *** 2007-01-18 11:35 gb * [r5735] : Bump. 2007-01-18 11:20 gb * [r5734] : *** empty log message *** 2007-01-18 11:17 gb * [r5733] : Some word-size conditionalization. Will need more work to handle struct return/asssignment correctly on all platforms. 2007-01-18 11:15 gb * [r5732] : Use CGFloat, type casts, etc. Likwise for :nteger, :nteger. 2007-01-18 11:13 gb * [r5731] : Use CGFloat, type casts, etc. 2007-01-18 11:12 gb * [r5730] : Use CGFloat, type casts, etc. Avoid deprecated NSFont stuff. 2007-01-18 11:11 gb * [r5729] : Handle ivars, class creation/initialization differently for ObjC 2.0. 2007-01-18 11:10 gb * [r5728] : More ObjC 2.0, ppc64 changes. 2007-01-18 11:09 gb * [r5727] : Note which protocols are actually used when walking classes. (There doesn't seem to be a good way of finding a Protocol object pre-objc-2.0, and we may need to test protocol compliance to disambiguate message sends in some cases.) 2007-01-18 11:07 gb * [r5726] : Conditionalize for darwin (not necessarily darwinppc). 2007-01-18 11:06 gb * [r5725] : Conditionalize for darwin (not necessarily darwinppc). May want to conditionalize for bundle executable name, as well. 2007-01-18 11:05 gb * [r5724] : WITH-FFCALL-RESULTS. Start to sketch out structure return code (structures are often "returned" by returning field values in registers pn DarwinPPC64.) 2007-01-18 11:04 gb * [r5723] : New ftd stuff; new attributes for structure args/results. 2007-01-18 11:03 gb * [r5722] : ffcall-returning registers. (Will need this for other platforms, too.) 2007-01-18 11:01 gb * [r5721] : Arch-specific macro interface for WITH-FFCALL-RESULTS. 2007-01-18 11:00 gb * [r5720] : Handle the case of an encoded structure type as a return value. 2007-01-18 11:00 gb * [r5719] : .SPpoweropen_ffcall_return_registers. 2007-01-18 10:59 gb * [r5718] : At least optionally, make structure return explicit (expecting it to be handled by higher-level code.) 2007-01-18 10:58 gb * [r5717] : ff-call-expand-function in ftd. Don't try to type-expand :registers form in EXTERNAL-CALL. 2007-01-18 10:55 gb * [r5716] : Get the contents of the register that may hold OS thread context. 2007-01-18 10:54 gb * [r5715] : ff-call-expander function in ftd. 2007-01-18 10:53 gb * [r5714] : WITH-FFCALL-RESULTS arch macro. 2007-01-18 10:52 gb * [r5713] : Poweropen ff-calls might need to return all possible result regs. (Maybe poweropen syscall, too.) 2007-01-18 10:51 gb * [r5712] : Replace unused subprim with .SPpoweropen-ff-call-return-registers. 2007-01-18 10:50 gb * [r5711] : Support :registers in %ff-call; when supported, backend will return all (potential) result registers in buffer at . 2007-01-16 09:20 gb * [r5709] : *** empty log message *** 2007-01-16 09:01 gb * [r5708] : Make sure that STREAM-READ-IVECTOR and STREAM-WRITE-IVECTOR are defined on binary stream classes. (These functions make less sense for character streams/strings than they used to.) 2007-01-16 08:39 gb * [r5707] : Ensure that the :FORMAT argument defaults to :BIVALENT 2007-01-16 08:37 gb * [r5706] : Put :OPENMCL-UNICODE-STRINGS on *FEATURES*. 2007-01-12 07:20 gb * [r5705] : *** empty log message *** 2007-01-12 06:58 gb * [r5704] : Transform (/ (* constant1 exp) constant2) into (* (/ constant1 constant2) exp). Not very general, but makes some foreign address arithmetic better. 2007-01-12 06:56 gb * [r5703] : Simple PAREF: (paref pointer foreign-pointer-or-array-type index) does simple arithmetic for you. 2007-01-12 06:54 gb * [r5702] : %FOREIGN-ARRAY-ACCESS-FORM. 2007-01-10 23:20 gb * [r5701] : *** empty log message *** 2007-01-10 22:50 gb * [r5700] : ppc2-vref1: lose extra arg to double->heap vinsn. 2007-01-09 19:20 gb * [r5699] : *** empty log message *** 2007-01-09 18:33 gb * [r5698] : When running an unwind-protect cleanup and multiple values are being thrown, be sure to set tcr.xframe from the target catch frame. This is necessary on general principles (GC safety), and especially necessary if the cleanup form does a non-local exit. 2007-01-09 05:20 gb * [r5697] : *** empty log message *** 2007-01-09 05:12 gb * [r5696] : Revive some of the old GNUstep support; add partial ObjC 2.0 support (for 64-bit targets only, so far.) 2007-01-09 05:11 gb * [r5695] : Conditionalize for darwin-target, not darwinppc-target. 2007-01-09 05:10 gb * [r5694] : Disable background layout in the listener; the whole two-thread scheme probably isn't viable, but doing this seems to suppress a timing-related symptom. 2007-01-09 05:09 gb * [r5693] : Document-begin-editing, document-end-editing: both seem to have been conditionalized backwards. 2007-01-06 00:20 gb * [r5692] : *** empty log message *** 2007-01-05 23:22 gb * [r5691] : Fix another typo ... 2007-01-05 23:20 gb * [r5690] : *** empty log message *** 2007-01-05 23:16 gb * [r5689] : Fix typo (ppc32::, not ppc2::). 2007-01-05 22:27 gb * [r5688] : Missing MISC-SET-C-S32 for PPC32. 2007-01-05 14:20 gb * [r5687] : *** empty log message *** 2007-01-05 13:20 gb * [r5686] : OPEN-SHARED-LIBRARY: freebsd version of #_dlopen doesn't return a link map, but calling #_dlinfo on the result (with the right arg) does obtain a link map. Revive shared libraries on FreeBSD, but don't do the strange Linux-specific re-opening there. (It seems like it's possible to #_dlopen something that already has a loaded link map - like the FreeBSD dynamic linker - or something.) 2007-01-04 13:20 gb * [r5685] : *** empty log message *** 2007-01-04 12:46 gb * [r5684] : %component-match-p: handle :WILD, :WILD-INFERIORS. 2007-01-04 12:45 gb * [r5683] : Try to check for old libc versions (which may use %gs instead of %fs for tls). 2007-01-04 12:44 gb * [r5682] : Explicitly require NUMBER-CASE-MACRO. 2007-01-04 12:44 gb * [r5681] : fixnum-sfloat-compare,fixnum-dfloat-compare: if the fixnum can be accurately represented in float, do so. 2007-01-03 11:20 gb * [r5680] : *** empty log message *** 2007-01-03 10:35 gb * [r5679] : Fix a few type errors, check for explicit vector-type length earlier. 2007-01-03 10:31 gb * [r5678] : %STD-NAME-COMPONENT, %STD-TYPE-COMPONENT: preserved :WILD, don't map it to "*". 2007-01-03 10:30 gb * [r5677] : More whack-a-mole with pathnames (specifically, preserving :WILD as name/type value and dealing with the consequences of doing so.) 2007-01-03 10:27 gb * [r5676] : Small (mostly cosmetic) change to RATIONAL. 2007-01-02 17:20 gb * [r5675] : *** empty log message *** 2007-01-02 16:57 gb * [r5674] : Keep fixing CTYPECASE. 2007-01-02 16:57 gb * [r5673] : DOH! STREAM-READ-LINE method on BASIC-CHARACTER-INPUT-STREAM returns both values. "Composite" (two-way, concatenated, etc) streams need a close protocol. 2007-01-02 14:20 gb * [r5672] : *** empty log message *** 2007-01-02 14:00 gb * [r5671] : EQUAL on physical pathnames ignores the (new) version slot, as does practically everything else. 2007-01-02 13:59 gb * [r5670] : Don't assume that fixnums can be accurately coerced to floats when comparing fixnums and floats; do the logical equvalent of calling RATIONAL on the float instead. 2007-01-02 07:20 gb * [r5669] : *** empty log message *** 2007-01-02 06:41 gb * [r5668] : ENOUGH-NAMESTRING: don't incorporate version into returned namestring for physical pathnames. 2006-12-31 17:20 gb * [r5667] : *** empty log message *** 2006-12-31 17:15 gb * [r5666] : Update. 2006-12-31 16:20 gb * [r5665] : *** empty log message *** 2006-12-31 16:13 gb * [r5664] : Bump image/abi version. 2006-12-31 15:57 gb * [r5663] : Bump fasl-min-vers. 2006-12-31 15:49 gb * [r5662] : Bump current fasl version. 2006-12-31 15:33 gb * [r5661] : Tweak the executable start address. 2006-12-31 10:20 gb * [r5660] : *** empty log message *** 2006-12-31 09:44 gb * [r5659] : On Darwin, use a .space directive to ensure that there's nothing interesting in the first few pages of the address space (and ensure that those pages are mapped.) 2006-12-31 09:43 gb * [r5658] : Sadly, we can't use -sectalign on x86-64 (ld64 does some floating-point arithmetic, fails to account for roundoff error, and claims that 0x1000 isn't a power of 2.) If you can't say anything good about a linker, it's best not to say anything at all. We can force the first segment to be mapped at 0x1000, and can waste a few pages to ensure that all low-memory pages that the lisp wants are mapped. 2006-12-31 09:33 gb * [r5657] : %SPLIT-COMPONENT: be a little more defensive. 2006-12-31 09:32 gb * [r5656] : NX-FORM-TYPE: handle quoted constants. 2006-12-31 09:29 gb * [r5655] : Bump $fasl-vers. 2006-12-31 09:28 gb * [r5654] : Bump; remove "Beta" designation. 2006-12-30 03:20 gb * [r5653] : *** empty log message *** 2006-12-30 03:13 gb * [r5652] : Move NIL up a page on x86-64 as well. 2006-12-29 23:20 gb * [r5651] : *** empty log message *** 2006-12-29 23:09 gb * [r5650] : Be a little clearer about where we're .orged. 2006-12-29 22:32 gb * [r5649] : Static area (including NIL) now one page higher on ppc32, as well. 2006-12-29 10:20 gb * [r5648] : *** empty log message *** 2006-12-29 09:38 gb * [r5647] : SQRT in LAP (todo: inline it.) 2006-12-29 09:20 gb * [r5646] : *** empty log message *** 2006-12-29 08:38 gb * [r5645] : Move ppc64::nil-value (and ppc64::reservation-discharge) up a page. 2006-12-29 08:37 gb * [r5644] : xload-static-space-address for ppc64 is #x2000; still #x1000 for ppc32. (need to set this for x86-64) 2006-12-29 08:36 gb * [r5643] : *xload-static-space-address* depends on the backend. ppc64 wants it one page higher than it has been, to avoid letting unmapped pages get claimed by the dynamic linker on Darwin. 2006-12-29 08:35 gb * [r5642] : Change seg1addr/sectalign, so that the image starts at #x1000 and aligned section data for the text section starts on the next page. 2006-12-29 08:33 gb * [r5641] : Org the spjump table 3 pages from where the aligned text section data starts. 2006-12-29 08:33 gb * [r5640] : nil_value, etc a page higher. 2006-12-29 08:32 gb * [r5639] : Don't remap ppc64 spjump table if it's already where it should be. 2006-12-29 08:32 gb * [r5638] : Access lisp_globals/nrs_symbols a page higher. 2006-12-29 08:30 gb * [r5637] : Move STATIC_BASE_ADDRESS up a page for ppc64. 2006-12-22 09:20 gb * [r5636] : *** empty log message *** 2006-12-22 08:40 gb * [r5635] : Ensure that sqrt() is prototyped. 2006-12-22 08:39 gb * [r5634] : Insist on at least Panther for PPC32/Darwin. Use dlopen() on 64-bit Darwin variants. Might want to do a runtime test and use dlfcn functions if at least Tiger. (The old dyld functions are deprecated under Leopard, and there's reason to believe that they don't work correctly on at least ppc64/ Leopard; attempting to use NSCreateObjectFileImageFromFile to open a 64-bit .dylib seems to lead to memory corruption rather than returning NSObjectFileImageInappropriateFile. The old functions work better than the dlfcn functions on pre-Tiger systems.) 2006-12-20 20:08 gb * [r5633] : *** empty log message *** 2006-12-19 08:24 gb * [r5632] : Just to be on the safe side (new pathname bugs, or old ?), append -BAK to (NAMESTRING (TRUENAME path)), not just (NAMESTRING path) when renaming. (MERGE-PATHNAMES call is producing a pathane with a non-null VERSION component, as it arguably should have been all along. 2006-12-18 20:22 gb * [r5631] : COERCE/MAKE-SEQUENCE try to simplify vector types into concrete vector types. 2006-12-18 20:04 gb * [r5630] : Try (halfheartedly) to maintain PATHNAME-VERSION of physical pathnames. (Totally meaningless, but failing to do this leads to many test failures.) 2006-12-18 20:03 gb * [r5629] : Don't call out to do sqrt. 2006-12-18 19:57 gb * [r5628] : Try (halfheartedly) to maintain PATHNAME-VERSION of physical pathnames. (Totally meaningless, but failing to do this leads to many test failures.) 2006-12-18 19:54 gb * [r5627] : Do single/double-float sqrt inline. 2006-12-16 12:28 gb * [r5626] : Recognize that fsqrt/fsqrts instructions have a minor opcode in frb and only have one source operand. 2006-12-16 12:27 gb * [r5625] : If we get an FP exception during a SQRT, note that there's only one operand and that it's in frb. 2006-12-16 12:25 gb * [r5624] : Handle (some) unimplemented instructions, notably fsqrt/fqrts, which are optional and not implemented on some PPC machines. 2006-12-16 12:24 gb * [r5623] : Add get_fpscr(). 2006-12-15 02:47 gb * [r5622] : Need to handle SIGQUIT on ppc, too. 2006-12-14 23:43 gb * [r5621] : Constraints on some double-float arithmetic vinsns should say :double-float, not :single-float. (I suspect cut-and-paste, frankly.) 2006-12-14 23:25 gb * [r5620] : *initial-lisp-thread*, not initial-thread. 2006-12-14 23:24 gb * [r5619] : Use frame pointer when signaling error. 2006-12-14 23:14 gb * [r5618] : Add a %PHYSICAL-PATHAME-VERSION slot to physical pathnames. (Yes, this is kind of silly.) 2006-12-14 23:01 gb * [r5617] : Recognize the kernel's attempt to signal errors for (e.g.) failed memory allocation. 2006-12-14 23:00 gb * [r5616] : If a memory allocation request fails, signal a lisp error (what a concept.) 2006-12-14 22:51 gb * [r5615] : Define DARWIN_USE_PSEUDO_SIGRETURN. Might be worth revisiting this. 2006-12-14 22:50 gb * [r5614] : Go back to using the pseudo_sigreturn mechanism on Darwin, since doing a real sigreturn from a fake context may not be reliable. When handling an interrupt with a partially created temp stack frame, do the right thing. (That basically means reset the frame before calling out and recreating it if we return from the callout.) Handle SIGQUIT on the altstack, to provide a way of forcibly terminating threads that's a bit more GC-safe than doing it via #_pthread_cancel. Note the tcr of the gc thread on gc entry; if an exception happens in that thread, print a message and sleep for 60 seconds. Note that there's still a race condition here: if thread A gets an exception and thread B gets the mach_exception_lock (to GC) before the exception thread can handle the message, we deadlock (thread A would be suspended in the kernel and can't respond to a suspend signal.) We can tell that this has happened and make a good guess as to why it happened, but dealing with it will require some thought. (If we've already issued a suspend signal and eventually raise the suspended thread's resume semaphore, it'll suspend and resume once the mach exception thread makes it runnable. Who cleans up its exception context in that case ? Does it matter ?) 2006-12-14 22:19 gb * [r5613] : Ensure that newly-created threads have all signals unmasked. If a tcr is dead, ensure that its stack areas are empty. Try to run shutdown_thread_tcr as a pthread cleanup action, hopefully before Mach clobbers the thread's signal mask. 2006-12-14 22:16 gb * [r5612] : lazarus() runs with a lock on the TCR_AREA_LOCK. 2006-12-14 22:15 gb * [r5611] : Don't declare do_memo_reset(), which hasn't existed in years. 2006-12-14 22:14 gb * [r5610] : Bump. 2006-12-14 22:14 gb * [r5609] : PREPARE-TO-QUIT: wait for the termination semaphore. 2006-12-14 22:13 gb * [r5608] : PROCESS objects are created with a termination semaphore. 2006-12-14 22:12 gb * [r5607] : THREAD-ENABLE: second arg is TERMINATION-SEMAPHORE. KILL-LISP-THREAD: don't use #_pthread_cancel, which may force the target thread to transition from running lisp code to running foreign code without advertising that fact. Send a SIGQUIT instead, and trust the SIGQUIT handler to manage that transition. Don't try to kill the initial thread. 2006-12-09 16:52 gb * [r5606] : STRING-OUTPUT-STREAMs need to override their CLEAR-OUTPUT methods. MAKE-STRING-INPUT-STREAM: handle displaced strings better (remember their displaced-index-offsets.) 2006-12-09 16:49 gb * [r5605] : Disable underflow traps in new threads on x86-64. (The PPC never got around to enabling them.) 2006-12-09 16:47 gb * [r5604] : WITH-OUTPUT-TO-STRING: if string provided, evaluate the ELEMENT-TYPE expression (for side-effects) and ignore the result of its evaluation. Some fixes in CTYPECASE handling, which hopefully didn't break anything else. 2006-12-09 16:36 gb * [r5603] : In aset2/aset3, ensure that the (final, 1d) index register is set - even if the index is a constant - if we think that the store needs EGC memoization. (Note that the test case that exposed this involved storing SINGLE-FLOATs, which don't need EGC write-barrier handling on 64-bit platforms.) In aref2, don't worry about reserving unscaled index registers when targeting the 1d index; the unscaled index registers aren't live after, could be NULL if the index is a constant, and any overlap between one of them and the 1d index should be handled in the vinsn(s). 2006-12-09 16:30 gb * [r5602] : Ensure that the (relatively new) built-in-classes SIMPLE-FIXNUM-VECTOR and FIXNUM-VECTOR have class-translations. 2006-12-09 16:27 gb * [r5601] : In list case of ELT: who says that value returned by NTHCDR is a LIST ? (It's not necessarily a list, so don't segfault taking its CAR without checking.) 2006-12-07 01:48 gb * [r5600] : x86-64 version of SFLOAT-SIGNIFICAND-ZEROS: count zero bits in the shifted float, not in %imm1 (which is undefined here ...) 2006-12-07 00:05 gb * [r5599] : DEFINE-COMPILER-MACRO: if a SETF name, use the right symbol for the named block. COMPILER-MACRO-FUNCTION, (SETF COMPILER-MACRO-FUNCTION): deal with (SETF NAME). 2006-12-06 05:44 gb * [r5598] : Update for 061205. 2006-12-06 03:31 gb * [r5597] : Change LDFLAGS, to keep ld64 from putting a linkedit segment in the wrong place and making 32-bit memory unmappable. 2006-12-06 03:30 gb * [r5596] : Bump. 2006-12-06 03:29 gb * [r5595] : Use _exit(), just in case. 2006-12-06 01:53 gb * [r5594] : *ALTERNATE-LINE-TERMINATOR* set to #\Return on #+darwin-target, not just #+darwinppc-target. Change doc string for *ALTERNATE-LINE-TERMINATOR*. 2006-12-04 22:13 gb * [r5593] : Append a patchlevel to the version string, at least until new tarballs can be built. 2006-12-04 22:12 gb * [r5592] : If stream element type upgrades to FIXNUM, upgrade it further (to largest supported SIGNED-BYTE element-type.) 2006-12-04 20:02 gb * [r5591] : BIGNUM-BIGNUM-GCD (whatever it's called): don't bother to shift by 0 bits. Do shift the initial U if non-zero trailing 0 BITS (was testing for non-zero trailing DIGITS.) 2006-12-04 19:52 gb * [r5590] : Choke earlier on unsupported stream element-types; do some upgrading to support some more element types (BIT -> (UNSIGNED-BYTE 8)). Maybe support arbitrarily large integer types someday soon. Maybe not. 2006-12-04 19:48 gb * [r5589] : Make built-in classes for istruct types (compiler stuff, mostly). 2006-12-04 19:46 gb * [r5588] : Ensure that %ASHL handles shift count of 32 (and returns 0 in that case.) 2006-12-04 19:46 gb * [r5587] : In BIGNUM-SHIFT-RIGHT-LOOP-1, use %ASHL, not %ILSL. (Don't want nonsense in the high 32 bits, just in case.) 2006-12-04 19:43 gb * [r5586] : Bugs in GCD-2. Long-standing bugs in GCD-2 ... 2006-12-04 16:26 gb * [r5585] : STREAM-CLEAR-INPUT: don't inherit method for STRING-INPUT-STREAMs. 2006-12-04 16:24 gb * [r5584] : BIGNUM-SHIFT-RIGHT-1: handle shift by 0 bits. BIGNUM-BIGNUM-GCD (whatever it's called): don't bother to shift by 0 bits. 2006-12-04 16:21 gb * [r5583] : Reset error-reentry count before doing %SLOT-UNBOUND-TRAP. 2006-12-04 16:21 gb * [r5582] : Default STREAM-POSITION method for STREAMs, returns NIL. STREAM-POSITION method for STRING-INPUT-STREAMs. 2006-12-04 16:19 gb * [r5581] : Derive ccl-directory from pathname containing heap image. 2006-12-03 14:41 gb * [r5580] : FORCE-OUTPUT, FINISH-OUTPUT on STRING-OUTPUT-STREAMs are no-ops. Need an ioblock-based character-read-vector function for string-input-streams. 2006-12-03 14:37 gb * [r5579] : Minimal DESCRIBE support for CCL::BASIC-STREAM. 2006-12-03 14:36 gb * [r5578] : If :element-type is supplied in WITH-OUTPUT-TO-STRING, don't quote it. 2006-12-03 12:11 gb * [r5577] : Define RIP-REGISTER-OFFSET, which isn't necessarily 16. 2006-12-03 12:10 gb * [r5576] : Use RIP-REGISTER-OFFSET, which isn't necessarily 16. Use WITH-ERROR-REENTRY-DETECTION n %xerr-disp; dtrt for %slot-unbound-trap. 2006-12-03 12:09 gb * [r5575] : Tweak stream hierarchy a little, to avoid some NO-NEXT-METHODs. 2006-12-03 09:42 gb * [r5574] : Handle VALUES type specifiers in THE forms a bit better. 2006-12-03 09:41 gb * [r5573] : Handle VALUES type specifiers in THE forms a bit. 2006-12-03 09:28 gb * [r5572] : Do vector bounds checking by clearing the low byte of the header, shifting it right, and comparing the index to that. (It's not reliable to shift the index left.) 2006-12-03 09:26 gb * [r5571] : Memory/vector encode functions: don't write BOM, leave that up to higher-level code. Start to write some higher-level code. 2006-12-03 09:23 gb * [r5570] : Fix some indentation. 2006-12-03 09:21 gb * [r5569] : CHECK-MISC-BOUNDS: shift the header word right (after clearing its low byte) and compare the index to that. (Shifting the index left - d'oh - obviously fails for indices of large magnitude.) 2006-11-30 04:07 gb * [r5568] : Indentation change. Add CDB-ENUMERATE-KEYS. 2006-11-30 01:37 gb * [r5567] : %DESTRUCTURE-LAMBDA-LIST: other callers of VERIFY-LAMBDA-LIST (hopefully) check to see if the alleged list is LISTP first. (I hope ...). Do so. 2006-11-29 01:42 gb * [r5566] : Wrong register typo in .SPaset2. 2006-11-28 23:14 gb * [r5565] : DEFAULT-FONT: retain the font. 2006-11-28 23:12 gb * [r5564] : Wrap a (primitive) WITH-AUTORELEASE-POOL around the code which detaches the dummy NSThread. 2006-11-28 23:09 gb * [r5563] : Don't claim that a callback parameter is ignored; use IGNORABLE instead. 2006-11-21 09:21 gb * [r5562] : *** empty log message *** 2006-11-21 08:28 gb * [r5561] : Typo in PATH-FROM-UNIX-ADDRESS conditionalization. 2006-11-21 08:20 gb * [r5560] : *** empty log message *** 2006-11-21 07:55 gb * [r5559] : INIT-UNIX-SOCKADDR: don't use %COPY-IVECTOR-TO-PTR to copy string to foreign memory. Basic file-socket streams. 2006-11-10 12:20 gb * [r5558] : *** empty log message *** 2006-11-10 12:08 gb * [r5557] : Update. 2006-11-10 11:28 gb * [r5556] : Bump min image version. 2006-11-10 11:20 gb * [r5555] : *** empty log message *** 2006-11-10 11:19 gb * [r5554] : bump current image version. 2006-11-10 11:19 gb * [r5553] : bump current, max image version. 2006-11-10 11:10 gb * [r5552] : Bump min fasl version. 2006-11-10 10:56 gb * [r5551] : bump current fasl version. 2006-11-10 10:47 gb * [r5550] : save_application: use #__exit, not #_exit 2006-11-10 10:46 gb * [r5549] : TYPE typo. 2006-11-10 10:32 gb * [r5548] : Bump max fasl version. 2006-11-10 10:25 gb * [r5547] : Bump. New snapshots today! 2006-11-10 00:20 gb * [r5546] : *** empty log message *** 2006-11-09 23:26 gb * [r5545] : QUIT uses #__quit. PREPARE-TO-QUIT kill stragglers without waiting as long. 2006-11-09 23:20 gb * [r5544] : *** empty log message *** 2006-11-09 22:50 gb * [r5543] : Use #_ _exit for fatal termination. Try to run lisp code after #_exit via lazarus(); this is for some cases of SAVE-APPLICATION (e.g., in the Cocoa GUI). The environment in which The Undead code runs is a little funky, especially on PPC. There might be GC implications (cstack will contain call history as of the point when #_exit called atexit handlers, vstack & tstack are emptied in lazarus()). We should barely run any code in this environment, but we will GC a time or two ... 2006-11-09 22:45 gb * [r5542] : Use _exit afer save-application. 2006-11-09 22:43 gb * [r5541] : Change the quit protocol. 2006-11-09 22:42 gb * [r5540] : :set-document-edited for listener "documents" is a no-op. 2006-11-09 22:39 gb * [r5539] : 3d-array accessor stuff for PPC as well. 2006-11-09 20:20 gb * [r5538] : *** empty log message *** 2006-11-09 19:50 gb * [r5537] : Always check that array-ctype's dimensions are a list before taking their length. 2006-11-09 19:20 gb * [r5536] : *** empty log message *** 2006-11-09 19:13 gb * [r5535] : 3d-unscaled-index does the right thing ... 2006-11-09 17:20 gb * [r5534] : *** empty log message *** 2006-11-09 17:08 gb * [r5533] : Three-arg aref/(setf aref) for x8664. 2006-11-09 13:20 gb * [r5532] : *** empty log message *** 2006-11-09 12:37 gb * [r5531] : Map array-type keywords to 3d-array type errors. 2006-11-09 12:37 gb * [r5530] : Operators for 3d array access. 2006-11-09 12:36 gb * [r5529] : Type-errors for 3d array types. 2006-11-08 20:20 gb * [r5528] : *** empty log message *** 2006-11-08 19:50 gb * [r5527] : In darwin Fake Structures For Leopard nonsense: define the stack context struct separately, make the ucontext stack field an instance of it. (Not -sure- if that was necessary, but :poweropen-target was accidentally on *FEATURES* and foreign struct layout was using read-time condtionalization to decide whether or not to use brain-dead structure-alignment rules leftover from AIX/PowerOpen. 2006-11-08 19:46 gb * [r5526] : Fix comment ("/*", not "/%" ...). 2006-11-08 19:40 gb * [r5525] : Try to infer/insert the array type when compiler-macroexpanding MAKE-ARRAY calls. 2006-11-08 19:39 gb * [r5524] : DarwinPPC32 FTD follows :poweropen-alignment rules. 2006-11-08 19:38 gb * [r5523] : parse-foreign-record-fields: conformance to poweropen structure layout rules should be an attribute of the FTD, not done via read-time conditionalization. 2006-11-08 12:20 gb * [r5522] : *** empty log message *** 2006-11-08 12:15 gb * [r5521] : Bump. 2006-11-08 12:15 gb * [r5520] : Some bivalent/binary fixes. TODO: get rid of "bivalent" streams, but allow switching some streams (sockets) between "character mode" and "binary mode". 2006-11-08 10:20 gb * [r5519] : *** empty log message *** 2006-11-08 09:39 gb * [r5518] : Typo in misc-ref-u64 vinsn emit. 2006-11-08 04:20 gb * [r5517] : *** empty log message *** 2006-11-08 04:10 gb * [r5516] : Transform three-argument AREF->%AREF3, ASET->%ASET3. 2006-11-08 03:40 gb * [r5515] : Implement %aset2 in the x8664 backend. 2006-11-08 03:39 gb * [r5514] : %aref3, %aset3. 2006-11-08 03:38 gb * [r5513] : Use %aref2,%aref3. 2006-11-08 03:37 gb * [r5512] : More (often redundant) vector-store vinsns. 2006-11-08 03:37 gb * [r5511] : .SParef3, .SPaset3. 2006-11-07 18:20 gb * [r5510] : *** empty log message *** 2006-11-07 17:28 gb * [r5509] : .SParef3, .SPaset3. 2006-11-07 17:27 gb * [r5508] : %ref3, %aset3. 2006-11-07 17:27 gb * [r5507] : .SPref3, .SPaset3. 2006-11-07 17:26 gb * [r5506] : Define mullr, which expands into one of mullw/mulld. 2006-11-07 11:20 gb * [r5505] : *** empty log message *** 2006-11-07 11:12 gb * [r5504] : :signed-64-bit-vector is signed ... 2006-11-07 05:20 gb * [r5503] : *** empty log message *** 2006-11-07 05:07 gb * [r5502] : nx-error-for-simple-2d-array-type: forgot the :fixnum-vector case. 2006-11-07 05:04 gb * [r5501] : Handle vector assignment differently; handle (setf (aref ...)) of simple 2d-arrays inline if the type is known/declared. 2006-11-07 05:02 gb * [r5500] : Add vinsns for misc-set-s* and variants. (Identical to the unsigned versions, but it helps not to have to remember which are implemented and which aren't.) 2006-11-07 05:01 gb * [r5499] : general-aset2, simple-typed-aset2. 2006-11-07 05:00 gb * [r5498] : New %ASET2 handling. 2006-11-07 04:59 gb * [r5497] : ACODE-FORM-TYPE: use var-inittype if var is never assigned to. 2006-11-06 05:20 gb * [r5496] : *** empty log message *** 2006-11-06 05:09 gb * [r5495] : Try to work around incompatibilites in Tiger/Leopard headers. 2006-11-06 05:08 gb * [r5494] : LOOKUP-PORT: use the macro HTONS, since the foreign function may not exist. 2006-11-06 04:20 gb * [r5493] : *** empty log message *** 2006-11-06 03:41 gb * [r5492] : 2d array error lookup -> nx. 2006-11-05 15:20 gb * [r5491] : *** empty log message *** 2006-11-05 14:55 gb * [r5490] : %VALID-REMAINING-TIMESPEC-TIME-P for x86-64, as well. 2006-11-05 14:52 gb * [r5489] : %AREF2, %ASET2 in x8664 lap. 2006-11-05 14:50 gb * [r5488] : pre-add displacement in .SParef2, .SPaset2. 2006-11-05 14:48 gb * [r5487] : .SParef2, .SPaset2: since we aren't testing simple-array-ness, add the (possibly 0) displacement from the header before handling the simple aref of the data vector. 2006-11-05 14:47 gb * [r5486] : PRINT-OBJECT method for EEPS: handle pointers on PPC(64), print 64-bit addresses on 64-bit platforms. 2006-11-05 14:46 gb * [r5485] : Bump; new binaries soon. 2006-11-05 14:46 gb * [r5484] : %NANOSLEEP calls %VALID-REMAINING-TIMESPEC-TIME-P, which is better than sleeping for 4G seconds after an EINTR on Leopard. 2006-11-05 14:44 gb * [r5483] : MACPTR->FIXNUM returns NIL if any non-fixnum bits would be set in the result. (Should probably make this true on all platforms; for PPC64, it's important for foreign function addresses that're ... suboptimally aligned.) 2006-11-05 14:42 gb * [r5482] : Try to detect the case where %NANOSLEEP's second (by reference) argument contains garbage after an EINTR. (The L word, again.) 2006-11-05 14:41 gb * [r5481] : %AREF2, %ASET2: in LAP, just call into the kernel. 2006-11-05 14:40 gb * [r5480] : Some platforms (the L-word) generate ppc64 functions whose addresses are not 0 mod 8, so (a) make MACPTR->FIXNUM return NIL if an address isn't representable as a fixnum and (b) make FOREIGN-SYMBOL-ENTRY return an address if it can't return a fixnum on PPC (really just affects PPC64.) 2006-11-05 14:37 gb * [r5479] : %aref2, %aset2: call into kernel. 2006-11-05 14:36 gb * [r5478] : 2d aref support/changes. 2006-11-05 14:36 gb * [r5477] : Get the PPC2 backend caught up with the x8664 wrt 2d aref. Stll a ways to go. 2006-11-05 14:35 gb * [r5476] : NX-ERROR-FOR-SIMPLE-2D-ARRAY-TYPE. 2006-11-04 07:20 gb * [r5475] : *** empty log message *** 2006-11-04 07:05 gb * [r5474] : Bump. 2006-11-04 06:20 gb * [r5473] : *** empty log message *** 2006-11-04 06:10 gb * [r5472] : fix parens. 2006-11-04 06:09 gb * [r5471] : arch gvector types contains :struct, not :structure. 2006-11-04 06:08 gb * [r5470] : aref2: new operator name(s), always call subprim. ppc2-fixed-call-builtin: (tail-)call subprims even if they aren't in the builtin functions table. 2006-11-03 22:20 gb * [r5469] : *** empty log message *** 2006-11-03 22:19 gb * [r5468] : New .SParef2, .SPaset2 subprims. 2006-11-03 22:18 gb * [r5467] : Comments/indentation. 2006-11-03 22:17 gb * [r5466] : Change decoding/handlng of type-error UUOs. 2006-11-03 22:16 gb * [r5465] : .SParef2, .SPaset2. 2006-11-03 22:16 gb * [r5464] : Define arrayH (array-header) layout. 2006-11-03 21:20 gb * [r5463] : *** empty log message *** 2006-11-03 20:46 gb * [r5462] : CANONICALIZE-SPECIALIZERS optionally returns its argument if all members are SPECIALIZERS. Use CANONICALIZE-SPECIALIZERS to canonicalize FIND-METHOD arguments, for the benefit of people who haven't read the MOP spec or who don't realize that the MOP defines some things differently from CLHS. 2006-11-03 20:38 gb * [r5461] : Consistently use :struct (not :structure) to refer to primitive structure type. Fix array-ctype-dimensions check in %aref1, add it to %aset1. 2006-11-03 20:37 gb * [r5460] : Ensure that :STRUCT (not :STRUCTURE) is on target-gvector-types. 2006-11-03 20:20 gb * [r5459] : *** empty log message *** 2006-11-03 19:53 gb * [r5458] : uuo_error_array_bounds, such as it is. 2006-11-03 19:52 gb * [r5457] : SParef2, .SPaset2. More to come ... 2006-11-03 19:52 gb * [r5456] : Define array-header layout. 2006-11-03 19:51 gb * [r5455] : New type-error uuo encodings, new type-error constants. 2006-11-03 19:50 gb * [r5454] : New range/encoding of type-error traps. 2006-11-03 19:49 gb * [r5453] : (temporary) mechanism to avoid constant-redefinition CERRORs. 2006-11-03 19:48 gb * [r5452] : New %type-error-typespecs%. 2006-11-03 19:47 gb * [r5451] : array--header -> array-header. 2006-11-03 19:47 gb * [r5450] : Fix misc-set-c-u64 braino; add/change some 2d array stuff. 2006-11-03 19:43 gb * [r5449] : Define .SParef2/.SPaset2 subprims; seem to be much (>5x) faster than the lisp/lap versions (though we can generally be faster still when we know the element-type and/or know whether the array is simple or explicitly displaced.) 2006-11-03 19:41 gb * [r5448] : Change handling of 2d-aref; lots of related changes. Split the code which handles bounds/type-checking in vref out from the code which actually does the reference, so that the latter can be used in multidimensional cases. Need to do this on PPC as well; need to do more (%aset2, 3d cases, general case.) 2006-11-03 19:38 gb * [r5447] : Change handling of 2d-aref. 2006-11-03 19:37 gb * [r5446] : aref2 -> simple-typed-aref2, add general-aref2. More changes to come (and this may affect compilation on PPC until some PPC compiler changes are made/checked-in.) 2006-11-03 19:35 gb * [r5445] : type-errors (as encoded in UUOs) are now in the range 128-255, rather than 64-127; more are defined. 2006-11-02 17:20 gb * [r5444] : *** empty log message *** 2006-11-02 16:43 gb * [r5443] : Move along, nothing to see here. 2006-11-02 16:37 gb * [r5442] : New error constants; work in progress ... 2006-11-01 19:20 gb * [r5441] : *** empty log message *** 2006-11-01 18:40 gb * [r5440] : Some new constants; more to check in later. 2006-10-31 11:20 gb * [r5439] : *** empty log message *** 2006-10-31 10:39 gb * [r5438] : You must be reading the OLD rule book (Leopard and #$EINTR). 2006-10-31 10:21 gb * [r5437] : handle_protection_violation: pass the tcr. Initialize "addr" before testing it ... 2006-10-31 10:21 gb * [r5436] : handle_protection_violation: pass the tcr. 2006-10-31 04:20 gb * [r5435] : *** empty log message *** 2006-10-31 03:40 gb * [r5434] : DEFSTRUCT-REF-TRANSFORM for typed vector expands into something more useful. 2006-10-31 03:39 gb * [r5433] : Fix ppc32 misc_ref_double_float_vector. 2006-10-30 08:20 gb * [r5432] : *** empty log message *** 2006-10-30 07:24 gb * [r5431] : Enable the destructive version of %BIGNUM-BIGNUM-GCD; conditionalize it for 64-bit targets and provide/fix supporting functions. Seems to work on ppc64; needs testing on x8664. 2006-10-28 00:20 gb * [r5430] : *** empty log message *** 2006-10-27 23:50 gb * [r5429] : New file. 2006-10-26 03:20 gb * [r5428] : *** empty log message *** 2006-10-26 02:36 gb * [r5427] : %BIVALENT-IOBLOCK-READ-U8-BYTE (and variants): always clear the untyi-char. %ENCODED-IOBLOCK-PEEK-CHAR: the read-char-when-locked function needs an argument. SETUP-IOBLOCK-OUTPUT: there are no special output functions for bivalent streams. %IOBLOCK-BINARY-READ-VECTOR: clear any untyi char. 2006-10-24 19:20 gb * [r5426] : *** empty log message *** 2006-10-24 18:35 gb * [r5425] : Bump. 2006-10-24 08:21 gb * [r5424] : *** empty log message *** 2006-10-24 07:29 gb * [r5423] : Fix :ALLOW-OTHER-KEYS check in _SPdestbind1 (still used in inlining, sometimes.) 2006-10-24 07:20 gb * [r5422] : *** empty log message *** 2006-10-24 07:05 gb * [r5421] : Change the date, add/change some content. 2006-10-24 07:05 gb * [r5420] : Reference the (bound) STREAM-VAR argument - not STREAM - in the expansion of WITH-OUTPUT-TO-STRING. 2006-10-24 06:56 gb * [r5419] : Reference the (bound) STREAM-VAR argument - not STREAM - in the expansion of WITH-OUTPUT-TO-STRING. 2006-10-24 06:55 gb * [r5418] : When recycling string-output-stream-ioblocks, set the IOBLOCK-CHARPOS to 0. (Failure to do this can screw up the pretty-printer, among other things.) 2006-10-24 06:52 gb * [r5417] : DEFAULT-CHARACTER-ENCODING for :SOCKET, *DEFAULT-SOCKET-CHARACTER-ENCODING*. 2006-10-23 23:20 gb * [r5416] : *** empty log message *** 2006-10-23 23:03 gb * [r5415] : Updated; almost ready to go (still a WITH-OUTPUT-TO-STRING/PPRINT (?) bug). 2006-10-23 22:43 gb * [r5414] : NORMALIZE-EXTERNAL-FORMAT: use *DEFAULT-EXTERNAL-FORMAT* value for :DEFAULT. 2006-10-23 22:42 gb * [r5413] : Use NORMALIZE-EXTERNAL-FORMAT in (SETF STREAM-EXTERNAL-FORMAT) methods. 2006-10-23 22:41 gb * [r5412] : Export some unicode/external-format stuff. 2006-10-23 15:20 gb * [r5411] : *** empty log message *** 2006-10-23 14:27 gb * [r5410] : Bump; binaries today, honest. 2006-10-23 14:27 gb * [r5409] : freelist string-output-stream-ioblocks. 2006-10-23 14:26 gb * [r5408] : Some/most/all of the WITH-IOBLOCK-OUTPUT-LOCKED macros were checking for the presence of an input lock ... 2006-10-23 12:20 gb * [r5407] : *** empty log message *** 2006-10-23 11:40 gb * [r5406] : Optimize write-string on string streams a little. 2006-10-23 11:26 gb * [r5405] : WITH-OUTPUT-TO-STRING: don't expand into code which evaluates the string form twice; don't do whatever was being done to default the element type of the string stream. 2006-10-23 11:24 gb * [r5404] : MAKE-STRING-OUTPUT-STREAM: skip some keyword/subtype checking if element-type not provided. (Could also skip it if it's explicit and obviously OK.) 2006-10-23 11:22 gb * [r5403] : %HEAP-BYTES-ALLOCATED: don't add "current" allocation if nothing's reserved. 2006-10-23 11:20 gb * [r5402] : *** empty log message *** 2006-10-23 11:19 gb * [r5401] : In handle_alloc_trap(), fix call to update_bytes_allocated (sign of "disp"). 2006-10-22 14:20 gb * [r5400] : *** empty log message *** 2006-10-22 14:02 gb * [r5399] : MULTIPLE-VALUE-PROG1: treat the first form like a PROGN if there are no other forms. 2006-10-22 14:01 gb * [r5398] : Don't disable interrupts when calling ioblock functions. New string-stream implementation. Better ... 2006-10-22 13:59 gb * [r5397] : If WITH-OUTPUT-TO-STRING gets the string, do so before the stream closes. 2006-10-22 13:59 gb * [r5396] : Don't assume that string output stream strings have a fill-pointer; use STREAM-POSITION instead. 2006-10-22 13:57 gb * [r5395] : Don't disable interrupts when calling ioblock functions, after all. Go through (new) ioblock-unread-char-function. 2006-10-22 13:55 gb * [r5394] : Stop trying to print strings via x/s 2006-10-22 01:20 gb * [r5393] : *** empty log message *** 2006-10-22 00:46 gb * [r5392] : Typo. 2006-10-22 00:41 gb * [r5391] : Initialize (zero, usually) more fields in pseudo sigcontexts. 2006-10-22 00:40 gb * [r5390] : Mostly locking changes (don't lock twice, etc.) 2006-10-22 00:36 gb * [r5389] : Remove vestival #+target-8-bit-chars stuff. 2006-10-21 23:20 gb * [r5388] : *** empty log message *** 2006-10-21 22:45 gb * [r5387] : New! Improved! Now with more iso-8859 encodings! 2006-10-19 20:20 gb * [r5386] : *** empty log message *** 2006-10-19 19:48 gb * [r5385] : Reference the source argument - not arg_z - in %safe-get-ptr. 2006-10-19 19:22 gb * [r5384] : %CSTR-POINTER: make #\Nul termination optional, use #\Sub char-code as replaement char. 2006-10-19 19:21 gb * [r5383] : Missing close paren. 2006-10-19 19:20 gb * [r5382] : *** empty log message *** 2006-10-19 19:20 gb * [r5381] : Careful BLTing strings to/from foreign memory. 2006-10-19 19:18 gb * [r5380] : Typos in UNBOX-S8. 2006-10-18 14:20 gb * [r5379] : *** empty log message *** 2006-10-18 14:04 gb * [r5378] : Bump. Binaries later today ? 2006-10-18 13:38 gb * [r5377] : The post-bootstrapping version of (SETF FIND-CLASS) only introduces (or removes) a type name if the class becomes (or ceases to be) a properly named class. The pre-bootstrapping version handles simpler cases of that; use ALIAS-CLASS to register some early class names as type names. 2006-10-18 13:36 gb * [r5376] : (SETF CLASS-NAME) can affect the type system, since only symbols which name properly-named classes are type specifiers. Changing the name of a class can either introduce or remove a properly-named class. Classes that have the kernel-p attribute shouldn't have their names changed casually. 2006-10-18 13:33 gb * [r5375] : Accept an :EXTERNAL-FORMAT argument to MAKE-SOCKET, pass it down to whatever function actually creates a stream socket. 2006-10-18 08:21 gb * [r5374] : *** empty log message *** 2006-10-18 08:12 gb * [r5373] : Allow closed-over addresses in conditional-form, NOT. 2006-10-18 07:43 gb * [r5372] : handle_fault() checks for safe_ref_addres. 2006-10-18 07:43 gb * [r5371] : %safe-get-ptr for x8664. 2006-10-18 07:41 gb * [r5370] : Compare stack addresses/vcells to nil. 2006-10-18 07:41 gb * [r5369] : Stuff for comparing memory (stack) addresses to NIL. Mark arg_z and rcx/nargs as being in use when restoring interrupt level inline. 2006-10-18 07:39 gb * [r5368] : We've had UUO-ERROR-UNBOUND for a few months now. The disassembler is always the last to know. 2006-10-18 01:20 gb * [r5367] : *** empty log message *** 2006-10-18 01:16 gb * [r5366] : add safe-ref-address to tcr. Handle memory faults to safe-ref-addres on PPC (so far). Try using SAFE-GET-PTR in objc runtime (untested). 2006-10-18 00:20 gb * [r5365] : *** empty log message *** 2006-10-17 23:34 gb * [r5364] : ppc backend %aref1/%aset1: notice array type in acode. 2006-10-17 20:20 gb * [r5363] : *** empty log message *** 2006-10-17 20:02 gb * [r5362] : BOM-ENCODING. 2006-10-17 20:02 gb * [r5361] : FILE-STRING-LENGTH actually does something. Clear the pending-bom flag when seeking. 2006-10-17 20:01 gb * [r5360] : (SETF STREAM-EXTERNAL-FORMAT) for ioblock-based character streams. 2006-10-17 19:45 gb * [r5359] : bind/unbind *interrupt-level* inline. 2006-10-17 19:44 gb * [r5358] : Act like ud2a takes a one-byte immmediate operand (xuuo). 2006-10-17 19:39 gb * [r5357] : Disable interrupts when necessary. 2006-10-17 19:39 gb * [r5356] : bind/unbind *interrupt-level* inline. 2006-10-16 00:20 gb * [r5355] : *** empty log message *** 2006-10-15 23:51 gb * [r5354] : Handle 32-bit character encodings. Start to extend STREAM-EXTERNAL-FORMAT, start to make it SETFable. 2006-10-15 23:51 gb * [r5353] : utf-32/ucs-4 and variants. 2006-10-15 23:50 gb * [r5352] : EXTERNAL-FORMATs are immutable, interned. 2006-10-14 14:20 gb * [r5351] : *** empty log message *** 2006-10-14 13:41 gb * [r5350] : Fix use of ^,<- vinsn macros inside ENSURING-NODE-TARGET. Inline the fixnum case of two-arg numeric comparisons. 2006-10-14 13:39 gb * [r5349] : INLINE-SUB2 support. 2006-10-14 13:20 gb * [r5348] : *** empty log message *** 2006-10-14 12:30 gb * [r5347] : Don't (DECLARE (OPTIMIZE (SAFETY 3))) in generated DEFSTRUCT accessors. 2006-10-14 12:27 gb * [r5346] : Fix many cases of ENSURING-NODE-TARGET. Do the fixnum case of two-argument subtraction inline. Inline the fixnum case of two-argument numeric comparisons. STRUCT-REF avoids typechecking when reckless. 2006-10-14 12:21 gb * [r5345] : It makes no sense to use the vinsn macros <- and ^ inside ENSURING-NODE-TARGET, so make it check for that. 2006-10-14 12:20 gb * [r5344] : *** empty log message *** 2006-10-14 12:20 gb * [r5343] : Do vector bounds-checking a little differently. Add xchg-registers. 2006-10-13 11:20 gb * [r5342] : *** empty log message *** 2006-10-13 10:54 gb * [r5341] : Use new NUMCMP operator. 2006-10-13 10:53 gb * [r5340] : Change inline expansion: use optimize settings in effect at call time, not those in effect at definition time. 2006-10-13 10:50 gb * [r5339] : Change >=-2, <=-2, /=-2 to avoid inlining issues. 2006-10-13 10:30 gb * [r5338] : Replace unused num= with numcmp. 2006-10-13 02:20 gb * [r5337] : *** empty log message *** 2006-10-13 01:46 gb * [r5336] : Terminate argument processing when "--" is seen. 2006-10-13 01:45 gb * [r5335] : Handle newline translation differenly (at the character I/O level, not the buffer level). All character encoding/decoding functions operate on octets, not necessarily code units. (The stream encode/decode functions are an exception; serialization and byte ordering are handled by the stream.) 2006-10-09 23:20 gb * [r5334] : *** empty log message *** 2006-10-09 22:34 gb * [r5333] : Write #\Sub's char-code if we can't encode a character in (implicit) iso-8859-1. 2006-10-09 22:33 gb * [r5332] : %COPY-STRING-TO-U8: use #\Sub (ascii 26) as a replacement character. 2006-10-09 22:31 gb * [r5331] : :stream-encode/decode functions always write/read something, even if it's a replacement character of some sort. 2006-10-09 21:20 gb * [r5330] : *** empty log message *** 2006-10-09 21:19 gb * [r5329] : More stuff about line termination. When will this end ? 2006-10-09 21:19 gb * [r5328] : Use a hashtable to map characters to their names. (Maybe a sparse vector ?). 2006-10-09 21:18 gb * [r5327] : Use a hashtable to map characters to their names. (Maybe a sparse vector ?). Define #\Sub and #\Replacement_Character. 2006-10-09 03:20 gb * [r5326] : *** empty log message *** 2006-10-09 03:17 gb * [r5325] : SELECT-BACKTRACE: it's been all NEW-BACKTRACE for a while now. 2006-10-09 03:16 gb * [r5324] : Need to #include to keep Panther happy. 2006-10-09 02:30 gb * [r5323] : Print hex with leading #x, not #0x ... 2006-10-09 02:30 gb * [r5322] : Do some sanity checks on SAVE-APPLICATION pathname, before it's too late ... 2006-10-09 02:29 gb * [r5321] : ASCII, some other changes. 2006-10-09 02:29 gb * [r5320] : Rearrange a few things; pass :line-termination option to MAKE-FD-STREAM. 2006-10-09 02:28 gb * [r5319] : Yet another line-termination translation scheme; hopefully, the last for a while. (More tedium before it's actually installed.) 2006-10-09 02:27 gb * [r5318] : Add character names for #\Line_Separator, #\Paragraph_separator. 2006-10-08 03:20 gb * [r5317] : *** empty log message *** 2006-10-08 02:35 gb * [r5316] : *TERMINAL-CHARACTER-ENCODING-NAME*. 2006-10-08 02:34 gb * [r5315] : In %ERROR, go straight to the kernel debugger if interactive streams aren't initialized yet. (We'd probably get there sooner or later in that case; try to get there sooner.) 2006-10-08 02:33 gb * [r5314] : Don't initialize interactive streams via DEF-LOAD-POINTERS. Wrap that up in a function, and set a flag when the function's called (from APPLICATION's TOPLEVEL-FUNCTION method.) 2006-10-08 02:32 gb * [r5313] : Handle the (new) -K/--terminal-encoding command-line argument. 2006-10-07 17:20 gb * [r5312] : *** empty log message *** 2006-10-07 17:14 gb * [r5311] : INPUT-STREAM-P, OUTPUT-STREAM-P are true of instances if INPUT-STREAM/ OUTPUT-STREAM, respectively (not just instances of FUNDAMENTAL-INPUT/OUTPUT-STREAM). 2006-10-07 17:13 gb * [r5310] : Do abort() if kernel debugger is entered when --batch is in effect. 2006-10-07 17:12 gb * [r5309] : QUIT: flush *STDOUT* in PROCESS-EXIT-APPLICATION. 2006-10-07 17:11 gb * [r5308] : PROCESS-EXIT-APPLICATION: flush *stdout* here. 2006-10-06 11:20 gb * [r5307] : *** empty log message *** 2006-10-06 10:49 gb * [r5306] : CODE-UNIT, not CODE-ELEMENT. Add a TRANSLATE slot to IO-BUFFER. 2006-10-06 10:40 gb * [r5305] : basic (vs fundamental) socket streams. 2006-10-06 10:39 gb * [r5304] : Real, live EXTERNAL-FORMAT structures. 2006-10-06 10:38 gb * [r5303] : Bump. 2006-10-06 10:38 gb * [r5302] : %COPY-STRING-TO-U8: use #\? as a replacement character. 2006-10-05 21:20 gb * [r5301] : *** empty log message *** 2006-10-05 20:46 gb * [r5300] : More register name aliases. FMH. 2006-10-05 20:46 gb * [r5299] : get_mxcsr, set_mxcsr. 2006-10-05 20:46 gb * [r5298] : Save the mxcsr, load it with a C-like value (all exceptions masked) before doing a callback. (Not doing this meant that tcr.foreign_mxcsr was often set to a "lispy" value, where most exceptions were unmasked, and this caused FP exceptions when calling out to foreign code with the "lispy" tcr.foreign_mxcsr in effect.) Mach exception stuff arranges to return via sigreturn. Didn't fix the spurious SIGTRAPs, but might be the right idea (and might save a context switch or two.) Mach exception stuff initializes rflags to exception rflags when arranging to resume thread at handler. Leaving them uninitialized was causing spurious SIGTRAPs, since the trace bit(s) may have been set. 2006-10-05 12:20 gb * [r5297] : *** empty log message *** 2006-10-05 12:09 gb * [r5296] : WITH-ENCODED-CSTR(s). (Hey, it's a start.) 2006-10-05 12:09 gb * [r5295] : CHAR-NAME: try to use at least 4 hex digits for #\U+xxxx syntax. 2006-10-05 12:08 gb * [r5294] : UCS-2, some other changes. 2006-10-05 12:07 gb * [r5293] : Try to get line-termination/external-format stuff working. 2006-10-05 11:35 gb * [r5292] : Assume that character data is always encoded as one or more 8-bit octets. 2006-10-05 11:20 gb * [r5291] : *** empty log message *** 2006-10-05 10:59 gb * [r5290] : Set class-prototypes for basic-streams just before we make the first one. 2006-10-02 14:20 gb * [r5289] : *** empty log message *** 2006-10-02 14:13 gb * [r5288] : Darwin uses _sigtramp glue to call handlers, so we need to make handlers do a sigreturn after switching off the signal stack, etc. 2006-10-02 14:11 gb * [r5287] : Don't use SIGRETURN to return from signal_handler() on Darwin (since we depend on pseudo_sigreturn handling for cleanup of tcr pending exception and since we can't easily restore the signal mask.) (Up until this point, SIGRETURN has been a nop on Darwin.) 2006-10-02 11:20 gb * [r5286] : *** empty log message *** 2006-10-02 10:59 gb * [r5285] : Compile on ppc,ppc64 osx 10.4 and earlier. 2006-10-02 10:47 gb * [r5284] : Compile on post OSX 10.4 systems, for ppc/ppc64. (Broken on systems <= 10.4). 2006-10-02 09:20 gb * [r5283] : *** empty log message *** 2006-10-02 08:46 gb * [r5282] : Try to make x86-64 version compile under 10.4 again; see if this breaks post-10.4 compilation. 2006-10-02 08:33 gb * [r5281] : Change things so that they compile for x86-64 on systems later than 10.4. Almost certainly break 10.4 compilation in the process. (Thanks, Apple! Any time I can cause you to waste hours on crap like this, just let me know!) 2006-09-28 03:20 gb * [r5280] : *** empty log message *** 2006-09-28 02:33 gb * [r5279] : DOCUMENTATION works on character-encodings. 2006-09-28 02:33 gb * [r5278] : Character-encodings have doc strings. 2006-09-27 12:20 gb * [r5277] : *** empty log message *** 2006-09-27 11:46 gb * [r5276] : More FILE-STREAM changes. Give BASIC-STREAM classes a class prototype. 2006-09-27 11:45 gb * [r5275] : Some changes to FD-STREAM initialization, to make it easier to use MAKE-FD-STREAM to open (seekable) FILE-STREAMs. 2006-09-27 11:41 gb * [r5274] : Decode and handle UUO_ERROR_DEBUG_TRAP_WITH_STRING. 2006-09-27 11:39 gb * [r5273] : Still get very occasional SIGFPE on return from (e.g.) #_asin on x86-64 Darwin. (#_asin often does a division-by-zero, but that should be masked when in foreign code.) Don't reload the lisp mxcsr until other registers have been popped, to make debugging easier in case it's the act of reloading the mxcsr that causes the spurious trap.) 2006-09-27 11:36 gb * [r5272] : DEBUG-TRAP-WITH-STRING. 2006-09-27 11:35 gb * [r5271] : Not sure when it happened, but %FASLOAD was parenthesized in such a way as to not return the error value on error. 2006-09-27 11:34 gb * [r5270] : On x8664, use a new UUO for BUG. (Not passing context into the kernel makes it too difficult to debug/get backtrace, etc.) 2006-09-27 11:33 gb * [r5269] : UUO-ERROR-DEBUG-TRAP-WITH-STRING. 2006-09-27 11:20 gb * [r5268] : *** empty log message *** 2006-09-27 11:05 gb * [r5267] : CLASS-PROTOTYPE method for CLASS class (mostly for BUILT-IN-CLASSes). 2006-09-27 03:20 gb * [r5266] : *** empty log message *** 2006-09-27 02:49 gb * [r5265] : SELECT-STREAM-CLASS: should probably take a BASIC-P argument. 2006-09-27 02:47 gb * [r5264] : More changes. Imagine that. 2006-09-27 02:46 gb * [r5263] : (Redundant) FIXNUM decls in LDB compiler macro. 2006-09-27 02:43 gb * [r5262] : Memory/vector encode/decode functions take/return strings, not chars. Some bom stuff. 2006-09-25 02:20 gb * [r5261] : *** empty log message *** 2006-09-25 02:14 gb * [r5260] : Check for assumptions about character size. Remove some %i-isms (fixnum arithmetic), make assertions about type of char-code. 2006-09-25 02:12 gb * [r5259] : Remove unused (probably ancient) CHKBOUNDS function. Simpler STRING-START-END. 2006-09-25 02:11 gb * [r5258] : Some stuff to infer line-termination. Probably belongs elsewhere; some encodings may not make it at all possible. 2006-09-25 02:07 gb * [r5257] : Load-character-constant: accept :u32const on PPC32, PPC64, too. 2006-09-25 02:00 gb * [r5256] : A little more u16 stuff. Remove line-termination slot from ioblock and corresponding arg from functions which initialize it. 2006-09-25 01:59 gb * [r5255] : Input-filter, output-filter in file-ioblock. 2006-09-25 01:58 gb * [r5254] : Buffer translation functions don't belong here; start thinking about BOM. 2006-09-25 01:57 gb * [r5253] : LOAD-CHARACTER-CONSTANT: it's been a :u32const for the last few weeks ... 2006-09-24 11:20 gb * [r5252] : *** empty log message *** 2006-09-24 10:29 gb * [r5251] : CHAR-NAME looks in the *CHARACTER-NAMES* array if the code is within its bounds. 2006-09-24 10:27 gb * [r5250] : Lots more character names. Maintain mappings in a vector and a hash table, rather than in an alist. 2006-09-24 10:27 gb * [r5249] : Doh! character constant fasl encoding ($fasl-char). 2006-09-23 12:20 gb * [r5248] : *** empty log message *** 2006-09-23 11:31 gb * [r5247] : Some other iso8859-n encodings. (Not sure which are the most important.) 2006-09-23 11:29 gb * [r5246] : Start to separate line-termination/external-format stuff. 2006-09-23 11:28 gb * [r5245] : Lots-o-changes. 2006-09-23 11:28 gb * [r5244] : Make initial interactive streams use UTF-8 (mostly for debugging utf-8 ...) 2006-09-21 10:20 gb * [r5243] : *** empty log message *** 2006-09-21 10:07 gb * [r5242] : don't clobber callback index when doing the DARWIN_GS_HACK stuff in _SPcallback (syscall sometimes clobbered r11, sometimes didn't.) 2006-09-21 10:05 gb * [r5241] : Require mach/port.h. 2006-09-21 10:05 gb * [r5240] : Save and restore errno in callback_for_interrupt. 2006-09-21 10:04 gb * [r5239] : Guess again about preprocessor stuff for x86-64 ucontext/mcontext on Darwin. 2006-09-21 10:02 gb * [r5238] : Use underscore prefix on all Darwin flavors. Add a type decl in %nanosleep. 2006-09-21 10:01 gb * [r5237] : sign-extend/zero-extend vinsns. 2006-09-21 10:01 gb * [r5236] : sign/zero-extend ffcall, syscall return values. (Probably only matters in a few cases, like the Darwin x86-64 version of #_nanosleep, which returns a non-sign-extended "int" value if the sleep is interruted.) 2006-09-20 13:20 gb * [r5235] : *** empty log message *** 2006-09-20 12:34 gb * [r5234] : Bump. 2006-09-20 11:20 gb * [r5233] : *** empty log message *** 2006-09-20 10:25 gb * [r5232] : slight improvements in signed-byte boxing, typechecking. 2006-09-20 10:22 gb * [r5231] : WRITE-BYTE for BASIC-STREAMs. 2006-09-20 10:21 gb * [r5230] : %SWAP-U16, %SWAP-U32. 2006-09-20 10:21 gb * [r5229] : %EVAL-REDEFs for new require-xx functions. 2006-09-20 10:20 gb * [r5228] : UTF-16, in native and byte-swapped versions. Need a "logical" UTF-16, which deals with byte-order marks and a means of installing a concrete encoding. Needs testing. 2006-09-20 10:20 gb * [r5227] : *** empty log message *** 2006-09-20 10:15 gb * [r5226] : Numerous changes. 2006-09-20 10:15 gb * [r5225] : %bswap16: fix type declaration. 2006-09-20 10:14 gb * [r5224] : vinsns for signed/unsigned byte 8,16,32,64. Tighten up some signed-byte boxing. 2006-09-20 10:12 gb * [r5223] : New require ops for signed/unsigned byte 8,16,32,64 2006-09-19 08:21 gb * [r5222] : *** empty log message *** 2006-09-19 07:38 gb * [r5221] : vinsns for REQUIRE-[US][8,16,32,64] 2006-09-19 07:37 gb * [r5220] : Need THREE-DIGIT-BIGNUM-HEADER to recognize some cases of (UNSIGNED-BYTE 64). 2006-09-19 07:36 gb * [r5219] : clean up an error message. 2006-09-19 07:36 gb * [r5218] : handle require-[us][8,16,32,64] operators. 2006-09-19 07:34 gb * [r5217] : REQUIRE-TYPE: transform into require-[us][8,16,32,64] when possible. 2006-09-19 07:34 gb * [r5216] : inline require-[us][8,16,32,64]. 2006-09-19 07:33 gb * [r5215] : New operators for require-[us][8,16,32,64]. 2006-09-19 07:20 gb * [r5214] : *** empty log message *** 2006-09-19 07:15 gb * [r5213] : x8664 vinsns for typechecking signed/unsigned bytes of sizes 8,16,32,64. 2006-09-19 06:36 gb * [r5212] : More changes; hard to bootstrap. 2006-09-19 04:20 gb * [r5211] : *** empty log message *** 2006-09-19 03:42 gb * [r5210] : Macros for locking the input/output locks of an ioblock, assuming that it exists. 2006-09-17 07:20 gb * [r5209] : *** empty log message *** 2006-09-17 06:45 gb * [r5208] : more changes 2006-09-13 13:20 gb * [r5207] : *** empty log message *** 2006-09-13 12:52 gb * [r5206] : Direct BASIC-STREAM stuff for WRITE-CHAR, WRITE-STRING, TERPRI. 2006-09-13 12:20 gb * [r5205] : *** empty log message *** 2006-09-13 12:08 gb * [r5204] : READ-LINE changes for BASIC-STREAMs. 2006-09-13 12:07 gb * [r5203] : GET-CHARACTER-ENCODING errors if the named encoding isn't defined. Terse PRINT-OBJECT method for CHARACTER-ENCODING. 2006-09-13 12:06 gb * [r5202] : More functions in the ioblock, some support for 8-bit encodings (like utf-8). Not finished, but getting hard to bootstrap; need to check in and build on all platforms before it gets even harder. 2006-09-12 12:20 gb * [r5201] : *** empty log message *** 2006-09-12 11:34 gb * [r5200] : Use %CSTR-POINTER (not %copy-ivector-to-pointer) in BUG. 2006-09-12 11:31 gb * [r5199] : Enter the debugger if altstack_signal_handler is entered in foreign code. (Maybe want to signal a lisp error in some cases, but this is currently fatal and we might as well die earlier than with a stack overflow.) 2006-09-12 11:29 gb * [r5198] : os_get_stack_bounds(): get stack direction right on FreeBSD. (Much better than randomly scribbling on memory in the GC, as the FreeBSD port has been doing for months now ...) 2006-09-12 04:20 gb * [r5197] : *** empty log message *** 2006-09-12 03:31 gb * [r5196] : x862-explicit-non-fixnum-type-p looks for intersection. 2006-09-12 03:21 gb * [r5195] : Partly inline 2-arg logior, logand. 2006-09-12 02:20 gb * [r5194] : *** empty log message *** 2006-09-12 02:10 gb * [r5193] : Character encodings contain a LITERAL-CHAR-CODE-LIMIT field. 2006-09-12 02:09 gb * [r5192] : extend IOBLOCK a bit. Start to provide functionality for encoded streams. 2006-09-12 02:08 gb * [r5191] : Partialy inline 2-arg logior/logand. 2006-09-09 11:20 gb * [r5190] : *** empty log message *** 2006-09-09 10:39 gb * [r5189] : REQUIRE-CHAR-CODE: compare against boxed char-code-limit. 2006-09-09 10:21 gb * [r5188] : NAME-CHAR: insist on something more reasonable for ^c syntax; don't do MOD for octal. 2006-09-09 10:21 gb * [r5187] : require-char-code checks agains boxed Unicode limit 2006-09-09 10:21 gb * [r5186] : *** empty log message *** 2006-09-09 09:56 gb * [r5185] : purify_range(), impurify_range(): handle x8664 functions, hash tables. (PPC versions should handle hash table key movement.) 2006-09-09 05:20 gb * [r5184] : *** empty log message *** 2006-09-09 04:47 gb * [r5183] : CHAR-NAME: use U+xxxx for chars with codes >= #x100. 2006-09-09 04:38 gb * [r5182] : FIXNUM->CHAR: return NIL for surrogate pairs. 2006-09-09 04:33 gb * [r5181] : Don't be overly unsafe in %CHAR-CODE. 2006-09-09 04:31 gb * [r5180] : Fix CFP-LFUN (fake stack frame case, sanity-check fn/lr). 2006-09-09 01:20 gb * [r5179] : *** empty log message *** 2006-09-09 00:55 gb * [r5178] : Don't be so reckless in %CODE-CHAR (don't necessarily want arbitrary expressions to be compiled unsafely; do want to avoid a little type-checking sometimes.) 2006-09-09 00:52 gb * [r5177] : FIXNUM->CHAR returns NIL for values in the surrogate-pairs range. (There are other things that it could plausibly return NIL for.) 2006-09-09 00:51 gb * [r5176] : Start to get purify working on x86-64; for now, just purify strings. 2006-09-09 00:48 gb * [r5175] : %CODE-CHAR asserts that code is (UNSIGNED-BYTE 8), which is still useful (and exactly what it used to do.) 2006-09-09 00:47 gb * [r5174] : Add l1-unicode. 2006-09-09 00:44 gb * [r5173] : Load l1-unicode. 2006-09-09 00:44 gb * [r5172] : New file. 2006-09-07 10:20 gb * [r5171] : *** empty log message *** 2006-09-07 09:50 gb * [r5170] : Larger PURESPACE_RESERVE for x8664. 2006-09-07 09:44 gb * [r5169] : Bump abi version. Make purespace-reserve an attribute of the backend (may need more pure space on x86-64.) 2006-09-07 09:43 gb * [r5168] : Bump abi version. 2006-09-07 09:43 gb * [r5167] : Bump. 2006-09-07 09:20 gb * [r5166] : *** empty log message *** 2006-09-07 08:58 gb * [r5165] : FreeBSD (at least with libthr) seems to mask synchronous signals. Unmask them in signal handlers. 2006-09-07 08:57 gb * [r5164] : suspend_resume_handler() : ignore signal mask. 2006-09-07 08:56 gb * [r5163] : Lose the CHAR_SIZE_32 bootstrapping conditional (char are 32 bits wide on all platforms now.) 2006-09-07 07:20 gb * [r5162] : *** empty log message *** 2006-09-07 06:40 gb * [r5161] : SIGBUS might be SIGNUM_FOR_PROCESS_INTERRUPT. 2006-09-07 06:20 gb * [r5160] : *** empty log message *** 2006-09-07 06:12 gb * [r5159] : _SPbuiltin_aref1: use box_fixnum_no_flags. 2006-09-07 06:12 gb * [r5158] : Define box_fixnum_no_flags (uses lea vice imul). 2006-09-07 03:20 gb * [r5157] : *** empty log message *** 2006-09-07 03:18 gb * [r5156] : Need MIN/MAX-8-BIT-IVECTOR-SUBTAG. 2006-09-07 01:20 gb * [r5155] : *** empty log message *** 2006-09-07 00:37 gb * [r5154] : misc_ref/misc_set: handle 32-bit strings. 2006-09-07 00:36 gb * [r5153] : Zero-extend early when extracting typecodes. 2006-09-07 00:36 gb * [r5152] : GET-OUTPUT-STREAM-STRING: use ENSURE-SIMPLE-STRING rather than horrible COERCE nonsense. 2006-09-07 00:35 gb * [r5151] : x8664 *immheader-1-array-element-types* contains entry for 32-bit BASE-CHAR. 2006-09-07 00:35 gb * [r5150] : zero-extend early when extracting typecodes. 2006-09-07 00:34 gb * [r5149] : 32-bit-chars. 2006-09-07 00:33 gb * [r5148] : Defer handling of the :rcontext pseudo-register name until the target backend is in place. 2006-09-06 23:20 gb * [r5147] : *** empty log message *** 2006-09-06 22:39 gb * [r5146] : 32-bit chars stuff for x8664. 2006-09-06 22:39 gb * [r5145] : kids: bourbon and editing COPY-READTABLE don't mix! 2006-09-06 22:38 gb * [r5144] : :verbose argument to rebuild-ccl. 2006-09-06 22:38 gb * [r5143] : process input/output: don't assume 8-bit strings. 2006-09-06 21:20 gb * [r5142] : *** empty log message *** 2006-09-06 20:42 gb * [r5141] : PPC64 callbacks are always #+poweropen-target; don't conditionalize the fp-arg-ptr argument (since that creates confustion when cross-compiling from a non-poweropen host.) 2006-09-06 06:20 gb * [r5140] : *** empty log message *** 2006-09-06 06:08 gb * [r5139] : remove 8-bit char stuff. 2006-09-06 06:08 gb * [r5138] : Forgot to set char-code-limit here. 2006-09-06 05:20 gb * [r5137] : *** empty log message *** 2006-09-06 04:49 gb * [r5136] : In DO-GC-AREAS, don't use comma before call to AREA-SUCC. (Cross-compilation/ word-size issue.) 2006-09-06 04:42 gb * [r5135] : If ENSURE-SIMPLE-STRING has to make a copy of a non-simple string, understand char width issues. 2006-09-06 04:41 gb * [r5134] : Some fixnum decls in %FIND-PKG. 2006-09-06 04:40 gb * [r5133] : ppc64 ARRAY-ELEMENT-SUBTYPES includes new string type. 2006-09-06 04:39 gb * [r5132] : cr3, nor c33, in ppc64 %init-misc. 2006-09-06 04:38 gb * [r5131] : NX-FORM-TYPEP uses NX-TARGET-TYPE. 2006-09-05 15:20 gb * [r5130] : *** empty log message *** 2006-09-05 15:13 gb * [r5129] : darwinppc32 backend uses 32-bit chars, too. 2006-09-05 15:01 gb * [r5128] : ppc32 defaults to 32-bit chars. 2006-09-05 11:20 gb * [r5127] : *** empty log message *** 2006-09-05 11:18 gb * [r5126] : Deal with 32-bit lisp strings. 2006-09-05 11:18 gb * [r5125] : Start to depend on calling print_lisp_object, since x/s doesn't deal with 32-bit lisp strings. 2006-09-05 11:17 gb * [r5124] : conditionalize purify_displaced_object() for char size. 2006-09-05 11:16 gb * [r5123] : find_symbol(): don't assume 8-bir lisp strings. 2006-09-05 11:15 gb * [r5122] : Try using "u8_t" and "u32_t" for lisp_char_code; this may be more portable. 2006-09-05 11:14 gb * [r5121] : SIMPLE-1D-ARRAY-SUBSEQ: handle new strings. 2006-09-05 11:13 gb * [r5120] : CALL-WITH-STRING-VECTOR: don't assume 8-bit char/strings. 2006-09-05 11:13 gb * [r5119] : Use heap-ivectors of element-type (UNSIGNED-BYTE 8) for character streams. (Should really depend on the coding system in use.) 2006-09-05 11:09 gb * [r5118] : Various archaic string functions: don't assume 8-bit strings. 2006-09-05 11:08 gb * [r5117] : Handle 32-bit strings in %UVECTOR-REPLACE. 2006-09-05 11:07 gb * [r5116] : No "x" in simple-base-string, when #-target-8-bit-chars. 2006-09-05 11:06 gb * [r5115] : REQUIRE-CHAR-CODE: comparison of OBJECT with TAG is not a CMPI. 2006-09-05 11:05 gb * [r5114] : PPC2-VREF: handle the 32-bit :SIMPLE-STRING case with constant index. 2006-09-05 00:20 gb * [r5113] : *** empty log message *** 2006-09-04 23:44 gb * [r5112] : Linux doesn't define uint8_t. Try "unsigned char", and cross fingers. 2006-09-04 23:38 gb * [r5111] : Define lisp_char_code type, use it when printing lisp string in the kernel debugger. 2006-09-04 23:28 gb * [r5110] : Bump. 2006-09-04 23:20 gb * [r5109] : *** empty log message *** 2006-09-04 23:18 gb * [r5108] : Sigh. Test for 8-bit-char case was conditionalized backwards in PPC64 %INIT-MISC. 2006-09-04 23:05 gb * [r5107] : Initial value of CHAR-CODE-LIMIT derived from the target arch (for bootstrapping.) 2006-09-04 23:04 gb * [r5106] : Use (UNSIGNED-BYTE 8) for array element type of rdtab.ttab. Don't treat (code-char #xca) as whitespace; do treat (code-char #xa0) as whitespace. (The former happens to be how the latter was encoded in MacRoman.) Handle U+ syntax in NAME-CHAR. 2006-09-04 22:20 gb * [r5105] : *** empty log message *** 2006-09-04 22:16 gb * [r5104] : LOAD-CHARACTER-CONSTANT: and the char code with #xff to get the low byte. 2006-09-04 22:15 gb * [r5103] : LOAD-CHARACTER-CONSTANT: and the char code with #xff or 255., not with #x255. (Cause lots of wackiness, mostly in readtable initialization.) 2006-09-04 22:12 gb * [r5102] : Use u32->char (which is defined) vice u8->char (which is no more) in %CHAR-CODE. 2006-09-04 12:20 gb * [r5101] : *** empty log message *** 2006-09-04 12:04 gb * [r5100] : Gak. Get guard/predicate syntax right in LOAD-CHARACTER-CONSTANT. 2006-09-04 12:03 gb * [r5099] : New constants in ppc32 typecode range assertion. 2006-09-04 10:20 gb * [r5098] : *** empty log message *** 2006-09-04 09:44 gb * [r5097] : ppc32/ppc64 %init-misc tweaks. 2006-09-04 09:43 gb * [r5096] : Typo in the (unused) 64-bit/no crf case of UNBOX-BASE-CHAR. 2006-09-04 09:20 gb * [r5095] : *** empty log message *** 2006-09-04 08:30 gb * [r5094] : Object-not- (mod #x110000) type error. 2006-09-04 08:29 gb * [r5093] : conditionalize %pname-hash for char size. 2006-09-04 08:26 gb * [r5092] : Conditionalize for CHAR_SIZE_32. 2006-09-04 08:25 gb * [r5091] : Conditionalize for #-target-8-bit-chars. 2006-09-04 08:23 gb * [r5090] : Deal - sometimes conditionally - with 32-bit characters. 2006-09-04 08:22 gb * [r5089] : ERROR-OBJECT-NOT-MOD-CHAR-CODE-LIMIT. 2006-09-03 22:20 gb * [r5088] : *** empty log message *** 2006-09-03 22:03 gb * [r5087] : conditionalize new string types. 2006-09-03 22:03 gb * [r5086] : %pname-hash: conditionalize for string element size. 2006-09-03 22:02 gb * [r5085] : Sketch out new string types (read-time conditionalized.) 2006-09-03 21:20 gb * [r5084] : *** empty log message *** 2006-09-03 21:02 gb * [r5083] : Put :target-8-bit-chars on *features*, for bootstrapping. 2006-09-03 20:21 gb * [r5082] : *** empty log message *** 2006-09-03 20:20 gb * [r5081] : More missing BASIC-STREAM methods. 2006-09-03 19:23 gb * [r5080] : :target-8-bit-chars on target features. 2006-09-03 19:21 gb * [r5079] : Some comments/indentation. 2006-09-03 19:20 gb * [r5078] : *** empty log message *** 2006-09-03 19:01 gb * [r5077] : Make CHAR-CODE-LIMIT an attribute of the target arch, for bootstrapping. 2006-09-02 18:20 gb * [r5076] : *** empty log message *** 2006-09-02 17:24 gb * [r5075] : Fix typo. 2006-09-02 17:20 gb * [r5074] : *** empty log message *** 2006-09-02 17:15 gb * [r5073] : hex/decimal integer printers: 3 is also a hex digit. 2006-09-02 17:11 gb * [r5072] : Install a "bogus signal handler" on Darwin, to deal with apparent race conditions/Mach kernel lossage. Don't do the whole msg_server here; just do the exc_server thing. Fix a couple of Mach exception-handling bugs; at least one more may remain. 2006-09-02 17:07 gb * [r5071] : Use a hlt instruction to trap in pseudo_sigreturn. Still get spurious SIGTRAPs. 2006-09-02 17:06 gb * [r5070] : Define some things in terms of USE_MACH_SEMAPHORES or USE_POSIX_SEMAPHORES. sem_wait_forever() does repeated timed waits, since I thought that Darwin was having trouble interrpting things that waited indefinitely. (That might just be a GDB bug. Who can tell ?) 2006-09-02 17:03 gb * [r5069] : Define some things in terms of USE_MACH_SEMAPHORES or USE_POSIX_SEMAPHORES. Note that as of Q3 2006, Darwin still doesn't implement enough POSIX semaphore functionality to make it possible to avoid the use of Mach semaphores. 2006-09-02 17:01 gb * [r5068] : Try to avoid things like %COPY-IVECTOR-TO-IVECTOR when strings are involved. 2006-09-02 17:01 gb * [r5067] : Things that're %HEAP-IVECTOR-P aren't bogus objects. 2006-09-02 17:00 gb * [r5066] : Try to avoid things like %COPY-IVECTOR-TO-IVECTOR when strings are involved. 2006-09-02 16:59 gb * [r5065] : AUTO-FLUSH-INTERACTIVE-STREAMS: use FORCE-OUTPUT. 2006-09-02 16:57 gb * [r5064] : %COPY-U8-TO-STRING, %COPY-STRING-TO-U8: new, to avoid %COPY-IVECTOR-TO-IVECTOR when strings are involved. %STR-FROM-PTR: take dest string as optional arg, don't assume 8-bit chars. %GET-CSTRING: change type decl, use %GET-UNSIGNED-BYTE. 2006-09-02 16:55 gb * [r5063] : %CSTR-POINTER and %CSTR-SEGMENT-POINTER: don't assume 8-bit chars. 2006-08-29 00:20 gb * [r5062] : *** empty log message *** 2006-08-28 23:35 gb * [r5061] : Fix tag check in 32-bit misc_set_single_float_vector. 2006-08-28 23:20 gb * [r5060] : *** empty log message *** 2006-08-28 23:01 gb * [r5059] : Bump max fasl version. 2006-08-28 22:48 gb * [r5058] : Use new string ops. Bump current fasl version. 2006-08-28 22:33 gb * [r5057] : Bump (max) fasl version. 2006-08-28 22:30 gb * [r5056] : xloader was missing a few new ops. 2006-08-28 12:20 gb * [r5055] : *** empty log message *** 2006-08-28 11:34 gb * [r5054] : New string stuff here, too. 2006-08-28 11:28 gb * [r5053] : New string stuff. 2006-08-28 11:20 gb * [r5052] : *** empty log message *** 2006-08-28 10:57 gb * [r5051] : More missing BASIC-STREAM methods (STREAM-LINE-COLUMN, STREAM-LISTEN, STREAM-EOFP.) 2006-08-28 10:57 gb * [r5050] : Use READ-CHAR instead of STREAM-READ-CHAR, since it may be faster in some cases. 2006-08-28 10:56 gb * [r5049] : Unscramble auto-flush-streams stuff (don't assume ioblock-based stream.) 2006-08-28 10:55 gb * [r5048] : Initial FD-STREAMs are BASIC-STREAMs. 2006-08-28 10:55 gb * [r5047] : Change some REQUIRE forms; cosmetic. (Maybe not even cosmetic.) 2006-08-28 10:54 gb * [r5046] : %unbox-u8, u8-vref stuff. 2006-08-28 10:53 gb * [r5045] : New fasl ops for new string encodings. 2006-08-28 10:52 gb * [r5044] : %CHAR-CODE: for how long has limit not been 2^16 ? For how long will limit be 2^8 ? 2006-08-28 10:51 gb * [r5043] : More fasl ops. 2006-08-28 10:51 gb * [r5042] : %UNBOX-U8, u8-vref stuff. 2006-08-28 10:50 gb * [r5041] : Add CHAR-CODE-LIMIT slot to TARGET-ARCH; may not be the best way to do this. 2006-08-26 20:20 gb * [r5040] : *** empty log message *** 2006-08-26 19:58 gb * [r5039] : Updated. 2006-08-26 19:50 gb * [r5038] : Updated. 2006-08-26 19:50 gb * [r5037] : Added. 2006-08-26 09:20 gb * [r5036] : *** empty log message *** 2006-08-26 09:17 gb * [r5035] : Bump. New snapshots soon, I hope. 2006-08-26 09:12 gb * [r5034] : Pass -g in ASFLAGS; same problems as always, but avoids problems with Darwin x86-64 ld64 segfaulting for unknown reasons. 2006-08-26 09:11 gb * [r5033] : Need to do an explicit sigreturn on Darwin (because of the sigaltstack hacks; Darwin uses sigtramp glue that we can't just return to after switching off of the alt stack.) 2006-08-26 09:08 gb * [r5032] : Check fpu overflow bit (not underflow) when deciding whether or not to signal overflow. 2006-08-26 09:08 gb * [r5031] : Need an indirection to access mcontext from ucontext on Darwin; invent xp-gp-regs to get at mcontext64.ss. 2006-08-26 09:06 gb * [r5030] : Add a declaration, to make this easier to step through. 2006-08-26 09:05 gb * [r5029] : SINGLE-FLOAT-HALF-PI: use f0 not e0, just in case. 2006-08-26 03:20 gb * [r5028] : *** empty log message *** 2006-08-26 02:35 gb * [r5027] : Hyphenate ADD-AUTO-FLUSH-STREAM and REMOVE-AUTO-FLUSH-STREAM. Export them, too. 2006-08-26 02:22 gb * [r5026] : READ-LINE on BASIC-INPUT-STREAM w/o STREAM-READ-LINE. 2006-08-26 02:22 gb * [r5025] : HANDLER-CASE: the :NO-ERROR clause can appear at most once but doesn't have to be the last clause. 2006-08-26 02:21 gb * [r5024] : BASIC-OUTPUT-STREAM isn't an INPUT-STREAM. Some STREAM-READ-LINE support for BASIC-CHARACTER-INPUT-STREAM. 2006-08-26 02:20 gb * [r5023] : *** empty log message *** 2006-08-26 02:19 gb * [r5022] : PREPARE-TO-QUIT: other threads should be dead by the time that streams on *OPEN-FILE-STREAMS* are closed, so unlock those streams with extreme prejudice before closing them. 2006-08-26 02:18 gb * [r5021] : FULL-PATHNAME: remove some weird old code that assumed that interfered with handling of relative logical pathnames. 2006-08-26 02:17 gb * [r5020] : Rename EVENT-POLL to AUTO-FLUSH-INTERACTIVE-STREAMS. Make AUTO-FLUSH-INTERACTIVE-STREAMS do conditional locking on each stream (doing nothing if it's not :SHARING :LOCK; unconditional locking on the *AUTO-FLUSH-STREAMS-LOCK*. ADD-AUTOFLUSH-STREAM ensures that the stream is :SHARING :LOCK. REMOVE-AUTOFLUSH-STREAM. 2006-08-26 02:13 gb * [r5019] : MAKE-MCL-LISTENER-STREAM uses ADD-AUTOFLUSH-STREAM, REMOVE-AUTOFLUSH-STREAM. 2006-08-24 05:20 gb * [r5018] : *** empty log message *** 2006-08-24 04:59 gb * [r5017] : Darwin needs to decode the mxcsr on SIGFPE. DARWIN_GS_HACK changes in altstack handlers for interrupt/suspend. 2006-08-24 04:58 gb * [r5016] : x86-64 Darwin's math library is pretty casual about causing spurious exceptions, so ignore them. 2006-08-24 04:57 gb * [r5015] : New file; like the linuxx8664 version, with different signals to be ignored by gdb. 2006-08-24 04:56 gb * [r5014] : Using ud2b to force an exception in pseudo_sigreturn() seems to cause/be related to spurious SIGTRAPs. Use int $0 instead. 2006-08-24 04:55 gb * [r5013] : First attempt at Darwinx8664 support; may need more work. 2006-08-22 08:22 gb * [r5012] : *** empty log message *** 2006-08-22 08:08 gb * [r5011] : Try not to use sprintf() (at least not for printing integers), since it seems to need access to pthread data (and calling print_lisp_object() from GDB might fail if %gs is pointing at the lisp TCR.) 2006-08-22 08:06 gb * [r5010] : Unbox %arg_z (boxed entrypoint) earlier in _SPffcall, before entering foreign context. (Need to check PPC version(s) as well.) Use DARWIN_GS_HACK to swap %gs memory around when transitioning between foreign code and lisp. 2006-08-22 08:04 gb * [r5009] : Fix a typo. 2006-08-22 08:04 gb * [r5008] : Define a few more things ... 2006-08-22 08:03 gb * [r5007] : Lots of changes, especially to Mach exception handling code. Do the DARWIN_GS_HACK things to overload %gs. 2006-08-22 08:02 gb * [r5006] : atomic_and(), pseudo_sigreturn() for DarwinX8664. 2006-08-22 08:01 gb * [r5005] : Some PPC64 conditionalizations should really have tested WORD_SIZE. DARWIN_GS_HACK stuff around suspend/resume signal handlers. (May not be necessary.) Ensure that tcr.linear gets set on x86, even if not HAVE_TLS. 2006-08-22 07:59 gb * [r5004] : Don't really flush cache lines on x8664. Some PPC64 conditionalizations should really have tested WORD_SIZE. 2006-08-22 07:58 gb * [r5003] : Define SYSCALL_SETS_CARRY_ON_ERROR for FreeBSDX664, DarwinX8664. 2006-08-22 07:57 gb * [r5002] : Much smaller pagezero, much lower seg1addr. Something in the (m4-generated) symbolic info in some .s files is screwy and causes ld64 to segfault; it's currently necesssary to use "strip -S" on the assembler-generated .o files. 2006-08-22 07:55 gb * [r5001] : Untested changes for Darwinx8664. 2006-08-22 07:54 gb * [r5000] : Fix typo in darwinx8664 xload-backend-info (compiler target is :darwinx8664, not :linuxx8664). 2006-08-22 07:53 gb * [r4999] : Dump concatenated functions from xfunctions when cross-compiling to x86 from x86. 2006-08-22 07:52 gb * [r4998] : MACHINE-VERSION: conditionalize on :DARWIN-TARGET (not :DARWINPPC-TARGET). 2006-08-22 07:50 gb * [r4997] : :DARWIN-TARGET (not :DARWINPPC-TARGET). 2006-08-22 07:50 gb * [r4996] : :DARWINX8664 stuff. 2006-08-22 07:49 gb * [r4995] : Bump. 2006-08-22 07:49 gb * [r4994] : Conditionalize on DARWIN-TARGET, not DARWINPPC-TARGET. 2006-08-22 07:49 gb * [r4993] : FILE-STREAM-EXTERNAL-FORMAT stuff for BASIC-STREAMs. 2006-08-22 07:47 gb * [r4992] : DarwinX8664 *features*. 2006-08-22 07:47 gb * [r4991] : Some DARWINPPC-TARGET conditionals really want to be DARWIN-TARGET. 2006-08-22 07:46 gb * [r4990] : YIELD-SYSCALL for DarwinX8664. 2006-08-22 07:45 gb * [r4989] : Use CROSS-CREATE-X86-FUNCTION if not compiling natively. Fix X862-FIXUP-FWD-REFS if not native function on x86. 2006-08-22 07:44 gb * [r4988] : Ensure that CROSS-CREATE-X86-FUNCTION is defined, even on x86. Use it if not compiling natively on x86. 2006-08-21 06:20 gb * [r4987] : *** empty log message *** 2006-08-21 05:59 gb * [r4986] : Define DARWIN_GS_HACK. 2006-08-21 05:58 gb * [r4985] : Still needs lots of work. 2006-08-21 05:58 gb * [r4984] : Include Mach stuff on Darwin; prototypes for DARWIN_GS_HACK %gs segment switching. 2006-08-21 05:57 gb * [r4983] : Manipulate the %gs register when a lisp thread starts, exits and DARWIN_GS_HACK is defined. This isn't so bad, but doing it around every ff-call might be. 2006-08-21 05:56 gb * [r4982] : DARWIN_GS_HACK macros. 2006-08-21 05:53 gb * [r4981] : Note the need for DARWIN_GS_HACK. 2006-08-21 05:51 gb * [r4980] : DARWIN_GS_HACK hacks. 2006-08-21 05:51 gb * [r4979] : Hope that the need for DARWIN_GS_HACK is also temporary. 2006-08-21 05:50 gb * [r4978] : Hope that the workaround for missing mcontext64 in x8664 Darwin is temporary. 2006-08-21 03:20 gb * [r4977] : *** empty log message *** 2006-08-21 02:35 gb * [r4976] : typo 2006-08-21 02:35 gb * [r4975] : spell name right 2006-08-20 23:20 gb * [r4974] : *** empty log message *** 2006-08-20 23:10 gb * [r4973] : :darwinx8664 stuff. 2006-08-20 23:10 gb * [r4972] : :struct-by-value: don't read-time conditionalize (do other things still use read-time conditionals ?) 2006-08-20 23:09 gb * [r4971] : Darwinx8664: overload %gs, setup syscall mask. 2006-08-20 23:08 gb * [r4970] : Handle syscall, ff-call for :darwinx8664 as well. 2006-08-20 23:06 gb * [r4969] : New file; or's in the bit for Unix syscalls on Darwinx8664. (Screw: should (carefully) review word-size issues.) 2006-08-20 23:03 gb * [r4968] : Mask for PPC. 2006-08-20 23:02 gb * [r4967] : Syscall conditionalization. 2006-08-20 23:01 gb * [r4966] : Syscall conditionalization. SELECTION-INPUT-STREAM created with :sharing :lock. 2006-08-20 23:00 gb * [r4965] : Fencepost in POINTER-IN-CFSTRING-SECTION-P. 2006-08-20 22:20 gb * [r4964] : *** empty log message *** 2006-08-20 22:17 gb * [r4963] : rename 2006-08-18 16:20 gb * [r4962] : *** empty log message *** 2006-08-18 15:50 gb * [r4961] : Enable FP underflow detection on x8664. Should do this on PPC as well, and should probably offer some backward-compatible behavior for reader. 2006-08-18 15:47 gb * [r4960] : Dress up the Darwin x86-64 backend a little. Explicily define lisp-context-register slot in each backend (%gs for those imolementations that follow the standard ABI, %fs for Darwin.) 2006-08-18 15:44 gb * [r4959] : Use :rcontext when necessary. 2006-08-18 15:43 gb * [r4958] : Recognize :rcontext as a reference to the backend's LISP-CONTEXT-REGISTER slot. 2006-08-18 15:42 gb * [r4957] : Backend now has a lisp-context-register slot. 2006-08-18 15:42 gb * [r4956] : No more (constant) rcontext. 2006-08-18 15:40 gb * [r4955] : Refer to rcontext segment register as :rcontext, since x86-64 Darwin doesn't laugh and play like the other children. 2006-08-18 15:29 gb * [r4954] : Copy some of the Mach exception code from the PPC version and tweak it a little; needs more work. 2006-08-18 15:27 gb * [r4953] : Provide some interim Darwin conditionalization. 2006-08-18 15:25 gb * [r4952] : Spell FREEBSD correctly. Generate an equate for HAVE_TLS. 2006-08-18 15:23 gb * [r4951] : Changes in purify_xp(). 2006-08-18 15:20 gb * [r4950] : *** empty log message *** 2006-08-18 15:19 gb * [r4949] : Darwin x86-64 assembler doesn't like absolute jumps, prefers slower ones. Move lock prefixes to their own lines, to satisfy you-know-who. If we don't have thread-local-storage (TLS), it's hard to avoid the get_tcr() call on each calback. 2006-08-18 15:13 gb * [r4948] : Tell m4 that we HAVE_TLS. 2006-08-17 17:20 gb * [r4947] : *** empty log message *** 2006-08-17 16:43 phil * [r4946] : Initial version 2006-08-17 16:40 phil * [r4945] : *** empty log message *** 2006-08-08 19:20 gb * [r4944] : *** empty log message *** 2006-08-08 19:07 gb * [r4943] : New file. 2006-08-08 19:07 gb * [r4942] : xp accessors for Darwinx8664. 2006-08-08 19:06 gb * [r4941] : REG_* constants for Darwin. 2006-08-08 19:05 gb * [r4940] : Darwin assembler doesn't like lock instructions on the same line as the locked instruction. (Maybe there's a better way to do this, than just pretending that lock is a standalone instruction, but that lock instruction does seem to be treated as a prefix.) 2006-08-08 19:03 gb * [r4939] : Treat X8664 like PPC64 (e.g., use dlfcn) on Darwin. 2006-08-08 19:02 gb * [r4938] : xcode 2.4 headers: doesn't define mcontext64, but lots of other things reference it. #define it here for now, but that'll likely break when the headers are fixed. 2006-08-08 18:58 gb * [r4937] : Define Iregnames for darwinx8664. 2006-08-08 18:58 gb * [r4936] : Define an IMAGE_BASE_ADDRESS for x8664/Darwin. 2006-07-29 15:20 gb * [r4935] : *** empty log message *** 2006-07-29 14:41 gb * [r4934] : STREAM-WRITE-BYTE, STREAM-WRITE-IVECTOR. :BASIC defaults to T in OPEN; BASIC-FILE-STREAM probably still needs READ/WRITE- SEQUENCE support. 2006-07-29 14:20 gb * [r4933] : *** empty log message *** 2006-07-29 13:39 gb * [r4932] : ALLOCATE-RESOURCE, FREE-RESOURCE: use lock. TODO: evaluate whether resources are worthwhile; used for string-output-stream allocation and DELETE-DUPLICATES. 2006-07-29 13:20 gb * [r4931] : *** empty log message *** 2006-07-29 12:27 gb * [r4930] : Resources aren't thread-private, need locks. 2006-07-29 12:26 gb * [r4929] : MAKE-FILE-STREAM: support for BASIC-STREAMs. 2006-07-29 12:26 gb * [r4928] : OPEN handles BASIC-STREAMs; not yet the default. 2006-07-29 12:25 gb * [r4927] : *PARSE-STRING-RESOURCE* probably hasn't been used in 10 or 12 years. 2006-07-29 10:21 gb * [r4926] : *** empty log message *** 2006-07-29 09:40 gb * [r4925] : Most BASIC-FILE-STREAM stuff. 2006-07-29 09:40 gb * [r4924] : BASIC-FILE-STREAM.EXTERNAL-FORMAT, at least for now. 2006-07-29 09:39 gb * [r4923] : BASIC-IO-STREAM; ALLOCATE-BASIC-STREAM makes one more slot in BASIC-FILE-STREAM. 2006-07-28 10:20 gb * [r4922] : *** empty log message *** 2006-07-28 10:17 gb * [r4921] : Bump. 2006-07-28 10:15 gb * [r4920] : A couple of more (soon-to-be-used) slots in IOBLOCK. 2006-07-28 07:20 gb * [r4919] : *** empty log message *** 2006-07-28 07:17 gb * [r4918] : Lots of changes; lots more to go. 2006-07-28 07:16 gb * [r4917] : Pass :CHARACTER-P T to MAKE-FD-STREAM. 2006-07-28 06:58 gb * [r4916] : Want :SHARING :LOCK on listener output. (SLIME probably faces the same issue; should probably do something in the auto-flush case. Of course, it's hard to know just what to do in that case. or at least what to do safely in that case.) 2006-07-28 06:49 gb * [r4915] : Define SELECT-STREAM-CLASS on FUNDAMENTAL-FILE-STREAM. 2006-07-28 06:49 gb * [r4914] : READ-BYTE for BASIC-STREAMs. 2006-07-28 06:48 gb * [r4913] : Need an explicit BASIC-STREAM-FLAG.OPEN-CHARACTER bit, due to bivalence. 2006-07-28 06:47 gb * [r4912] : BASIC-STREAM-IOBLOCK. 2006-07-25 10:20 gb * [r4911] : *** empty log message *** 2006-07-25 09:48 gb * [r4910] : Compiler-macro for BASIC-STREAM-P. 2006-07-24 11:20 gb * [r4909] : *** empty log message *** 2006-07-24 10:55 gb * [r4908] : READ-CHAR bypasses STREAM-READ-CHAR for BASIC-STREAMs. 2006-07-24 10:55 gb * [r4907] : Some BASIC-STREAM stuff starts to appear. 2006-07-24 10:51 gb * [r4906] : Add a plist to basic-/basic-file-stream. 2006-07-24 09:20 gb * [r4905] : *** empty log message *** 2006-07-24 08:43 gb * [r4904] : FILE-STREAM is now abstract (and some sublasses may get that way); FUNDAMENTAL-FILE-STREAM is an implementation of it (and BASIC-FILE-STREAM soon will be.) 2006-07-23 09:20 gb * [r4903] : *** empty log message *** 2006-07-23 09:05 gb * [r4902] : *class-table* entry for BASIC-STREAM. 2006-07-23 09:03 gb * [r4901] : STREAM-IOBLOCK now takes 2 required args and is inlined. Some bug fixes; STREAM class no longer has slots. IOBLOCK-READ-CHAR-FUNCTION used by READ-CHAR. %ioblock-read-byte variants; fix read-byte botch (should vector according to element-type.) 2006-07-23 08:56 gb * [r4900] : STREAM-IOBLOCK now takes 2 required args. 2006-07-23 08:54 gb * [r4899] : Bump. 2006-07-23 08:53 gb * [r4898] : Start to define what a BASIC-STREAM looks like. 2006-07-23 08:53 gb * [r4897] : Move stream-ioblock access/locking macros here. Likewise WITH-EAGAIN. 2006-07-23 07:20 gb * [r4896] : *** empty log message *** 2006-07-23 07:12 gb * [r4895] : Fix a few botches. :private arg to stream creation primitives replaced with :sharing, which can have values: :private (the default; stream can only be accessed by the creating thread) :lock (the old, often slow, default); most accesses to stream involve locking :nil or :external (stream code doesn't enforce exclusion). 2006-07-23 06:20 gb * [r4894] : *** empty log message *** 2006-07-23 05:57 gb * [r4893] : Bootstrapping stuff. Move along, nothing to see here. 2006-07-20 10:20 gb * [r4892] : *** empty log message *** 2006-07-20 09:43 gb * [r4890] : First (rough) cut at making streams "private" (effectively always locked by the "owning" thread) to reduce some locking overhead. More to come, some of which will be a little tricky to bootstrap. 2006-07-20 07:20 gb * [r4889] : *** empty log message *** 2006-07-20 07:09 gb * [r4888] : subtag_lisp_thread -> subtag_basic_stream. 2006-07-20 07:09 gb * [r4887] : new (istruct-based) LISP-THREAD accessors. 2006-07-20 07:08 gb * [r4886] : Bump. 2006-07-20 07:07 gb * [r4885] : Introduce "owner" slot in basic ioblock. A little tricky to bootstrap. 2006-07-20 07:06 gb * [r4884] : Use new (istruct-based) lisp-thread accessors. 2006-07-20 07:03 gb * [r4883] : Get rid of the lisp-thread subtag (lisp-thread's are ISTRUCTs now); re-use it for (to-be-introduced) BASIC-STREAM. 2006-07-15 21:20 gb * [r4882] : *** empty log message *** 2006-07-15 21:02 gb * [r4881] : Insert some fresh lines in the IMPOSSIBLE-NUMBER condition :REPORT method. 2006-07-15 21:00 gb * [r4880] : x862-%aref1,x862-%aset1: don't try to do simple vector operations unless we're sure that the vector is a (SIMPLE-ARRAY * (*)) and we know the element-type (these were leaping into the simple-typed-vector case when the element-type was known but the vector wasn't known to be SIMPLE. 2006-07-14 12:20 gb * [r4879] : *** empty log message *** 2006-07-14 11:57 gb * [r4878] : NX-FORM-TYPEP: unqoute constants. 2006-07-13 22:20 gb * [r4877] : *** empty log message *** 2006-07-13 22:01 gb * [r4876] : SPECIALIZED-ARRAY-ELEMENT-TYPES needs to know about FIXNUM arrays. 2006-07-11 21:20 gb * [r4875] : *** empty log message *** 2006-07-11 20:23 gb * [r4874] : Destination and 1st source arg overlap (2-address instructions) in natural logical, arithmetic operations. 2006-07-05 09:20 gb * [r4873] : *** empty log message *** 2006-07-05 08:38 gb * [r4872] : Update. 2006-07-05 08:22 gb * [r4871] : Reluctantly put :MCL back on *features*; its removal needs to be better publicized. 2006-07-05 05:20 gb * [r4870] : *** empty log message *** 2006-07-05 04:24 gb * [r4869] : Use NATIVE-TRANSLATED-NAMESTRING on the PROGRAM arg to RUN-PROGRAM. 2006-07-05 04:23 gb * [r4868] : Bump. 2006-06-29 08:21 gb * [r4867] : *** empty log message *** 2006-06-29 08:11 gb * [r4866] : Bump image-abi-version. 2006-06-28 12:20 gb * [r4865] : *** empty log message *** 2006-06-28 11:24 gb * [r4864] : *rtld-default* wants to be a null pointer on DarwinPPC32 (at the very least, that platform's xFindSymbol gets confused by anything but a null pointer/global lookup.) 2006-06-28 11:20 gb * [r4863] : *** empty log message *** 2006-06-28 11:09 gb * [r4862] : Bump. Call the FreeBSD/AMD64 version "Alpha", which may be a bit generous (pretty regular deadlock, random thread creation errors, other bugs.) 2006-06-28 11:07 gb * [r4861] : new; awfully similar to linux x86-64 version. 2006-06-28 10:20 gb * [r4860] : *** empty log message *** 2006-06-28 10:04 gb * [r4859] : FreeBSD doesn't do a good job of identifying the cause of XMM FP exceptions. Try to do that by looking at MXCSR bits. 2006-06-28 10:03 gb * [r4858] : FreeBSD syscalls set carry flag to indicate error return. 2006-06-28 10:03 gb * [r4857] : Some FreeBSD changes/fixes. 2006-06-28 10:02 gb * [r4856] : Remove a #+freebsd-target breakpoint. 2006-06-28 00:20 gb * [r4855] : *** empty log message *** 2006-06-28 00:08 gb * [r4854] : fix (?) nested #+/#-. Maybe. 2006-06-27 23:30 gb * [r4853] : Fix Linux def of IS_MAYBE_INT_TRAP. 2006-06-27 23:27 gb * [r4852] : (possibly) make get_r_debug more portable; possibly break it completely ... 2006-06-27 15:20 gb * [r4851] : *** empty log message *** 2006-06-27 14:55 gb * [r4850] : map the altstack read/write/execute. 2006-06-27 14:54 gb * [r4849] : Fix os_get_stack_bounds() for FreeBSD. 2006-06-27 14:54 gb * [r4848] : FreeBSD needs prototypes from 2006-06-27 14:53 gb * [r4847] : New (renamed) kernel-build-direcory for :freebsdx8664. 2006-06-27 14:53 gb * [r4846] : freebsd x8664 GPR indices. (Needs testing.) Don't conditionalize as much code. Indices/offsets are platform-specific, but accessors defiend in terms of them needn't be. 2006-06-27 14:51 gb * [r4845] : Remove 1st line (bogus comment.) 2006-06-27 14:50 gb * [r4844] : More freebsd x8664 features. 2006-06-27 14:50 gb * [r4843] : Update for FreeBSD/amd64. 2006-06-27 14:47 gb * [r4842] : FreeBSD backend has :freebsdx8664-target as a target-specific-feature. 2006-06-27 11:20 gb * [r4841] : *** empty log message *** 2006-06-27 11:06 gb * [r4840] : directory rename 2006-06-27 10:59 gb * [r4839] : FreeBSD xload-backend. 2006-06-27 10:47 gb * [r4838] : Some macros for FreeBSD changes. 2006-06-27 10:46 gb * [r4837] : FreeBSD changes (at least enough to recognize UUOs, which generate SIGBUS vice SIGSEGV.) 2006-06-27 10:45 gb * [r4836] : Define freebsd_sigreturn. TODO (maybe): copy the sigtramp code when evacuating the altstack. 2006-06-27 10:44 gb * [r4835] : FreeBSD sigreturn from suspend_resume_handler. 2006-06-27 10:44 gb * [r4834] : Some FreeBSD conditionalization; get_r_debug() for ELF platforms. (Needs testing under Linux.) 2006-06-27 10:43 gb * [r4833] : Don't import metering-control; do import get_r_debug(). 2006-06-27 10:42 gb * [r4832] : New file. Should check word-size stuff (here and in other versions.) 2006-06-27 10:42 gb * [r4831] : Find MACHINE-VERSION string via #_sysctl on FreeBSD. 2006-06-27 10:39 gb * [r4830] : Signed char nonsense (should use target ftd, not reader macros here.) 2006-06-27 10:38 gb * [r4829] : Some FreeBSD support. 2006-06-27 10:36 gb * [r4828] : Some FreeBSD changes/conditionalization. Maybe not enough. 2006-06-27 10:35 gb * [r4827] : Recognize FREEBSD platform-os-name. 2006-06-27 10:34 gb * [r4826] : Ignore *fsl. 2006-06-27 10:34 gb * [r4825] : FreeBSD syscalls. 2006-06-27 10:33 gb * [r4824] : Some FreeBSD features. 2006-06-27 10:33 gb * [r4823] : Lots of FreeBSD conditionalization, changes. MAY HAVE BROKEN LINUX PORTS; need to recompile on those platforms and test. Now depends on an imported lisp-kernel function on ELF platforms (Linux/FreeBSD), and therefore depends on recent kernel changes. 2006-06-27 10:29 gb * [r4822] : Fix platform-syscall-mask in freebsd backend. 2006-06-27 10:27 gb * [r4821] : Don't import metering-control; do import get-r-debug (for Elf shared libraries.) A yield-syscall constant for FreeBSD. (This is all a bad idea ...) 2006-06-27 10:26 gb * [r4820] : Don't import metering-control; do import get-r-debug (for Elf shared libraries.) 2006-06-27 10:26 gb * [r4819] : Use the non-eabi syscall on freebsd as well. 2006-06-26 23:20 gb * [r4818] : *** empty log message *** 2006-06-26 22:38 gb * [r4817] : SET-BIGITS-AND-HEADER-FOR-FIXNUM-OVERFLOW: need to change the (wrong) sign of the top 3 (fixnumshift) bits only. Where did the constant #xffff000000000000 come from ? 2006-06-26 22:35 gb * [r4816] : Fix some indentation. 2006-06-26 03:20 gb * [r4815] : *** empty log message *** 2006-06-26 02:32 gb * [r4814] : FreeBSD x8664 backend; change initialiation of x8664 foreign-type stuff. 2006-06-26 02:31 gb * [r4813] : PLATFORM-OS-FREEBSD. 2006-06-24 02:20 gb * [r4812] : *** empty log message *** 2006-06-24 01:52 gb * [r4811] : Ignore Developer subdirectory. c 2006-06-24 01:28 gb * [r4810] : Update some. 2006-06-24 01:20 gb * [r4809] : *** empty log message *** 2006-06-24 01:11 gb * [r4808] : New file. 2006-06-24 00:56 gb * [r4807] : Bump. 2006-06-24 00:52 gb * [r4806] : Export DEFLOADVAR. 2006-06-22 20:20 gb * [r4805] : *** empty log message *** 2006-06-22 19:39 gb * [r4804] : new 2006-06-22 19:30 gb * [r4803] : #+linuxppc-target -> #+linux-target. 2006-06-22 19:28 gb * [r4802] : Bump. 2006-06-21 21:20 gb * [r4801] : *** empty log message *** 2006-06-21 21:15 gb * [r4800] : Define new string subtags; handle them in ppc64 misc_ref/misc_set. 2006-06-21 21:14 gb * [r4799] : Fix %INIT-MISC (both ppc2 and ppc64 versions) to handle FIXNUM vectors, new UCS4 strings. 2006-06-21 21:14 gb * [r4798] : Fix %INIT-MISC (bot ppc2 and ppc64 versions) to handle FIXNUM vectors, new UCS4 strings. 2006-06-21 21:13 gb * [r4797] : Use 10.4u headers from XCode SDK. 2006-06-21 21:11 gb * [r4796] : Add SUBTAG-NEW-STRING. 2006-06-21 21:11 gb * [r4795] : MAX-32-BIT-IVECTOR-SUBTAG wasn't updated with the introduction of fixnum vectors, (unused) UCS4 "new"strings. 2006-06-16 10:20 gb * [r4794] : *** empty log message *** 2006-06-16 09:50 gb * [r4793] : Re-compute static NSString info after opening an extension framework. 2006-06-16 09:45 gb * [r4792] : Use new functions for augmenting interfaces, loading framework. 2006-06-16 09:43 gb * [r4791] : Juggle things around so that add-on framework bundles get loaded on appication re-launch. REVIVE-OBJC-CLASSES tells MAP-OBJC-CLASSES not to hit the database, since (a) it may not be there, or may not be where we can find it yet and (b) it's probably more reasonable to do that than it would be to teach CLOS about classes that're "discovered" when the appication's launched on a newer OS release than it was saved on. 2006-06-16 09:39 gb * [r4790] : Touch tha application bundle before saving an image into it. That -may- convincve LaunchServices to do the right thing. 2006-06-16 09:38 gb * [r4789] : Reload any addon/framework bundles that had been loaded before the image was saved when starting a bundled application. Juggle things around, so that that works (need to be able to use low-level bridge features early.) 2006-06-16 09:36 gb * [r4788] : Disable the #? reader macro. 2006-06-14 22:20 gb * [r4787] : *** empty log message *** 2006-06-14 21:42 gb * [r4786] : Fix #+APPLY+ lfun-bits. 2006-06-13 03:20 gb * [r4785] : *** empty log message *** 2006-06-13 02:24 gb * [r4784] : PPC version of %find-register-argument-value had the same sort of fencepost as the x86-64 version had a few days ago. 2006-06-11 06:20 gb * [r4783] : *** empty log message *** 2006-06-11 05:52 gb * [r4782] : %ARGLIST-INTERNAL: if generic-function, return its GENERIC-FUNCTION-LAMBDA-LIST. 2006-06-11 05:51 gb * [r4781] : Give reader/writer method args more descriptive names. (Barely more descriptive ...) 2006-06-11 02:20 gb * [r4780] : *** empty log message *** 2006-06-11 01:25 gb * [r4779] : Arrgh. In %MAYBE-STD-SETF-SLOT-VALUE-USING-CLASS, we know nothing about the type of the instance, so don't use INSTANCE.SLOTS (instead, let INSTANCE-SLOTS do a TYPECASE; it's inlined at this point.) This was keeping non-standard instances - like ObjC classes and metaclasses - from having some of their slots initialized correctly. 2006-06-11 01:22 gb * [r4778] : PRINT-OBJECT methods for objc-class, objc-metaclass: don't die if the class is unnamed (makes it easier to print classes for debugging during initialization.) 2006-06-10 19:20 gb * [r4777] : *** empty log message *** 2006-06-10 19:05 gb * [r4776] : Fix LinuxPPC64 versions of XP-FPSCR-INFO and XP-DOUBLE-FLOAT. 2006-06-08 11:20 gb * [r4775] : *** empty log message *** 2006-06-08 11:01 gb * [r4774] : Update a bit. 2006-06-08 10:20 gb * [r4773] : *** empty log message *** 2006-06-08 09:56 gb * [r4772] : Bump image-abi-version. 2006-06-08 09:46 gb * [r4771] : Bump (min) fasl version. 2006-06-08 09:35 gb * [r4770] : Bump (current) fasl-version. 2006-06-08 09:28 gb * [r4769] : -O2. 2006-06-08 09:26 gb * [r4768] : Bump (max) fasl version. 2006-06-08 09:20 gb * [r4767] : *** empty log message *** 2006-06-08 09:18 gb * [r4766] : Bump; give x86-64 port the benefit of the doubt and claim that it's a beta. 2006-06-08 09:18 gb * [r4765] : Enable -O2. 2006-06-08 09:17 gb * [r4764] : Some commented-out debugging options; enable -O2. 2006-06-08 08:55 gb * [r4763] : thread_signal_setup() moved here. Handle suspend (and maybe someday resume, again) signals on the altstack, since we don't really want a signal context in the middle of the tagged lisp stack ... 2006-06-08 08:52 gb * [r4762] : thread_signal_setup() moved to PLATFORM-exceptions.c; some constants/prototypes moved to Threads.h. 2006-06-08 08:51 gb * [r4761] : suspend_resume_handler() installed via install_signal_handler(); thread_signal_setup() moved here, to do that installation. 2006-06-08 08:50 gb * [r4760] : Don't enable_fp_exceptions() twice. Don't call (or define) prepare_for_the_worst(), which is kind of too bad ... 2006-06-08 08:49 gb * [r4759] : suspend/resume constants, prototypes. 2006-06-08 07:20 gb * [r4758] : *** empty log message *** 2006-06-08 07:11 gb * [r4757] : Back out of ill-advised changed to %LOAD; let OPTIMIZE proclamations persist beyond the LOAD. 2006-06-08 04:20 gb * [r4756] : *** empty log message *** 2006-06-08 03:45 gb * [r4755] : No nore %tsp/%Rnext_tsp. 2006-06-08 03:38 gb * [r4754] : Don't maintain current/next tsp in MMX regs. Do maintain them in TCR. 2006-06-08 03:37 gb * [r4753] : Don't maintain %tsp when stack-consing; update (decrement) tcr.next_tsp atomically, use %stack_temp as a temporary (old tsp), finish frame by storing into tcr.save_tsp. 2006-06-08 03:34 gb * [r4752] : tsp is in the TCR now. 2006-06-08 03:33 gb * [r4751] : No more %tsp, %Rnext_tsp. Define %stack_temp. 2006-06-08 03:32 gb * [r4750] : No more Itsp, Inext_tsp. 2006-06-08 03:30 gb * [r4749] : %CURRENT-TSP: the tsp is in the TCR now. 2006-06-08 03:30 gb * [r4748] : Don't try to maintain %tsp (since it no longer exists ...). Use %stack-temp in some cases. 2006-06-08 03:29 gb * [r4747] : No more %tsp or %next-tsp. %stack-temp is a symbolic name for an MMX register that can be used for (e.g.) linking frames on the temp/foreign stacks (or for any other purpose ...) 2006-06-07 22:20 gb * [r4746] : *** empty log message *** 2006-06-07 21:29 gb * [r4745] : No more %Rforeign_sp. Keep tcr.next_tsp in sync with %next_tsp Check to ensure that %tsp,tcr.save_tsp,and tcr.next_tsp all have the same value when allocating on the temp stack. (This will soon change, so that %tsp and %next_tsp will not be canonical and the tcr fields are, then will change so that %tsp and %next_tsp will cease to exist.) 2006-06-07 21:25 gb * [r4744] : No more %Rforeign_sp. 2006-06-07 21:25 gb * [r4743] : Keep tcr.next_tsp in sync with %Rnext_tsp. No more %Rforeign_sp, but we can use %mm5 as a temporary. 2006-06-07 21:23 gb * [r4742] : Find the foreign RSP from the TCR; don't need FP regs for this. (They're sometimes invalid and sometimes missing, which is the point of recent changes.) 2006-06-07 21:21 gb * [r4741] : No foreign_sp. Rename %next_tsp to %Rnext_tsp to avoid potential m4 lossage. 2006-06-07 21:20 gb * [r4740] : *** empty log message *** 2006-06-07 21:20 gb * [r4739] : No Iforeign_sp. 2006-06-07 21:20 gb * [r4738] : Initialize tcr->next_tsp on X86. 2006-06-07 21:19 gb * [r4737] : Bump. 2006-06-07 21:19 gb * [r4736] : Get foreign-sp from the TCR. 2006-06-07 21:18 gb * [r4735] : More bootstrapping: no more foreign-sp register (though some code still uses %mm5 as a temporary, the "canonical" foreign SP is maintained in the TCR.) 2006-06-07 11:20 gb * [r4734] : *** empty log message *** 2006-06-07 10:53 gb * [r4733] : More bootstrapping: continue to maintain the foreign sp in both the TCR and the MMX register, but consider the TCR version to be canonical. 2006-06-07 10:52 gb * [r4732] : When making a catch/unwind-protect frame, get the foreign SP out of the TCR. 2006-06-07 09:20 gb * [r4731] : *** empty log message *** 2006-06-07 08:46 gb * [r4730] : More bootstrapping: if the %tsp MMX register doesn't match the value in the tcr, trap. Likewise for %foreign-sp. 2006-06-07 08:21 gb * [r4729] : *** empty log message *** 2006-06-07 07:30 gb * [r4728] : Try to keep tcr.save_tsp in synch with %tsp, and likewise for %Rforeign_sp/tcr.foreign_sp. (The idea is to keep the temp/foreign stack pointers in the tcr and only use the MMX registers as temporaries, since it seems that handlers for asynchronous signals can't count on the MMX register values in the signal context being correct. This seems to either be a Linux kernel bug or ill-considered if it's intentional behavior, but it seems better to work around it this way.) This is just an early step in bootstrapping that change; it's being checked into CVS so that it doesn't get lost (and so that I can recover to this point if/when I hose things on a later step.) 2006-06-07 07:24 gb * [r4727] : Add a next-tsp slot to the tcr. 2006-06-07 07:22 gb * [r4726] : Add a next_tsp slot to the tcr. 2006-06-06 12:21 gb * [r4725] : *** empty log message *** 2006-06-06 10:25 gb * [r4724] : no hope 2006-06-06 10:22 gb * [r4723] : %find-register-argument-value: Current frame (cfp argument) could be an exception callback frame (xcf-p may be true of it), in which case the saved register value is in the xcf.xp context. (Basically, this means look at the first frame if it's an xcf, otherwise skip it and start with the child.) 2006-06-06 08:21 gb * [r4722] : *** empty log message *** 2006-06-06 07:48 gb * [r4721] : Using imul to box fixnums is smaller and seems to be faster than using lea with a scale factor. 2006-06-06 06:20 gb * [r4720] : *** empty log message *** 2006-06-06 05:56 gb * [r4719] : I don't think that suspend_resume_handler wants to have suspend signals unmasked, does it ? 2006-06-05 19:20 gb * [r4718] : *** empty log message *** 2006-06-05 18:43 gb * [r4717] : area-readonly is in ccl:: now. 2006-06-03 09:20 gb * [r4716] : *** empty log message *** 2006-06-03 09:11 gb * [r4715] : In _SPnthrowvalues: don't skip the last catch frame, do skip others (test was backwards. As it probably had been in the single-value case at one point.) 2006-06-02 10:20 gb * [r4714] : *** empty log message *** 2006-06-02 09:48 gb * [r4713] : Don't print trace indent/outdent stuff to *terminal-io*; do print it to *trace-output*. 2006-06-02 02:20 gb * [r4712] : *** empty log message *** 2006-06-02 02:20 gb * [r4711] : There's a T in subtag. 2006-06-02 02:17 gb * [r4710] : X862-MVCALL wasn't handling things like: (multiple-value-bind (...) (multiple-value-call #'foo ...)) correctly. 2006-06-02 02:16 gb * [r4709] : Defvars for global optimization quantities moved to level-0. 2006-06-02 02:15 gb * [r4708] : Bind the variables which hold global compiler optimization quantities, since people persist in doing things like: (declaim (optimize )) without bothering to use EVAL-WHEN or otherwise thinking very clearly. Note that this means that anyone who wanted such a declamation to have persistent global effects is now SOL. Blame whoever wrote (for instance) clx/glx.lisp. 2006-06-02 02:11 gb * [r4707] : Move *nx-speed*, etc here, so that they can be referenced early in the cold load. 2006-06-01 19:20 gb * [r4706] : *** empty log message *** 2006-06-01 19:13 gb * [r4705] : In %FIND-REGISTER-ARGUMENT-VALUE, pass a pointer to ENCODED-GPR-LISP if we have to get the register value from an exception frame. 2006-05-31 02:20 gb * [r4704] : *** empty log message *** 2006-05-31 01:27 gwking * [r4703] : Updated asdf to the current CVS head (revision 1.98) 2006-05-30 23:20 gb * [r4702] : *** empty log message *** 2006-05-30 23:12 gb * [r4701] : bump again. 2006-05-30 22:31 gb * [r4700] : When unwinding the stack, try to ensure that tcr.xframe is updated from the catch frame (a) in all execution paths and (b) before RSP is updated. 2006-05-30 22:30 gb * [r4699] : Handle (simple) fixnum vectors. 2006-05-30 22:29 gb * [r4698] : Add a DEFTYPE for fixnum vectors. 2006-05-30 22:29 gb * [r4697] : Update some more ... 2006-05-30 22:23 gb * [r4696] : There are two 'e's in UNPROCESSED. Apparently. 2006-05-30 22:20 gb * [r4695] : *** empty log message *** 2006-05-30 22:15 gb * [r4694] : updated 2006-05-30 08:21 gb * [r4693] : *** empty log message *** 2006-05-30 07:58 gb * [r4692] : Bump. 2006-05-30 07:48 gb * [r4691] : altstack_interrupt_handler: having made copies of the signal context and siginfo on the foreign stack, don't pass the originals to interrupt_handler(). (altstack_interrupt_handler() looks suspiciously like altstack_signal_handler(), which probably had the same bug a few months ago ...) Since the alternate signal stack (as we use it) isn't reentrant, the next signal (perhaps a suspend signal) overwrote the context. Which is one reason to have made a copy in the first place ... 2006-05-30 07:43 gb * [r4690] : In THREAD-ENABLE, use the new %TIMED-WAIT-ON-SEMAPHORE-PTR; regardless of whether or not a timeout is a good idea here, make sure that the timeout isn't fooled by #$EINTR. 2006-05-30 07:39 gb * [r4689] : Rename the old %TIMED-WAIT-ON-SEMAPHORE-PTR to %PROCESS-WAIT-ON-SEMAPHORE-PTR, since it basically waits indefinitely. Split out the functionality in TIMED-WAIT-ON-SEMAPHORE into a new %TIMED-WAIT-ON-SEMAPHORE-PTR function, so that it's possible to do a "timed wait" that handles interruptions, etc. when the caller just has a "raw" (native) semaphore. Change callers, endure warnings. New binaries soon. 2006-05-30 07:36 gb * [r4688] : When waiting for a lock, use the new %PROCESS-WAIT-ON-SEMAPHORE-PTR. 2006-05-30 05:20 gb * [r4687] : *** empty log message *** 2006-05-30 04:34 gb * [r4686] : The clear_tcr_save_allocptr_tag_instruction opcode array was one byte too small ... 2006-05-29 20:20 gb * [r4685] : *** empty log message *** 2006-05-29 20:03 gb * [r4684] : In %OS-GETCWD: don't trust syscall-based getcwd (it doesn't work as it once did/is documented to on linuxx8664); use the library function on all platforms. 2006-05-29 20:02 gb * [r4683] : Clear interrupt_pending flag before processing interrupt. 2006-05-28 23:20 gb * [r4682] : *** empty log message *** 2006-05-28 23:01 gb * [r4681] : If we handle a SIGNAL_FOR_PROCESS_INTERRUPT trap, clear tcr->interrupt_pending. Back out of the hack that tried to work around a Darwin bug (FP exception flags were cleared in the MSR after a thread had been interrupted via pthread_kill()), since the fix seems not to be GC-safe. Add some debugging code to print when the exception lock is acquired/released. 2006-05-28 07:20 gb * [r4680] : *** empty log message *** 2006-05-28 07:10 gb * [r4679] : long gone 2006-05-24 03:20 gb * [r4678] : *** empty log message *** 2006-05-24 02:25 gb * [r4677] : D'oh! 2006-05-23 05:20 gb * [r4676] : *** empty log message *** 2006-05-23 05:01 gb * [r4675] : New file. 2006-05-23 04:57 gb * [r4674] : PROCESS-SUSPEND: don't allow current thread to suspend itself. 2006-05-23 04:57 gb * [r4673] : Use the SP red zone to load SINGLE-FLOAT constants. SINGLE-FLOAT foreign args are stored in the low 32 bits of each 64-bit arg word, regardless of whether they're also passed in f1-f13 as well. 2006-05-23 04:55 gb * [r4672] : Callback fixes: SINGLE-FLOATs passed in memory are passed in the low 32 bits of the 64-bit word. 2006-05-23 04:54 gb * [r4671] : Use the red zone (on SP) for LOAD-SINGLE-FLOAT-CONSTANT, LOAD-DOUBLE-FLOAT-CONSTANT. 2006-05-23 04:53 gb * [r4670] : "poweropen ABI" (Darwin) callback fixes: SINGLE-FLOATs passed in memory are passed as ... SINGLE-FLOATs, so don't use %GET-SINGLE-FLOAT-FROM-DOUBLE-PTR to access them. 2006-05-22 10:20 gb * [r4669] : *** empty log message *** 2006-05-22 09:53 gb * [r4668] : Bump. 2006-05-22 09:32 gb * [r4667] : fixnum->float vinsns. 2006-05-22 09:31 gb * [r4666] : Numeric contagion on binops, fixnum->float. 2006-05-22 09:29 gb * [r4665] : Do stupid self-tail-call tricks on PPC as well. 2006-05-22 09:28 gb * [r4664] : Load tcr.cs-limit a little earlier when creating a stack frame inline. Doesn't seem to affect latency much. 2006-05-22 09:27 gb * [r4663] : Comment formatting on the first line of the file. 2006-05-22 09:26 gb * [r4662] : Check argument counts on self-calls. (AFAIK, this hasn't been done for at least 20 years in MCL/OpenMCL ...) 2006-05-21 14:20 gb * [r4661] : *** empty log message *** 2006-05-21 13:34 gb * [r4660] : Long-standing X862-COPY-REGISTER bug. 2006-05-21 13:32 gb * [r4659] : Implement (late) div2; use (simple) numeric-contagion stuff for arithmetic ops. 2006-05-21 13:31 gb * [r4658] : Use div2 for /-2. 2006-05-21 13:31 gb * [r4657] : Binop numeric contagion. 2006-05-21 13:20 gb * [r4656] : *** empty log message *** 2006-05-21 12:49 gb * [r4655] : ACODE-REAL-CONSTANT-P. 2006-05-21 12:49 gb * [r4654] : Weird case in ppc2-copy-register wasn't weird, just wrong. 2006-05-21 12:20 gb * [r4653] : *** empty log message *** 2006-05-21 12:17 gb * [r4652] : Fixnum vector stuff. 2006-05-21 12:17 gb * [r4651] : PPC64: %typep recognizes SIMPLE-FIXNUM-VECTORs. 2006-05-21 12:17 gb * [r4650] : Fixnum vector stuff. FIXNUM->FPR uses SP red zone. 2006-05-21 12:15 gb * [r4649] : %fixnum-to-single, %fixnum-to-double. Fixnum vector vref/vset. 2006-05-21 12:15 gb * [r4648] : %fixnum-to-single, %fixnum-to-double. 2006-05-21 12:14 gb * [r4647] : More operator result types. 2006-05-21 11:20 gb * [r4646] : *** empty log message *** 2006-05-21 10:35 gb * [r4645] : Integrate specialized FIXNUM array-element-type into type system, CLOS. (Knock wood.) 2006-05-21 10:34 gb * [r4644] : Support for (out-of-line) fixnum-vector access (.SPmisc_ref, .SPmisc_set). 2006-05-21 10:33 gb * [r4643] : New subtag for :fixnum-vector. 2006-05-21 10:32 gb * [r4642] : 64-bit vset wasn't handling non-constant indices correctly. 64-bit vset, vref: handle :FIXNUM-VECTOR. 2006-05-21 07:20 gb * [r4641] : *** empty log message *** 2006-05-21 06:22 gb * [r4640] : Spell istruct-typep correctly. 2006-05-21 00:20 gb * [r4639] : *** empty log message *** 2006-05-21 00:13 gb * [r4638] : Do some self-tail calls (fixed args, no stack args) without restoring/re-saving NVRs. 2006-05-19 21:20 gb * [r4637] : *** empty log message *** 2006-05-19 21:07 gb * [r4636] : RUN-PROGRAM: waiting for the child's completion semaphore (when :WAIT is true) happens inside an UNWIND-PROTECT cleanup and therefore needs to explicitly allow interrupts. 2006-05-19 21:06 gb * [r4635] : Add SEMAPHOREP, make it a type-predicate for SEMAPHORE. 2006-05-19 19:20 gb * [r4634] : *** empty log message *** 2006-05-19 18:23 gwking * [r4633] : Added help and redisplay to select-item-from-list. Also changed it so that it displays all items before trying to get an answer. 2006-05-19 18:20 gb * [r4632] : *** empty log message *** 2006-05-19 18:00 gwking * [r4631] : The *fsl wasn't cutting it for me..., take two 2006-05-19 17:57 gwking * [r4630] : The *fsl wasn't cutting it for me... 2006-05-19 17:50 gb * [r4629] : Bump. 2006-05-19 17:49 gb * [r4628] : Change the EVAL-WHEN that requires syscall definitions so that it can be moused on. 2006-05-19 17:49 gb * [r4627] : .SPeabi-syscall -> .SPsyscall. 2006-05-19 17:45 gb * [r4626] : Inline ADD2, unless we know that at least one arg couldn't be a fixnum. (Should do the same for SUB2.) This does indeed burn cache lines, but going out of line isn't cheap either. (100 million iterations of (+ x (+ y z)) - where X Y and Z are in fact fixnums and no overflow is involved - seems to be about 2X faster when inlined this way. Not sure if that'll be the case on PPC.) 2006-05-19 17:40 gb * [r4625] : Use NX-TARGET-TYPE in some of the acode-type stuff. 2006-05-18 21:20 gb * [r4624] : *** empty log message *** 2006-05-18 20:47 gb * [r4623] : When unwinding the stack (hopefully, in all places that do so ...), update tcr.xframe from catch_frame.xframe on all execution paths (in particular, do so before SP is restored from the catch_frame, since a tcr.xframe "beyond" SP is clearly bogus.) 2006-05-18 09:21 gb * [r4622] : *** empty log message *** 2006-05-18 08:34 gb * [r4621] : Bump. 2006-05-18 05:20 gb * [r4620] : *** empty log message *** 2006-05-18 05:12 gb * [r4619] : "late" acode-type stuff (add2/sub2/mul2) for PPC, too. 2006-05-18 04:53 gb * [r4618] : %FP-ERROR-FROM-STATUS: don't cons an &REST arg. Should do a quicker test (logandc2 and a shift) to see if any unmasked exceptions were raised before even calling this ... 2006-05-18 04:51 gb * [r4617] : If we have to spill SINGLE-FLOAT registers, spill them (boxed) on the vstack. 2006-05-18 04:50 gb * [r4616] : Work harder to find acode types in some cases. 2006-05-18 04:50 gb * [r4615] : Handle (new-style) byte registers in OP-REG. Don't want to use RET very often if at all, but disassembler should treat as as a :JUMP. 2006-05-18 04:48 gb * [r4614] : MUL2, DIV2. 2006-05-18 04:48 gb * [r4613] : A little of that there type-inferencing (or, more accurately, paying a little more attention here and there.) 2006-05-13 03:20 gb * [r4612] : *** empty log message *** 2006-05-13 02:40 gwking * [r4611] : added examples/addressbook.lisp 2006-05-12 01:20 gb * [r4610] : *** empty log message *** 2006-05-12 00:42 gb * [r4609] : Stop processing command-line arguments when "--" is encountered; anything after that "--" is retained in CCL:*UNPROCESSED-COMMAND-LINE-ARGUMENTS*. 2006-05-12 00:42 gb * [r4608] : Export command-line stuff 2006-05-12 00:40 gb * [r4607] : Lennart Staflin's fix to PRINT-OBJECT method for LOCKs. 2006-05-12 00:39 gb * [r4606] : In %DIFECTORY-LIST-NAMESTRING: all strings are base-strings (and always have been in OpenMCL.) 2006-05-12 00:37 gb * [r4605] : CCL:BASE-CHARACTER -> CL:BASE-CHAR. 2006-05-11 14:20 gb * [r4604] : *** empty log message *** 2006-05-11 14:11 gwking * [r4603] : Added a check that the directory exists to fid-open-output in order to make it clearer what is wrong when one is unable to rebuild interface files because they are trying to do so in a place without FFI files to read! 2006-05-11 11:20 gb * [r4602] : *** empty log message *** 2006-05-11 11:15 gb * [r4601] : Bump lisp-implementation-version, image-abi-version. 2006-05-11 11:04 gb * [r4600] : Bump (min) fasl version. 2006-05-11 10:52 gb * [r4599] : Bump (current) FASL version. 2006-05-11 10:52 gb * [r4598] : FASL-FILE-P looks for inital #xFF00 (big-endian) instead of just #xff, to avoid UTF-16 BOM confusion. 2006-05-11 10:35 gb * [r4597] : Bump (max) fasl version. 2006-05-11 10:32 gb * [r4596] : Little-endianness is a target-specific feature. 2006-05-11 10:20 gb * [r4595] : *** empty log message *** 2006-05-11 09:35 gb * [r4594] : Juggle PPC32 typecodes around: get rid of the (unused) old 16-bit string; add an (as yet unused) UCS-4 string type. These changes generally involve cross-compilation (or some very tricky bootstrapping that I didn't attempt) FASL versions and kernel ABI versions will need to change, and new binaries will be made available soon. 2006-05-11 02:20 gb * [r4593] : *** empty log message *** 2006-05-11 02:20 gb * [r4592] : Change CHARCODE-SHIFT constant on PPC32, so that the (currently 8-bit) character code is stored in the #x0000xx00 byte of a CHARACTER, rather than in the #x00xx0000 byte, This is hard to bootstrap natively (which is why I was trying to make ppc64->ppc32 cross-compilation work.) Change related macros/visnsns as well. 2006-05-11 01:20 gb * [r4591] : *** empty log message *** 2006-05-11 01:00 gb * [r4590] : Reintroduce a TARGET-ARCH-CASE, but I think that this is just hard to do right the way that it's being done. 2006-05-11 00:59 gb * [r4589] : BOGUS-THING-P is hard to do right when cross-compiling (because word-size differences aren't handled correctly by macros.) Until the macros are fixed, don't even try ... 2006-05-11 00:58 gb * [r4588] : In NEW-CLASS-WRAPPER-HASH-INDEX, user target::target-most-positive-fixnum (not the host's MOST-POSITIVE-FIXNUM). 2006-05-11 00:57 gb * [r4587] : Use NX-TARGET-TYPE in at least a few places. 2006-05-11 00:56 gb * [r4586] : XCOMPILE-LAMBDA binds *TARGET-BACKEND*. 2006-05-11 00:55 gb * [r4585] : NX-TARGET-TYPE: there may be other differences between host and target types, but try to turn references to FIXNUM to the appropriate target-specific type. 2006-05-10 03:20 gb * [r4584] : *** empty log message *** 2006-05-10 03:01 gb * [r4583] : UNBOX-BASE-CHAR just checks that the lowbyte is SUBTAG-CHARACTER; it used to rotate things around to check the high bits of the char code, but all chars have been base-chars for several years now ... UNBOX-CHARACTER was unused. (This is the first step of what may be several bootstrapping steps to change PPC32::CHARCODE-SHIFT from 16 to 8, so that we have 24 easily-accessed char-code bits on PPC32.) 2006-05-09 18:20 gb * [r4582] : *** empty log message *** 2006-05-09 17:48 gb * [r4581] : No psn in cs_area. 2006-05-09 09:20 gb * [r4580] : *** empty log message *** 2006-05-09 08:42 gb * [r4579] : Back out of the extra Boolean arg to mark_simple_area_range(). 2006-05-09 08:33 gb * [r4578] : Endianness is also target-specific. 2006-05-09 08:25 gb * [r4577] : Thread stack areas are "owned by" the containing TCR, to make low-level debugging a bit easier. 2006-05-09 08:25 gb * [r4576] : Mostly some added consistency/sanity checks and debugging printfs. 2006-05-09 08:24 gb * [r4575] : Note that pc_luser_xp() needs an extra (signed_natural *) arg, for allocptr displacement. 2006-05-09 08:23 gb * [r4574] : Lots of #ifdef DEBUG fprintfs. Change the way that pc_luser_xp() and normalize_tcr() handle alloc regs/ saved TCR fields. The general idea is that the after they run, all threads have tcr->save_allocptr/tcr_save_allocbase set to VOID_ALLOCPTR; if the thread was interrupted in lisp code, the allocbase register will be set to VOID_ALLOCPTR and allocptr will be set to either VOID_ALLOCPTR or to VOID_ALLOCPTR-disp. On an interrupt, the caller needs to pass in a signed_natural * to pick up the allocptr displacement (if consing was interrupted at the alloc trap); the interrupt handler uses this value to fix up alloptr when it returns if consing was interrupted. This is all pretty clearly right, but something still goes wrong under some combination of high thread stress, frequent allocation, and blocking ff-calls. 2006-05-09 08:21 gb * [r4573] : *** empty log message *** 2006-05-09 08:17 gb * [r4572] : No psn currently available. 2006-05-09 08:16 gb * [r4571] : Bug() passes xp to lisp_Debugger(), where it may come in handy ... 2006-05-09 08:16 gb * [r4570] : mark_simple_area_range() takes an added Boolean arg. 2006-05-09 08:13 gb * [r4569] : $xfunbnd was used after all. Restore it, and recycle $XNOCDR. 2006-05-09 08:13 gb * [r4568] : area struct's owner is a (TCR *). 2006-05-09 08:12 gb * [r4567] : $xfunbnd was used after all. Restore it. 2006-05-09 08:11 gb * [r4566] : :64-bit-target is ... target-specific. 2006-05-09 08:11 gb * [r4565] : :32-bit-target is ... target-specific. 2006-05-09 07:20 gb * [r4564] : *** empty log message *** 2006-05-09 06:40 gb * [r4563] : Foreign thread callback stuff moved here. 2006-05-09 06:39 gb * [r4562] : Support for BOGUS-THING-P on x8664. 2006-05-09 06:39 gb * [r4561] : Move foreign-thread callback stuff up to l1-lisp-threads.lisp. 2006-05-09 05:20 gb * [r4560] : *** empty log message *** 2006-05-09 04:58 gb * [r4559] : misc_ref/misc_set jump tables on PPC64 handle array and vector headers. 2006-05-09 04:57 gb * [r4558] : Synch errors $xbadvec, $xsetbadvec (neither of which should ever happen) with the kernel. 2006-05-09 04:56 gb * [r4557] : %PROCESS-PRESET-INTERNAL doesn't need the TCR either. 2006-05-09 04:55 gb * [r4556] : $XFUNBND is no more. 2006-05-09 04:55 gb * [r4555] : Define the format strings for $xbadvec, $xsetbadvec. 2006-05-08 10:20 gb * [r4554] : *** empty log message *** 2006-05-08 09:26 gb * [r4553] : .SPstkconslist_star: ensure that flags are set to the result of decrementing %nargs at the end of the loop. 2006-05-08 09:25 gb * [r4552] : uuo-error-udf: %err-disp-internal wants a list of condition args, not a single arg. 2006-05-08 09:24 gb * [r4551] : %PROCESS-PRESET-INTERNAL doesn't need CS-AREA. 2006-05-08 09:23 gb * [r4550] : %FUNCTION: signal error via uuo_error_udf, not as a type error. 2006-05-07 23:20 gb * [r4549] : *** empty log message *** 2006-05-07 22:50 gb * [r4548] : INSPECTOR objects haven't had FORWARDED-P slots in a long time (they probably haven't -needed- them in even longer; I think that this had to do with the old partial MOP.) Don't define the slot for FUNCTION-INSPECTOR, and don't call the accessor in STANDARD-OBJECT-SETF-LINE-N. 2006-05-07 11:20 gb * [r4547] : *** empty log message *** 2006-05-07 10:39 gb * [r4546] : New file. 2006-05-07 10:34 gb * [r4545] : Combine the AREA_LOCK and TCR_LOCK into a single lock (TCR_AREA_LOCK). (This makes it a little harder to deadlock if code that wants both locks isn't careful to try to obtain them in some canonical order.) 2006-05-07 10:29 gb * [r4544] : Finally find those pesky registers. 2006-05-07 10:29 gb * [r4543] : Stub BOGUS-THING-P moves here. 2006-05-07 10:28 gb * [r4542] : FOREIGN-SYMBOL-ENTRY: return NIL if xFindSymbol/dlsym returns 0 on x8664. 2006-05-07 10:27 gb * [r4541] : Use PSRLQ/PSLLQ for single-float boxing/unboxing. (A few months ago, I'd mistakenly decided that it was faster to go through memory than to use these instructions. It seems to actually be substantially slower, as one would expect.) 2006-05-07 10:25 gb * [r4540] : Add support for PSRLQ/PSRLDQ/PSLLQ/PSLLDQ (sse2 shifts.) 2006-05-07 10:21 gb * [r4539] : Fix doc in comments. 2006-05-07 10:21 gb * [r4538] : GET/SET-REGISTER-VALUE. 2006-05-07 10:20 gb * [r4537] : YES-OR-NO-P: bell goes to *QUERY-IO*. 2006-05-07 10:20 gb * [r4536] : *** empty log message *** 2006-05-07 10:19 gb * [r4535] : Move GET/SET-REGISTER-VALUE to TARGET-backtrace.lisp. 2006-05-07 10:18 gb * [r4534] : Move (implementation-dependent) BOGUS-THING-P to TARGET-thread-utils.lisp. 2006-05-07 10:17 gb * [r4533] : Move (implementation-dependent) BOGUS-THING-P support here. 2006-05-07 10:16 gb * [r4532] : Don't set "owner" field of tcr.cs_area to psn. 2006-05-06 22:20 gb * [r4531] : *** empty log message *** 2006-05-06 21:27 gb * [r4530] : Ash left by constant: prefer the fixnum case to the "natural" case. 2006-05-06 21:26 gb * [r4529] : Pass the right number of args to NATURAL-SHIFT-LEFT and NATURAL-SHIFT-RIGHT. 2006-05-06 21:26 gb * [r4528] : Indentation. 2006-05-06 08:21 gb * [r4527] : *** empty log message *** 2006-05-06 08:09 gb * [r4526] : Let's see: if thread X holds local A and tries to obtain lock B, and thread Y holds lock B and tries to obtain lock A, what is that a classic case of ? Try to avoid that classic deadlock case in shutdown_thread_tcr() when HAVE_TLS is in effect. suspend_other_threads wants to hold two locks at the same time, and that's always a bit dangerous; it might be safer to merge TCR_LOCK and AREA_LOCK into a single lock that must be owned in order to manipulate either the tcr or area queues. 2006-05-06 08:04 gb * [r4525] : Make (%REALPATH "") behave like %REALPATH of the current directory namestring, as it does by default on some platforms (but not Linux.) 2006-05-06 02:20 gb * [r4524] : *** empty log message *** 2006-05-06 02:08 gb * [r4523] : x862-ff-call: really implement the simple-foreign-args stuff. Box u64/s64, or don't ... but don't do the el-bizarro <- with a (temp) node target. 2006-05-06 02:06 gb * [r4522] : Get PC-FROM-XCF right (it needs to be relative to the tagged entrypoint, not to the start of the underlying function-vector.) 2006-05-05 20:20 gb * [r4521] : *** empty log message *** 2006-05-05 19:24 gb * [r4520] : Preserve NEXT-METHOD-CONTEXT when heap-consing an &REST arg. (%allocptr and %next-method-context are the same register, so &rest clobbbers %next-method-context.) Could make separate entry points, so that this is only done when %next-method-context needs to be preserved ... 2006-05-05 19:21 gb * [r4519] : Use RESERVED-FRAME-MARKER. 2006-05-05 09:20 gb * [r4518] : *** empty log message *** 2006-05-05 08:23 gb * [r4517] : New file. Needs some work. 2006-05-05 08:23 gb * [r4516] : New file. 2006-05-05 08:22 gb * [r4515] : *** empty log message *** 2006-05-05 07:37 gb * [r4514] : Bump du jour. 2006-05-05 07:36 gb * [r4513] : %FF-CALL: maintain the distinction between the first 6 words of the outgoing frame (from which GPRs will be loaded) and other words (which stay on the stack); after the first 8 FP args, floats have to be passed on the stack, even if there's still room in the first 6 words of the outgoing frame. Need to understand the struct-by-value rules better: can a struct be passed partly in registers and partly on the stack ? RTFM (or, in this case, TF ABI spec.) 2006-05-05 07:31 gb * [r4512] : In X862-FF-CALL: - the general integer case should store relative to "other-offset" (rather than "gpr-offset") when more than 6 (not 8) GPR args have been passed. - both the single and double-float cases were incrementing "other-offset" twice when more than 8 FP args were involved. This had the effect of storing floats into every other word and writing beyond the bounds of the outgoing stack frame, possibly overwriting return addresses on the foreign stack. 2006-05-05 05:20 gb * [r4511] : *** empty log message *** 2006-05-05 05:12 gb * [r4510] : James Bielman's fix to READ-TOPLEVEL-FORM (INPUT-STREAM) (EOF handling/peek-char.) 2006-05-05 01:20 gb * [r4509] : *** empty log message *** 2006-05-05 00:30 gb * [r4508] : When binding FP args, decrement FPR-ARG-OFFSET by 8, not -8. (This could cause values outside of the stack frame to be read, but I still haven't found what's writing beyond the frame bounds.) 2006-05-05 00:27 gb * [r4507] : If its argument is an exception callback frame, %CFP-LFUN tries to determine the relative PC associated with the frame's nominal function and return that as a second value (just as it does for a real frame.) I think that this should work a high percentage of the time, but callers should generally be prepared to deal with a NULL second return value from %CFP-LFUN. 2006-05-04 17:20 gb * [r4506] : *** empty log message *** 2006-05-04 17:12 gb * [r4505] : More callback issues: since we're using the same stack pointer to return values as we are to find arguments, don't try to "optimize" its initialization. 2006-05-04 01:20 gb * [r4504] : *** empty log message *** 2006-05-04 00:52 gb * [r4503] : Clean up x8664 callbacks a little. Don't coerce single float results to double; bind stack-ptr variable even if no args (might be a return value form that references it.) Failure to do the latter was creating a weird free variable reference that wasn't handled by the fasloader. 2006-05-04 00:50 gb * [r4502] : $fasl-vmksym-special wasn't implemented. 2006-05-04 00:49 gb * [r4501] : Indent a comment. 2006-05-04 00:20 gb * [r4500] : *** empty log message *** 2006-05-04 00:16 gb * [r4499] : In an %FF-CALL that might involve a variable-length arglist (at this level, always), set %rax to the number (0-8) of FP arguments passed in xmm registers. 2006-05-03 20:20 gb * [r4498] : *** empty log message *** 2006-05-03 19:26 gb * [r4497] : In find_foreign_rsp(), mask with ~15, not ~!5. (This was causing segfaults on return from interrupts sometimes.) 2006-05-03 19:24 gb * [r4496] : CLOSURE-CLOSED-OVER-VALUES, which is needed by SWANK/SLIME. 2006-05-03 19:23 gb * [r4495] : FRAME-SUPPLIED-ARGS, hopefully in a form that'll work on all platforms. (Note that FRAME-SUPPLIED-ARGS seems to get many things wrong.) 2006-05-03 18:20 gb * [r4494] : *** empty log message *** 2006-05-03 17:37 gb * [r4493] : You mist have yesterday's version. 2006-05-03 17:37 gb * [r4492] : Implement #'%FF-CALL. Not too much LAP. 2006-05-03 17:36 gb * [r4491] : Put :X86-64 and :AMD64 on *FEATURES* if #+x8664-target. 2006-05-03 17:36 gb * [r4490] : Bad typo in UNBOX-S64, and it can just check for a 2-digit bignum header if it has something of tag TAG-MISC. 2006-05-03 17:34 gb * [r4489] : X862-COPY-REGISTER: handle :S64 destinations. 2006-05-03 17:33 gb * [r4488] : Recognize uuo-error-eep-unresolved. 2006-05-03 00:20 gb * [r4487] : *** empty log message *** 2006-05-03 00:07 gb * [r4486] : Bump. 2006-05-03 00:07 gb * [r4485] : Comment out stuff that's (mostly ?) unused and currently PPC-specific, if it works at all ... 2006-05-02 16:33 gb * [r4484] : *** empty log message *** 2006-05-02 16:27 gb * [r4483] : Don't definea STREAM-CLEAR-INPUT method for STRING-INPUT-STREAMs; it's not clear what such a method should do (and fairly clear that it shouldn't do what the previous definition was doing.) 2006-05-02 16:26 gb * [r4482] : clear input from *query-io*, not from *terminal-io* in YES-OR-NO-P, Y-OR-N-P 2006-05-02 16:26 gb * [r4481] : x8664 lap 2006-05-02 16:26 gb * [r4480] : hons area stuff 2006-04-29 15:20 gb * [r4479] : *** empty log message *** 2006-04-29 15:14 gb * [r4478] : Write-protect handlers for stack overflow (on value/temp stacks). 2006-04-29 15:13 gb * [r4477] : Handle stack overflow in cmain callback. 2006-04-29 15:10 gb * [r4476] : x86-gc.o. No linker script. Use -lthr, which offers 1:1 threads. 2006-04-29 15:09 gb * [r4475] : Remove the __sighandler typedef. 2006-04-29 15:01 gb * [r4474] : xpMMXreg for FreeBSD. 2006-04-29 15:00 gb * [r4473] : Don't use __sighandler_t. (See if anything cares ...) 2006-04-29 14:59 gb * [r4472] : FreeBSD register constants. 2006-04-29 14:58 gb * [r4471] : FreeBSD regiser names. 2006-04-29 14:58 gb * [r4470] : Update IMAGE_BASE_ADDRESS for FreeBSD. 2006-04-29 04:20 gb * [r4469] : *** empty log message *** 2006-04-29 03:28 gb * [r4468] : bump. 2006-04-29 03:27 gb * [r4467] : changes to print-call-history (and "internal" version). 2006-04-29 03:24 gb * [r4466] : Fix typo. 2006-04-29 03:23 gb * [r4465] : Add :NFRAMES break loop command. Change the command/keyword parser a little, so that :FOO is just :FOO and not (:FOO). Change :B break loop command (and :F) to allow focusing on a range of frames, and change PRINT-CALL-HISTORY to more easily support that. 2006-04-29 02:20 gb * [r4464] : *** empty log message *** 2006-04-29 01:37 gb * [r4463] : Remove a quote from a LAP LOAD-CONSTANT form. 2006-04-29 01:34 gb * [r4462] : Factor out a little more. Time to see what's broken on PPC ... 2006-04-29 01:34 gb * [r4461] : Make parens around colon-commands optional. 2006-04-28 10:20 gb * [r4460] : *** empty log message *** 2006-04-28 10:11 gb * [r4459] : 'lregs' calls the kernel debugger's register display code. 2006-04-28 10:11 gb * [r4458] : Don't claim that the (legitimate) last frame on the stack is bogus. 2006-04-28 10:10 gb * [r4457] : Create fake stack frames (sort of ...) when calling back for an exception. 2006-04-28 09:36 gb * [r4456] : A few symbolic register names. xcf struct. 2006-04-28 09:36 gb * [r4455] : FAKE-STACK-FRAME is #+ppc-target. 2006-04-28 09:36 gb * [r4454] : Implemented some stuff. Backtrace (sort of) works on x8664 now; need to find register values, hide tagged return addresses, a few other things. 2006-04-28 09:34 gb * [r4453] : Stuff moved here from elsewhere. 2006-04-28 09:34 gb * [r4452] : LAST-CATCH-SINCE moved elsewhere. 2006-04-28 09:33 gb * [r4451] : Flags register accessor. New exception callback scheme. 2006-04-28 09:33 gb * [r4450] : The lowlevel system keeps changing. 2006-04-28 09:32 gb * [r4449] : New exception callback stuff. 2006-04-28 09:32 gb * [r4448] : No fake stack frames on x86. 2006-04-28 09:32 gb * [r4447] : Move more stuff to arch-specific files. 2006-04-28 09:31 gb * [r4446] : No fake-stack-frames on x86, but we do have exception callback frames ... 2006-04-28 09:30 gb * [r4445] : Tweak :report method for wrong-number-of-args errors. 2006-04-28 09:30 gb * [r4444] : Implement APPLY+. 2006-04-28 09:29 gb * [r4443] : eXception Callback Frame. 2006-04-28 09:27 gb * [r4442] : Add missing IN-PACKAGE. CHECK-NARGS: don't enforce upper bound if it's null. 2006-04-28 09:20 gb * [r4441] : *** empty log message *** 2006-04-28 09:17 gb * [r4440] : x86 xFLAGS bits. 2006-04-28 09:16 gb * [r4439] : Allow (FUNCTION (SETF name)) in IGNORE/IGNORABLE/DYNAMIC-EXTENT declarations. 2006-04-26 20:20 gb * [r4438] : *** empty log message *** 2006-04-26 20:18 gb * [r4437] : Allow (SETF NAME) in DEFINE-COMPILER-MACRO. 2006-04-26 20:17 gb * [r4436] : Disassemble is supposed to signal an error (or "be prepared" to do so) if its argument isn't a function designator, so don't quietly call METHOD-FUNCTION for the user. 2006-04-26 19:56 gb * [r4435] : Use REPLACE-FUNCTION-CODE (not (SETF (UVREF ...))) in ENCAPSULATE, REMOVE-ENCAPSULATION. 2006-04-26 10:20 gb * [r4434] : *** empty log message *** 2006-04-26 09:59 gb * [r4433] : zero out the exception bits whenever mxcsr is stored in tcr.foreign_mxcsr, so no exceptions will be pending when we start an ff-call (and the mxcsr is loaded from that location.) tcr.ffi_exception will then contain non-zero exception bits iff the most recent ff-call raised a (probably masked) FP exception. TODO: make this reentrant (PROCESS-INTERRUPT should preserve this.) 2006-04-26 09:55 gb * [r4432] : define mxcsr exception bits. 2006-04-26 09:23 gb * [r4431] : Update comment on tcr.ffi_exception. 2006-04-26 09:22 gb * [r4430] : Maintain tcr.ffi_exception on return from ffcall. (Should maybe clear exception bits on entry to foreign state, so we're sure that any exception bits set in the tcr.ffi_exception slot were caused by the most recent ff-call.) 2006-04-26 09:20 gb * [r4429] : Check MXCSR status bits from last ffcall, compare them to unmasked exception bits in current MXCSR, maybe raise FP exception. (Note: compile this change and rebuild image before making corresponding kernel change.) 2006-04-26 09:20 gb * [r4428] : *** empty log message *** 2006-04-26 08:31 gb * [r4427] : .SPbuiltin-logxor was calling out to LOGAND-2, not to LOGXOR-2. Sheesh. 2006-04-26 08:21 gb * [r4426] : *** empty log message *** 2006-04-26 08:01 gb * [r4425] : %ILOGBITP treats "large" bit numbers as tests of the sign bit. (This is probably best viewed as being "formally undefined", but it's not too expensive to check for that. The case where the bit number's negative is definitely undefined; I think that it also gets treated as a test of the sign bit.) 2006-04-26 07:58 gb * [r4424] : Logbitp calls out if bit number negative. 2006-04-26 07:32 gb * [r4423] : If both args fixnums, handle bits >= 61. (Should check here for negative bit .) 2006-04-26 07:31 gb * [r4422] : If we get the "no FP regs in context" error - which typically indicates a fault during exception processing - pass the context in to the kernel debugger. 2006-04-26 07:30 gb * [r4421] : Change EQL doc string. 2006-04-26 07:20 gb * [r4420] : *** empty log message *** 2006-04-26 06:56 gb * [r4419] : EQL on two bignums works much better if it compares all corresponding digits, rather than repeatedly comparing the 0th digits N times ... 2006-04-26 00:20 gb * [r4418] : *** empty log message *** 2006-04-25 23:38 gb * [r4417] : _SPmvslide: skip the loop if nargs is 0, not if disp is 0. ash: always make a signed integer out of the result, since arg is always signed. 2006-04-25 23:37 gb * [r4416] : Negate floats by fliping the sign bit, so (for instance) (- 0.0d0) is -0.0d0. 2006-04-25 23:36 gb * [r4415] : Hash bignums 32 bits (not 64!) at a time. 2006-04-25 23:35 gb * [r4414] : %%INEG: need to ensure that source operand is in target reg (2-address issue.) 2006-04-24 23:20 gb * [r4413] : *** empty log message *** 2006-04-24 22:42 gb * [r4412] : How hard can misc_set_s32 be ? Apparently, pretty hard. It was doing a logical shift right when typechecking the value, so negative values were rejected. 2006-04-24 22:41 gb * [r4411] : Call out to handle SIGFPE. cmain callback now takes a 'signo' argument. 2006-04-24 22:40 gb * [r4410] : Initial support for handling arithmetic errors. Will need more work to get operation and operands right. 2006-04-24 22:39 gb * [r4409] : REBUILD-CCL: :reload arg defaults to T. 2006-04-24 20:20 gb * [r4408] : *** empty log message *** 2006-04-24 19:58 gb * [r4407] : SETQ (not #'SET) in the uuo-error-reg-not-type case. 2006-04-24 19:57 gb * [r4406] : Generally want to use FPZERO, not FP0, for 0.0[d,f]0. (One case wanted to use fp1, but that's another story.) 2006-04-24 19:55 gb * [r4405] : REQUIRE-REAL, REQUIRE-NUMBER: mask wont fit in 32 bits. OTOH, doesn't need to be boxed, either. 2006-04-24 18:20 gb * [r4404] : *** empty log message *** 2006-04-24 18:02 gb * [r4403] : misc_set_s32: register botch. _SPtcallnfngen: several bugs, which were already fixed in similar variants. 2006-04-24 18:01 gb * [r4402] : Use FORCE_DWS_MARK to force use of the Deutch/Waite/Schorr marker. 2006-04-24 18:00 gb * [r4401] : Only turn on GC debug bit if GC_INTEGRITY_CHECKING is defined (e.g., in Makefile.) 2006-04-24 17:59 gb * [r4400] : Remove debugging FORMAT call. 2006-04-24 17:58 gb * [r4399] : %ROUND-NEAREST-SHORT-FLOAT->FIXNUM: argument is a SINGLE/SHORT-FLOAT, so don't access as a DOUBLE-FLOAT. 2006-04-24 17:58 gb * [r4398] : %MAKE-FLOAT-FROM-FIXNUMS: mask the high and low words of the mantissa, to ensure that no hidden bits are set. (Some such bits were set, causing many double-float constansts - like MOST-POSITVE-DOUBLE-FLOAT - to be bogus NaNs). 2006-04-24 09:20 gb * [r4397] : *** empty log message *** 2006-04-24 08:21 gb * [r4396] : *** empty log message *** 2006-04-24 08:21 gb * [r4395] : Expect a MANAGED_STATIC section in the image. (Don't expect it to have any contents yet; that'll require several more changes. This is all intended to support PURIFY on x8664, where we'll want to ensure that any constant referenced by a function in readonly/pure memory has a fixed, static address.) 2006-04-24 08:14 gb * [r4394] : Bump IMAGE_ABI_VERSION. New managed_static section makes NUM_IMAGE_SECTIONS 4. 2006-04-24 08:13 gb * [r4393] : Typo in the definition of AREA_MANAGED_STATIC. 2006-04-24 08:12 gb * [r4392] : Global pointers to more areas. 2006-04-24 08:12 gb * [r4391] : EGC write-barrier labels are the addresses of instructions (not pointers thereto ...) 2006-04-24 08:11 gb * [r4390] : Add conditionalization to force DWS marker (mostly so that I could step through it to remember how it worked ...). 2006-04-24 08:10 gb * [r4389] : write-barrier exported symbols: remove indirection. 2006-04-24 08:09 gb * [r4388] : Remove a sanity check. 2006-04-24 08:09 gb * [r4387] : Create (empty) managed-static section (mostly for x8664 PURIFY, but might be useful on PPC someday.) 2006-04-24 08:08 gb * [r4386] : New image-abi-version, for managed-static section. 2006-04-24 08:07 gb * [r4385] : GET-DESCRIPTOR-FOR: try not to leak fds, take CLOSE-ON-ERROR more seriously. 2006-04-24 08:06 gb * [r4384] : (eq %FOREIGN-STACK-POINTER %CURRENT-FRAME-PTR) on PPC32/64. 2006-04-24 07:20 gb * [r4383] : *** empty log message *** 2006-04-24 06:50 gb * [r4382] : Bump. Getting there ... 2006-04-24 06:50 gb * [r4381] : Use %GET-NATURAL where possible. Note that TCRs aren't guaranteed to be in the low 32 bits on x8664, so pass them as a pointer to avoid word-size issues. 2006-04-24 06:49 gb * [r4380] : Implement things that had been missing. 2006-04-24 06:48 gb * [r4379] : Recognize recently-defined UUOs. 2006-04-24 06:48 gb * [r4378] : Fix FIXUP-OPCODE-ORDINALS (again). 2006-04-24 06:47 gb * [r4377] : UUO-ERROR-ARRAY-BOUNDS. 2006-04-24 06:47 gb * [r4376] : A little word-size cleanup. 2006-04-24 06:47 gb * [r4375] : %FOREIGN-STACK-POINTER. 2006-04-24 06:46 gb * [r4374] : WITH-C-FRAME, WITH-VARIABLE-C-FRAME: use (new) %FOREIGN-STACK-POINTER, not %CURRENT-FRAME-PTR (differ on x8664; need to boostrap on PPC). Parse %FOREIGN-STACK-POINTER. 2006-04-22 10:21 gb * [r4373] : *** empty log message *** 2006-04-22 09:30 gb * [r4372] : If we're using thread-local-storage (HAVE_TLS is defined), make shutdow_thread_tcr() dequeue the tcr, because (a) nothing else will and (b) the tcr will likely become unmapped soon, and we don't want things that walk the tcr list to segfault. 2006-04-22 09:27 gb * [r4371] : Something's clearly wrong in find_foreign_rsp; call Bug() and exit() if the context has no fpregs. (That's a symptom, not the proble, but it's slightly better to do this than to double fault, which is what seems to be happening.) 2006-04-21 16:20 gb * [r4370] : *** empty log message *** 2006-04-21 16:12 gb * [r4369] : implement pc_luser_xp; just has to worry about consing, write-barrier stuff. 2006-04-21 16:11 gb * [r4368] : Define X86 flags register bits (condition codes.) 2006-04-21 16:11 gb * [r4367] : change the signature of pc_luser_xp (again.) Define enumeration constants for matching the instructions that can occur during consing. 2006-04-21 16:09 gb * [r4366] : Preserve %temp0 in store_node_conditional, in case pc_luser_xp() needs to complete the memoization. 2006-04-21 09:20 gb * [r4365] : *** empty log message *** 2006-04-21 08:50 gb * [r4364] : Use only the value of the first instance of :ALLOW-OTHER-KEYS to determine whether unknown keys are allowed. 2006-04-21 08:49 gb * [r4363] : Tweak things so that %rip and %flags line up with GPRs. 2006-04-21 08:48 gb * [r4362] : SIMPLE-1D-ARRAY-SUBSEQ: 64-bit vector case is #+64-bit-target, not #+ppc64-target. 2006-04-21 08:47 gb * [r4361] : %SIMPLE-BIT-BOOLE: shift by the right amount. 2006-04-21 08:46 gb * [r4360] : %UVECTOR-REPLACE: 64-bit vectors aren't unique to PPC64 (but do depend on #+64-bit-target). 2006-04-21 08:46 gb * [r4359] : The binding(s) of *INTERRUPT-LEVEL* confuse PROGV; remove them, for now. 2006-04-21 06:20 gb * [r4358] : *** empty log message *** 2006-04-21 05:35 gb * [r4357] : In the bignum case of .SPinteger-sign, sign-extend the bignum's sign digit instead of ... whatever the code was trying to do there. In the lexpr-has-two-values case of .SPspread-lexprz, load arg_y and arg_z from the correct offsets within the lexpr. 2006-04-20 11:20 gb * [r4356] : *** empty log message *** 2006-04-20 10:48 gb * [r4355] : Get the link-inverting marker written; force it to always be used, for now. 2006-04-19 23:20 gb * [r4354] : *** empty log message *** 2006-04-19 23:11 gb * [r4353] : (Very) different MAXIMUM_MAPPABLE_MEMORY for Solaris. 2006-04-19 23:10 gb * [r4352] : (Very) different IMAGE_BASE_ADDRESS for Solaris. 2006-04-19 07:20 gb * [r4351] : *** empty log message *** 2006-04-19 06:35 gb * [r4350] : Write a few paragraphs outlining how the reversing-pointer marker should work. (Perhaps it's a strong argument for implementing "purify" soon.) 2006-04-19 06:34 gb * [r4349] : function-boundary-marker. 2006-04-19 06:33 gb * [r4348] : Add node_aligned and indirect_node macros. 2006-04-19 02:20 gb * [r4347] : *** empty log message *** 2006-04-19 01:35 gb * [r4346] : Comparison was backwards in multiple-value case of THROW: don't want to do unbind_to if tcr's db-link matches target catch frame's. 2006-04-19 01:33 gb * [r4345] : Add UUO-ERROR-DEBUG-TRAP, remove UUO-ERROR-STACK-OVERFLOW. Still need to get this completely in synch with x86-asm. 2006-04-19 01:33 gb * [r4344] : Say more about the availability of :R in break messages. Say more about :C in :R. 2006-04-19 01:32 gb * [r4343] : Remove the #+x86-target debugging code in QUIT's PROCESS-INTERRUPT call, since the closure- trashing bug seems to have been fixed. 2006-04-19 01:31 gb * [r4342] : Recognize (uuo-error-not-tag %rn x8664::subtag-catch-tag) as CANT-THROW-ERROR. (Yes, this is a horrible kludge, but it only happens once.) 2006-04-19 01:20 gb * [r4341] : *** empty log message *** 2006-04-19 01:19 gb * [r4340] : Remove the #+x86-target debugging code in PROCESS-EXIT-APPLICATION, since the closure- trashing bug seems to have been fixed. 2006-04-19 01:18 gb * [r4339] : EQL on bignums wasn't using X8664::MISC-DATA-OFFSET, so it was comparing a few bytes beyond the ends of its arguments and ignoring the first few bytes (as well as doing misaligned memory references.) 2006-04-19 01:16 gb * [r4338] : Insert the PPC version of APPLY+ here. This'll require some thought, mostly because of the reserved-frame issue(s). 2006-04-18 13:20 gb * [r4337] : *** empty log message *** 2006-04-18 12:21 gb * [r4336] : Comment mishaps. These wacky adventures are about to end, I'm afraid. 2006-04-18 12:20 gb * [r4335] : *** empty log message *** 2006-04-18 12:04 gb * [r4334] : Daily bump. 2006-04-18 11:59 gb * [r4333] : Comment indentation. 2006-04-18 11:56 gb * [r4332] : Move APPLY+ here. 2006-04-18 11:55 gb * [r4331] : Move APPLY+ (which is defined in LAP) to ppc-def (that means that there will need to be an APPLY+ for x8664). 2006-04-18 11:38 gb * [r4330] : Try to get the compiler and x8664-vinsn-templates to agree on vinsn-names and implement some vinsns that had been missing. (About 15 or so undefined vinsns remain.) 2006-04-18 11:37 gb * [r4329] : The mechanism that tries to fixup opcode ordinals in vinsn templates (e.g., in case new instructions are defined in the assembler) wasn't checking to see if vinsn templates named in the hash table were defined. (It was apparently getting away with this because CONSP was broken.) So, don't try to fixup opcode ordinals in NIL, and optionally report templates that're referenced but not defined. 2006-04-18 11:20 gb * [r4328] : *** empty log message *** 2006-04-18 10:38 gb * [r4327] : CONSP was using the EXTRACT-FULLTAG vinsn (which only sets the low 8 bits of its destination), then comparing all 64-bits of the destination to a zero-extended 32-bit constant. (E.g., CONSP had little chance of being right all other things being equal, but may have been right more than one in 2^56 times if some random preceding instruction had zeroed the upper 56 bits.) (It's also shorter to do an 8-bit compare.) 2006-04-18 09:20 gb * [r4326] : *** empty log message *** 2006-04-18 09:11 gb * [r4325] : box-fixnum arg order in %DFLOAT-HASH. 2006-04-18 08:51 gb * [r4324] : Now that all assemblers are happy (well, as happy as they're going to be ...), back to finding bugs. MULTIPLY-FIXNUMS has to unbox one arg (the source). 2006-04-18 08:21 gb * [r4323] : *** empty log message *** 2006-04-18 07:26 gb * [r4322] : Some Solaris changes. 2006-04-18 07:20 gb * [r4321] : *** empty log message *** 2006-04-18 06:48 gb * [r4320] : The version of gm4 that Sun installs in /usr/sfw/bin is both newer and buggier than the version available on the "Solaris Companion" disk, which gets installed in /opt/sfw/bin. Ordinarily, I'd wonder why a "companion" disk offers to install software that's bundled as part of the installation (especially old versions of that software.) Now I know. 2006-04-18 06:45 gb * [r4319] : # __line__ wasn't a comment. 2006-04-18 06:45 gb * [r4318] : Fix changecom. 2006-04-18 06:20 gb * [r4317] : *** empty log message *** 2006-04-18 06:11 gb * [r4316] : x8664 linux as doesn't like typos, but is otherwise fine with single-line /* */. 2006-04-18 06:09 gb * [r4315] : Whoops. Darwin as doesn't like #. I dont like Darwin as, but we're back to single-line /* */, which is often pretty ugly and not well-supported by Emacs. 2006-04-18 05:23 gb * [r4314] : Use #/newline for comment delimiters, which is supported by Emacs and by all current GAS versions (but see ARM conventions.) No more jerking around with this, knock wood. 2006-04-18 05:23 gb * [r4313] : _SPspread_lexprz wasn't handling the case where caller provided args before the lexpr, which is much harder than the simple case. Use #/newline for comment delimiters, which is supported by Emacs and by all current GAS versions (but see ARM conventions.) No more jerking around with this, knock wood. 2006-04-18 05:20 gb * [r4312] : %FIXNUM-GCD was horrible, espcially when numbers differed greatly in magnitude. Current (binary) algorithm is usually OK. 2006-04-18 05:20 gb * [r4311] : *** empty log message *** 2006-04-18 05:17 gb * [r4310] : Condition name should be UNDEFINED-X86-LAP-LABEL, since it's (a) referenced and (b) that's not the problem. Name wasn't spelled correctly in code that signals the condition, anyway. 2006-04-17 23:20 gb * [r4309] : *** empty log message *** 2006-04-17 22:34 gb * [r4308] : pc_luser_xp always takes a tcr argument, takes a new boolean to distinguish current tcr from others. (The current tcr is only lusered on an interrupt; others are only lusered on GC entry.) 2006-04-17 22:33 gb * [r4307] : X86 prototype for pc_luser_xp moved here. (No deep reason not to share the new x86 signature on all architectures.) 2006-04-17 22:32 gb * [r4306] : PPC prototype for pc_luser_xp moved here. (No deep reason not to share the new x86 signature on all architectures.) 2006-04-17 22:31 gb * [r4305] : prototype for pc_luser_xp is platform-dependent. 2006-04-17 13:20 gb * [r4304] : *** empty log message *** 2006-04-17 12:49 gb * [r4303] : In the closure created in QUIT, only do CCL::DBG if (a) we're on X86 and (b) the inner closure is bad already. 2006-04-17 12:20 gb * [r4302] : *** empty log message *** 2006-04-17 12:17 gb * [r4301] : Don't know if there's such a thing as truly portable type name. 2006-04-17 12:16 gb * [r4300] : Well, // style comments seem to only work if they're at the start of a line in Linux GAS. Try /* */ comments, but limit them to a single line. (May ultimately be better to use m4 dnl - carefully - and avoid end-of-line comments.) 2006-04-17 11:28 gb * [r4299] : // comments. 2006-04-17 11:28 gb * [r4298] : timeradd, timersub. Other Solaris conditionalization. 2006-04-17 11:27 gb * [r4297] : Solaris-related changes (xp accessors, etc.) 2006-04-17 11:26 gb * [r4296] : Solaris-related changes (mostly in the sigstack trampoline stuff.) 2006-04-17 11:25 gb * [r4295] : // comments, Solaris-friendly integer types. 2006-04-17 11:23 gb * [r4294] : Some Solaris conditionalization. Solaris mmap() is ... unusual. 2006-04-17 11:21 gb * [r4293] : Use C++-style single-line comments, since Solaris version of GAS doen't handle multi-line comments (/* ... */). It -seems- that other assemblers do handle them correctly; we'll see. Some Solaris conditionalization (ELFstabs, etc.) 2006-04-17 11:20 gb * [r4292] : *** empty log message *** 2006-04-17 11:20 gb * [r4291] : Conditionalize for Solaris (type names). We'll see what this breaks. 2006-04-17 11:19 gb * [r4290] : Use C++-style single-line comments, since Solaris version of GAS doen't handle multi-line comments (/* ... */). It -seems- that other assemblers do handle them correctly; we'll see. 2006-04-17 11:19 gb * [r4289] : PLATFORM_OS is PLATFORM_OS_SOLARIS when SOLARIS is defined. 2006-04-17 11:18 gb * [r4288] : Solaris conditionalization; print registers in predictable order. 2006-04-17 11:15 gb * [r4287] : IMAGE_BASE_ADDRESS for SolarisX64. Wrong. 2006-04-17 11:15 gb * [r4286] : Conditionalize for Solaris. TODO: drop Jaguar support, use POSIX (vice Mach) semaphores on all platforms ? 2006-04-17 11:01 gb * [r4285] : new file 2006-04-17 10:56 gb * [r4284] : Make it a little easier to change the name/usage of the linker script. Stop using the linker script (since spjump is the only thing in the kernel that has to be at a fixed address and we can copy it there.) Don't disable the EGC. 2006-04-17 10:54 gb * [r4283] : Define spjump_start as the origin of the spjump table, exported for C (so that spjump can be mapped into low memory.) 2006-04-17 10:54 gb * [r4282] : Use reserved_frame_marker. If the tlb's too small when trying to bind a special, push the binding index of the symbol we're binding (for the benefit of the handler.) 2006-04-17 10:52 gb * [r4281] : Check for cases where a function has more code words than it has elements (can't happen, unless the function's code-words word is garbage, which seems to happen when closures get consed up in non-zeroed memory.) Make sure that we mark and forward Itemp2 (= %rcx = %temp2 = %nargs.q), which sometimes contains something tagged and non-immediate. 2006-04-17 10:48 gb * [r4280] : new_heap_segment() (slowly) checks to ensure that everythings's zeroed. x8664 implementation of extend_tcr_tlb(). on a gc trap, use tcr->save_allocptr (not the random value in %rbx) in call to update_bytes_allocated(). normalize_tcr() - alway set tcr->save_allocptr/tcr->save_allocbase to VOID_ALLOCPTR (I -think- that this explains the crash-on-quit bug, and possibly some other instabiliy.) handle_exception: keep trying to guess how to distinguish memory faults from other sources of SIGSEGV. On debug traps, advance the PC past the UUO before entering the debugger. Handle XUUO_TLB_TOO_SMALL. 2006-04-17 10:42 gb * [r4279] : reserved_frame_marker. 2006-04-17 10:42 gb * [r4278] : Remove implementation of extend_tcr_tlb(). 2006-04-17 10:41 gb * [r4277] : Prototype for PPC version of extend_tcr_tlb(). 2006-04-17 10:41 gb * [r4276] : PPC version of extend_tcr_tlb(). 2006-04-17 10:40 gb * [r4275] : Remap the spjump table on x8664, so that we don't need to use a linker script. 2006-04-17 10:40 gb * [r4274] : extend-tcr-tlb: signature defined by target xxx-exceptions.h now. 2006-04-17 10:39 gb * [r4273] : Some comment formatting/editing. 2006-04-17 10:38 gb * [r4272] : Bump. 2006-04-17 10:38 gb * [r4271] : Some debugging code in the code which creates the final closure. It seems that when it's bad, it's bad on creation (here.) 2006-04-17 10:37 gb * [r4270] : Some debugging code in PROCESS-EXIT-APPLICATION, to try to catch whatever's causing the final closure to be trashed. I don't think that other threads were getting their allocptr's nullified correctly by the GC/normalize_tcr(); with the current debugging code in place, it seems like the closure is consed on top of something other than free, thread-private heap memory (when it's bad, it fails the first check here. I'm not sure that the latest round of GC/normalize-tcr() changes have fixed the problem, but I haven't seen it in a while now, either.) 2006-04-17 10:32 gb * [r4269] : %FUNCTION-CODE-BYTE (for partial disassembly, etc.). 2006-04-17 10:31 gb * [r4268] : Use RESERVED-FRAME-MARKER in RESERVE-OUTGOING-FRAME; don't use ENSURE-RESERVED-FRAME at all (anymore.) Subprims are down in low memory again; make INIT-CLOSURE skip setting the high byte of .SPcall-closure's address if it's zero. 2006-04-17 10:28 gb * [r4267] : Define RESERVED-FRAME-MARKER, which may help backtrace. 2006-04-17 10:27 gb * [r4266] : Split up RAW-FRAME-REF (interface/implementation.) 2006-04-15 16:20 gb * [r4265] : *** empty log message *** 2006-04-15 15:20 gb * [r4264] : Ensure that callback trampolines are allocated in executable memory. 2006-04-15 15:20 gb * [r4263] : *** empty log message *** 2006-04-15 15:12 gb * [r4262] : Refresh external entry points before reviving callbacks. Reset callback storage before reviving callbacks. 2006-04-15 15:10 gb * [r4261] : Don't prefix the unknown trap error message with a %. (Where did that come from ?) 2006-04-15 15:08 gb * [r4260] : Don't (knock wood) need to do anything in RESET-CALLBACK-STORAGE for PPC, since malloc()ed memory is executable. 2006-04-14 15:20 gb * [r4259] : *** empty log message *** 2006-04-14 15:02 gb * [r4258] : Jeez. Save buffer, why dontcha. 2006-04-14 14:21 gb * [r4257] : ENCODE-UNIVERSAL-TIME: treat year 0-99 as shorthand for the nearest matching year. 2006-04-14 14:20 gb * [r4256] : FORMAT-PRINT-NUMBER was printing commas even when : was missing. 2006-04-14 05:20 gb * [r4255] : *** empty log message *** 2006-04-14 05:09 gb * [r4254] : Eschew missing parens. 2006-04-14 05:08 gb * [r4253] : Change NTH-VALUE-IN-FRAME call. 2006-04-14 05:08 gb * [r4252] : DBG-FORM checks the second return value from ARGLIST-FROM-MAP. 2006-04-14 05:02 gb * [r4251] : Don't pass a "conser" function around (the idea is leftover from 68K MCL). ARGLIST-FROM-MAP tries to be consistent about returning a second value; returns values NIL, T for functions that take no args and don't have a symbol map. 2006-04-14 04:53 gb * [r4250] : Try to make args to NTH-VALUE-IN-FRAME et al a little saner. 2006-04-14 04:37 gb * [r4249] : Conditionalize on word size, not PPC processor type. 2006-04-14 04:36 gb * [r4248] : Bump. 2006-04-14 04:35 gb * [r4247] : Remove the DBG trap from #'APPLY. Still trying to find the cause of the crash-on-quit bug. 2006-04-13 14:20 gb * [r4246] : *** empty log message *** 2006-04-13 13:30 gb * [r4245] : Migrate some stuff here. 2006-04-13 13:30 gb * [r4244] : Migrate some stuff to target-backtrace. 2006-04-13 13:20 gb * [r4243] : *** empty log message *** 2006-04-13 12:45 gb * [r4242] : mark_root: need to handle TRAs earlier. 2006-04-13 12:44 gb * [r4241] : Set GC debug bits manually here on x8664. 2006-04-13 12:44 gb * [r4240] : XP-ARGUMENT-LIST: only the low 13 bytes of %nargs_q contain arg count. 2006-04-13 12:43 gb * [r4239] : No EVAL-PROGRAM-ERROR. Please. 2006-04-13 12:43 gb * [r4238] : Trying to find what's (sometimes) trashing a closure during QUIT; embed a #+x86-target breakpoint in APPLY. 2006-04-13 12:42 gb * [r4237] : Bad typo in %FIXNUM-GCD. 2006-04-13 12:41 gb * [r4236] : Lap expression changes; recycle constant operands. 2006-04-13 12:41 gb * [r4235] : Freelist a lot of things; use "vector-lists" instead of adjustable arrays, simplify handling of constant expressions. Still conses hysterically; need to find out where and tone it down a bit, if possible. 2006-04-13 12:39 gb * [r4234] : Maybe no lap immediate operands; maybe no lap constant-expressions. 2006-04-13 12:39 gb * [r4233] : Don't need *lap-constant-0-expression*. 2006-04-13 07:20 gb * [r4232] : *** empty log message *** 2006-04-13 06:46 gb * [r4231] : Kernel depends on Makefile, linker script. 2006-04-13 06:45 gb * [r4230] : Use linker script from ld 2.16.1 (Thanks to James Bielman). 2006-04-13 01:20 gb * [r4229] : *** empty log message *** 2006-04-13 00:53 gb * [r4228] : Try to avoid use of PROVIDE_HIDDEN, since some versions of GNU ld don't like it. (Most of this seems to have to do with C++ static constructors, which aren't really relevant here.) 2006-04-13 00:28 gb * [r4227] : Use movd (not movq) when moving between %Rforeign_sp and GPRs. 2006-04-13 00:22 gb * [r4226] : egc flag 2006-04-13 00:20 gb * [r4225] : *** empty log message *** 2006-04-13 00:08 gb * [r4224] : Add current/prev (temporarily) to help find where we're losing here. 2006-04-13 00:07 gb * [r4223] : pc_luser_xp: notice if we're in the write barrier code, crap out if so. 2006-04-13 00:06 gb * [r4222] : Disable EGC if DISABLE_EGC is defined (presumably in a Makefile). 2006-04-12 21:20 gb * [r4221] : *** empty log message *** 2006-04-12 20:50 gb * [r4220] : Bump. 2006-04-12 20:46 gb * [r4219] : new target backtrace stuff. 2006-04-12 20:45 gb * [r4218] : Start splitting things up, ideally into a mostly abstract API and gory platform- dependent implementation details. 2006-04-12 04:20 gb * [r4217] : *** empty log message *** 2006-04-12 03:40 gb * [r4216] : Only way to stress-test EGC is to turn the damned thing on. (A little painful with integrity checking enabled, as it still is by default.) TODO: - sleep (me) - non-recursive (link-inverting) marker - pc-lusering (pc_luser_xp()) - backtrace - ANSI CL tests - purify - 2d-aref - FP exception handling - check UUO usage in kernel, disassembler, exception handlers - stack overflow recovery 2006-04-12 03:24 gb * [r4215] : Skip over function code in check_refmap_consistency. 2006-04-12 03:23 gb * [r4214] : Write barrier memoization sets bits backwards. 2006-04-12 01:20 gb * [r4213] : *** empty log message *** 2006-04-12 01:11 gb * [r4212] : Change LABELS for new closure format. 2006-04-12 01:11 gb * [r4211] : 3 words of code + inner function in a closure, no lfname slot in closure. 2006-04-11 23:20 gb * [r4210] : *** empty log message *** 2006-04-11 22:29 gb * [r4209] : ada prints active dynamic (GC) area. 2006-04-11 22:29 gb * [r4208] : Maintain the MXCSR on transitions between lisp/foreign code. 2006-04-11 22:29 gb * [r4207] : Add write barrier stuff (may have bit-order problems.) Maintain the MXCSR on transitions between lisp/foreign code. 2006-04-11 22:28 gb * [r4206] : Access globals in expressions. 2006-04-11 22:27 gb * [r4205] : Start to implement EGC for X8664; this'll need some bootstrapping stuff (function boundary markers) for a few reasons, as will DWS marker. 2006-04-11 22:27 gb * [r4204] : If we get an FP exception, report the siginfo error code. 2006-04-11 22:26 gb * [r4203] : Define MXCSR fields in tcr. 2006-04-11 22:25 gb * [r4202] : Define MXCSR bits. 2006-04-11 22:25 gb * [r4201] : Setup the MXCSR somewhat sanely when starting a lisp thread. (It'll be saner when FP exceptions are handled.) 2006-04-11 22:24 gb * [r4200] : Recognize what complaints about non-uvectors are trying to tell us. 2006-04-11 22:23 gb * [r4199] : PRINT-OBJECT method for TRAs, symbol/function vectors. 2006-04-11 22:23 gb * [r4198] : Spell TAGGED-RETURN-ADDRESS with more Ses. Define classes for function/symbol vectors. 2006-04-11 22:22 gb * [r4197] : %RETURN-ADDRESS-OFFSET. 2006-04-11 22:21 gb * [r4196] : Closure initialization needs to add a function boundary marker (but we can stop padding each closure with a null function name slot.) 2006-04-11 22:20 gb * [r4195] : *** empty log message *** 2006-04-11 22:20 gb * [r4194] : Define FUNCTION-BOUNDARY-MARKER, which should never exist in the wild. 2006-04-11 22:19 gb * [r4193] : Emit a function boundary marker between code and constants. (For GC reasons.) Change closure initialization to account for it. 2006-04-11 22:18 gb * [r4192] : Emit a function boundary marker between code and constants. (For GC reasons.) 2006-04-11 22:17 gb * [r4191] : Require more stuff. 2006-04-11 22:17 gb * [r4190] : Require X8664ENV on X8664. See if we can get by without 68K condition code ... 2006-04-11 13:20 gb * [r4189] : *** empty log message *** 2006-04-11 12:53 gb * [r4188] : We'll need to handle FP exceptions soon. 2006-04-11 12:53 gb * [r4187] : Infinity constants conditionalized on/off during MXCSR adventures. On now. 2006-04-11 12:52 gb * [r4186] : Comment indentation. No such thing as an interpreted function; hasn't been for years. 2006-04-11 12:51 gb * [r4185] : Fixup lfun bits of %FIXNUM-SET et al. 2006-04-11 12:51 gb * [r4184] : Bumped yesterday. 2006-04-11 12:50 gb * [r4183] : Comments. 2006-04-11 12:50 gb * [r4182] : CLASS-OF handles non-simple vectors, at least. 2006-04-11 12:50 gb * [r4181] : MXCSR stuff. Better, mostly because a fixnum constant is boxed in %SET-MXCSR. 2006-04-11 12:49 gb * [r4180] : Get info about register usage. 2006-04-11 12:49 gb * [r4179] : Natural arithmetic; start work on 2d aref. Fix MISC-SET-U64 (or some flavor thereof), which though that it was dealing with a scaled index. 2006-04-11 12:48 gb * [r4178] : Natural arithmetic; start work on 2d aref. 2006-04-11 12:48 gb * [r4177] : FRAGs are freelisted. Really need to do something about FRAG-CODE-BUFFERs, and need to tone down expression overkill. 2006-04-11 12:46 gb * [r4176] : Vinsn template fixup mechanism, for when opcode ordinals change. 2006-04-11 12:46 gb * [r4175] : UUO-ERROR-STACK-OVERFLOW -> UUO-ERROR-ARRAY-RANK. Vinsn template fixup mechanism, for when opcode ordinals change. 2006-04-11 12:45 gb * [r4174] : Try to unscramble MXCSR bits/masks. 2006-04-11 12:45 gb * [r4173] : NX-U31-CONSTANT-P. 2006-04-11 12:44 gb * [r4172] : WITH-DLL-NODE-FREELIST: syntax to allow specialized list header initialization. 2006-04-11 12:44 gb * [r4171] : Type errors for 2d float arrays. 2006-04-10 16:20 gb * [r4170] : *** empty log message *** 2006-04-10 16:19 gb * [r4169] : _SPcallback is now called with the foreign SP properly obeying C alignment conventions. In order to get it aligned for lisp usage during the callback, an extra word needs to be pushed before the backpointer to the old foreign RSP. 2006-04-10 16:18 gb * [r4168] : Ensure that the copied sigcontext is 16-byte aligned, so that the foreign stack pointer is aligned according to C conventions when we enter the signal handler on that stack. 2006-04-10 16:13 gb * [r4167] : Define FP infinities, since we're running with FP exceptions masked. (We don't handle FP exceptions on X8664 yet, but this was crashing due to ldmxcsr/stmxcsr confusion, not because an unmasked FP exception was raised and unhandled.) 2006-04-10 16:10 gb * [r4166] : Remove the WARN and the conditionalization around bit/bitfield accessors on X8664. 2006-04-10 16:09 gb * [r4165] : Repeat after me: to load the mxcsr from memory, we say "ldmxcsr". We want to do such a load in order to write new values into the mxcsr. To store it in memory, we say "stmxcsr". We want to do such a store (and a subsequent load from memory) in order to access the value of the mxcsr. Simple, no ? 2006-04-10 16:07 gb * [r4164] : %GET-BIT/%SET-BIT. 2006-04-10 15:20 gb * [r4163] : *** empty log message *** 2006-04-10 14:29 gb * [r4162] : Float addition/multiplication: either arg can overlap result (and often will). Float subtraction/division: caller should ensure that result and second arg are disjoint. 2006-04-10 14:27 gb * [r4161] : Float arithmetic is two-address; ensure that result doesn't overlap second argument for non-commutative operations. 2006-04-10 14:26 gb * [r4160] : Show FPRs on LinuxX8664. 2006-04-10 11:20 gb * [r4159] : *** empty log message *** 2006-04-10 11:12 gb * [r4158] : Don't free dead TCRs if we didn't malloc them. 2006-04-10 11:12 gb * [r4157] : REBUILD-CCL paramaters for :LINUXX8664. 2006-04-10 11:11 gb * [r4156] : No need to reference PPC package, and good reason not to. 2006-04-10 10:39 gb * [r4155] : signed 32-bit, 64-bit integers in fasl files are always big-endian. 2006-04-10 10:38 gb * [r4154] : mark_root(): if we're marking a function reachable via a TRA, update the variable "dnode" so that the correct bits get set. 2006-04-10 10:37 gb * [r4153] : Remove some breakpoints. 2006-04-10 10:37 gb * [r4152] : %SIMPLE-BIT-BOOLE. Could stand to be a bit simpler. 2006-04-10 10:36 gb * [r4151] : Whoops: %SET-NEEDS-REHASHING is called from the MAKE-LOAD-FORM method for hashtables. 2006-04-09 15:20 gb * [r4150] : *** empty log message *** 2006-04-09 15:17 gb * [r4149] : Inflate control stack size on x8664, until DWS marker works (and do so correctly this time). 2006-04-09 15:12 gb * [r4148] : Inflate control stack size on x8664, until DWS marker works. 2006-04-09 15:09 gb * [r4147] : Now that bitvectors work, remove #+x86-target DBG trap. 2006-04-09 14:57 gb * [r4146] : Set/clear rehash bits via SBIT, not AREF. (Should be exactly the same code, but may not have been 15 years ago or whenever this was written.) 2006-04-09 14:56 gb * [r4145] : Clear bits in a bit vector via btrX (Bit Test and Reset), not btcX (Bit Test and Complement.) 2006-04-09 10:20 gb * [r4144] : *** empty log message *** 2006-04-09 09:52 gb * [r4143] : Parens around negative displacement in condition_to_boolean. Not clear why this is necessary, but it seems to be. 2006-04-09 09:24 gb * [r4142] : Use shorter sequence in condition_to_boolean. 2006-04-09 09:20 gb * [r4141] : *** empty log message *** 2006-04-09 08:32 gb * [r4140] : Using cmovCC intstead of setCC for cr-bit->boolean is 1 instruction and 2 bytes shorter, all other things being equal. 2006-04-09 08:21 gb * [r4139] : *** empty log message *** 2006-04-09 07:37 gb * [r4138] : reg/reg variants of bsfX, bsrX had operands backwards. 2006-04-09 07:37 gb * [r4137] : Comment formatting. 2006-04-09 07:29 gb * [r4136] : %BIGNUM-SIGN-DIGITS: bsr/xorl != clz when (zerop n). 2006-04-09 04:20 gb * [r4135] : *** empty log message *** 2006-04-09 03:25 gb * [r4134] : Forget about blocking SIGINT in current thread before entering kernel debugger. 2006-04-09 03:21 gb * [r4133] : early def of XLOAD-LEVEL-0: require (TARGET-XLOAD-MODULES). 2006-04-09 03:21 gb * [r4132] : misc_set_s16: forgot a lognot. 2006-04-09 03:20 gb * [r4131] : *** empty log message *** 2006-04-09 03:20 gb * [r4130] : 64-bit vectors: target::, not ppc64::. 2006-04-09 02:46 gb * [r4129] : Add missing paren ... 2006-04-09 02:45 gb * [r4128] : Recognize a :struct-by-value case. 2006-04-09 02:44 gb * [r4127] : X8664 ABI seems to pass structs by value, sort of. 2006-04-09 02:43 gb * [r4126] : Add -D_GNU_SOURCE to CFLAGS. 2006-04-09 02:42 gb * [r4125] : Port some of this to x8664; need to add code to maintain "used" bitmap. 2006-04-09 02:42 gb * [r4124] : Conditionalize out the FP infinity stuff in the short term. 2006-04-09 02:41 gb * [r4123] : %MAP-LFIMMS: lfun-vector changes. Should someday fix this crap. 2006-04-09 02:40 gb * [r4122] : Conditionalize, a little. 2006-04-09 02:40 gb * [r4121] : SAVE-DB-FUNCTION appears to be unused. 2006-04-09 02:39 gb * [r4120] : PRINT-A-NAN not #+ppc-target. 2006-04-09 02:39 gb * [r4119] : Error reentry stuff here. 2006-04-09 02:39 gb * [r4118] : Workaround ffi translator lossage in _inet_ntoa (just as Darwin has for years; why not fix this ?) 2006-04-09 02:38 gb * [r4117] : Error-reentry stuff to l1-readloop-lds (nothing to do with PPC or traps ...) 2006-04-09 02:37 gb * [r4116] : Ignore *CL*/*cl*, not necessarily *CCL*/*ccl*. 2006-04-08 11:20 gb * [r4115] : *** empty log message *** 2006-04-08 10:46 gb * [r4114] : X862-MVCALL needs to always save and restore, even if there's only one values-producing form (that form might produce more values than will fit in argument registers, and we therefore have to reserve space for a stack frame.) 2006-04-08 10:45 gb * [r4113] : Need to actually define INSERT-IMM64. 2006-04-08 10:43 gb * [r4112] : Scanning backwards for instructions depends on a reasonable architecture. 2006-04-08 10:43 gb * [r4111] : TARGET-XLOAD-MODULES for x8664. 2006-04-08 10:42 gb * [r4110] : Hide ppc32 double-float-vector stuff in target arch. 2006-04-08 10:41 gb * [r4109] : ignore *.*fsl 2006-04-08 05:20 gb * [r4108] : *** empty log message *** 2006-04-08 04:46 gb * [r4107] : XCOMPILE-LAMBDA is #+ppc-target for now. 2006-04-08 04:45 gb * [r4106] : Provide an FTYPE declaration for %DEFINE-PPC-LAP-FUNCTION. 2006-04-08 04:25 gb * [r4105] : Bump. 2006-04-08 04:20 gb * [r4104] : *** empty log message *** 2006-04-08 04:19 gb * [r4103] : ARCH::OPCODE, ARCH::OPERAND moved to CCL package. 2006-04-08 04:18 gb * [r4102] : OPCODE, OPERAND moved here and from arch:: to ccl::. 2006-04-08 04:17 gb * [r4101] : OPERAND, OPCODE belong elsewhere (in RISC-LAP, for instance). That's a different package, too. 2006-04-08 03:20 gb * [r4100] : *** empty log message *** 2006-04-08 02:46 gb * [r4099] : %CONS-FAKE-STACK-FRAME here. Really. 2006-04-08 02:45 gb * [r4098] : *IMMHEADER-1-TYPES* was wrong, too. Should I maybe look at the others, or just fix them as I stumble on them ? 2006-04-08 02:44 gb * [r4097] : jns->jne when checking for a DOUBLE-FLOAT. 2006-04-08 02:41 gb * [r4096] : Print file pos in FASL-DISPATCH debug message. 2006-04-08 02:40 gb * [r4095] : #+x8664: Check for fulltag-immheader-1, not fulltag-immheader-0 twice in fasl-scan-dispatch FASL-DUMP-SYMBOL needs a SYMPTR->SYMVECTOR. (Note that FASLOAD has never handled uninterned symbols that need a binding index.) 2006-04-07 14:20 gb * [r4094] : *** empty log message *** 2006-04-07 13:55 gb * [r4093] : Formatting changes, mostly. 2006-04-07 13:54 gb * [r4092] : multiple-value-calls need to reserve an outgoing stack frame when more than 3 args/ values are involved. 2006-04-07 13:53 gb * [r4091] : _SPrecover-values-for-mvcall (reserves outgoing frame if necessary.) 2006-04-07 13:52 gb * [r4090] : %CONS-FAKE-STACK-FRAME moved to lispequ. 2006-04-07 13:20 gb * [r4089] : *** empty log message *** 2006-04-07 12:28 gb * [r4088] : pop argument regs in the right order. 2006-04-07 12:27 gb * [r4087] : VALUES accounts for reserved frame. 2006-04-07 10:20 gb * [r4086] : *** empty log message *** 2006-04-07 09:59 gb * [r4085] : Don't push the function on the "value" stack when doing a MULTIPLE-VALUE-CALL, since its presence there can interfere with multiple-value continuations. (Can it be popped before recovering multiple values ? I didn't think of that, but I suspect that that'd be simpler than what I did; check this. Maybe there is -or once was- an issue with _SPrecovervalues bashing a register.) Pushing the function on the temp stack while collecting multiple values avoids the issue (but it may be far better to simply vpop it before the call to _SPrecovervalues.) Most of the issues related to X862-MVCALL calling itself recursively have to do with cleaning the function off of the stack in non tail-call cases. Since the function's not there anymore, the rationale for splitting this into a base and recursive case isn't very strong. Changes in X862-MVCALL exposed bugs in X862-INVOKE-FN: we don't want to emit or otherwise reference the mvpass label if it's null. 2006-04-07 09:51 gb * [r4084] : TEMP-PUSH-NODE, TEMP-POP-NODE. 2006-04-07 04:20 gb * [r4083] : *** empty log message *** 2006-04-07 03:27 gb * [r4082] : CDB-HASH-MASK happens to be = PPC32::TARGET-MOST-POSITIVE-FIXNUM. 2006-04-07 03:26 gb * [r4081] : Bugs, not worth describing in detail. 2006-04-07 03:25 gb * [r4080] : Bugs, not worth describing in detail ... 2006-04-07 03:24 gb * [r4079] : Simplify CFP-LFUN. 2006-04-07 03:23 gb * [r4078] : Fenceposts with the last gpr/fpr args. 2006-04-07 03:22 gb * [r4077] : tcr->linear_end -> tcr->save_rbp. 2006-04-07 03:21 gb * [r4076] : Maintain tcr->save_rbp on ffcall/callback. 2006-04-07 03:20 gb * [r4075] : Setup tcr->save_rbp when doing a callback. Pass rpc to %ERR-DISP, as it expects. 2006-04-07 03:20 gb * [r4074] : *** empty log message *** 2006-04-06 13:20 gb * [r4073] : *** empty log message *** 2006-04-06 13:03 gb * [r4072] : Call out to handle errors. 2006-04-06 13:03 gb * [r4071] : Ignore more signals. 2006-04-06 12:44 gb * [r4070] : Use new UUOs for slot-boundp checks, eep.address. 2006-04-06 12:43 gb * [r4069] : X862-FIXUP-FWD-REFS: don't assume that functions are uvectors or that their constants start at index 1. PROGV: don't keep interrupts off in the body. 2006-04-06 12:20 gb * [r4068] : *** empty log message *** 2006-04-06 11:58 gb * [r4067] : RESOLVE-EEP uuo needs source and dest registers to be recoverable. Use (int ($ #x8x)) for SLOT-UNBOUND, since it needs slot-vector, index, and dest reg to be recoverable. Note that the use of this range of (int N) operands may lead to confusion on platforms that use int #x80 for syscalls. That (a) doesn't apply to x8664, since SYSCALL is present and used on x8664 and (b) would only be a problem if the dest reg was %rax, which it won't be. 2006-04-06 11:53 gb * [r4066] : DEFSETF for INTERRUPT-LEVEL here. 2006-04-06 11:53 gb * [r4065] : Better (useful) X8664 *FEATURES*. 2006-04-06 11:52 gb * [r4064] : Remove some #+x86-target DBG traps. Add a few fixnum decls in and around DO-REHASH. 2006-04-06 11:51 gb * [r4063] : %RETURN-ADDRESS-FUNCTION: if argument is a "real" tagged return address, return the associated function. 2006-04-06 11:50 gb * [r4062] : Flesh this out some. The FAKE-STACK-FRAME stuff isn't quite the right idea on X8664; may need to revise it. 2006-04-06 11:49 gb * [r4061] : Remove HANDLE-GC-HOOKS. 2006-04-06 11:49 gb * [r4060] : DEFSETF for INTERRUPT-LEVEL moved ... anywhere but a platform-specific file. 2006-04-06 11:48 gb * [r4059] : HANDLE-GC-HOOKS isn't really platform-specific, so move it here. 2006-04-06 11:47 gb * [r4058] : Hashing code expects instance.hash to be a fixnum, so violating that expectation (as was done a few weeks ago for debugging) fouls things up royally. This may have led to instability on PPC as well. 2006-04-06 11:44 gb * [r4057] : Return false (for now) on unhandled UUO. (Need to merge code from another machine, which will try to call out to lisp.) 2006-04-06 11:43 gb * [r4056] : register_sigint_handler on x8664. 2006-04-05 13:20 gb * [r4055] : *** empty log message *** 2006-04-05 12:57 gb * [r4054] : nthrow variants: set %Rforeign_sp, not tcr slot. _SPcallback aligns the foreign stack the way that lisp expects. 2006-04-05 12:56 gb * [r4053] : Handle more stuff, including interrupt callbacks. 2006-04-05 12:55 gb * [r4052] : GETUID was wrong. 2006-04-05 12:55 gb * [r4051] : VERY early callback for interrupt handling. 2006-04-05 12:54 gb * [r4050] : Comment format 2006-04-05 12:54 gb * [r4049] : Arg order in lap macros. 2006-04-05 12:54 gb * [r4048] : Get *IMMHEADER-2-TYPES* (or whatever it's called) right. 2006-04-05 12:53 gb * [r4047] : DEFCALLBACK guts for X8664. 2006-04-05 11:20 gb * [r4046] : *** empty log message *** 2006-04-05 11:20 gb * [r4045] : Do some callbacks. Let's see if they work yet. 2006-04-05 11:19 gb * [r4044] : Use interrupt-checking macros. Fix callback typos. 2006-04-05 11:19 gb * [r4043] : Check for interrupts, and do so when we know that they've just become enabled. 2006-04-05 11:17 gb * [r4042] : Set high bit of tcr->interrupt pending when there's ... a pending interrupt. 2006-04-05 11:17 gb * [r4041] : Define xuuo stuff. 2006-04-05 04:20 gb * [r4040] : *** empty log message *** 2006-04-05 03:47 gb * [r4039] : callback changes. 2006-04-05 03:47 gb * [r4038] : rip "=", not "-". 2006-04-05 03:46 gb * [r4037] : Ignore SIGPWR, until we know better. 2006-04-05 01:20 gb * [r4036] : *** empty log message *** 2006-04-05 00:51 gb * [r4035] : Wimp out, back to 128GB for LinuxPPC64. 2006-04-05 00:50 gb * [r4034] : Don't use %GET-NATURAL in Linux version of PIPE, since file descriptors are always C ints. 2006-04-05 00:20 gb * [r4033] : *** empty log message *** 2006-04-05 00:03 gb * [r4032] : callback implementation moved to the backend. 2006-04-05 00:02 gb * [r4031] : ARGLIST-FROM-MAP: get &rest name right when &key is also present. 2006-04-04 22:20 gb * [r4030] : *** empty log message *** 2006-04-04 22:10 gb * [r4029] : linuxppc64 can't mmap 512GB; try 256GB. 2006-04-04 21:52 gb * [r4028] : Bump. New images soon. 2006-04-04 21:52 gb * [r4027] : Add slots for callback definition. 2006-04-04 21:52 gb * [r4026] : Remove a #-bccl. 2006-04-04 21:26 gb * [r4025] : print_unsigned_hex: conditionalize on word size. 2006-04-04 21:26 gb * [r4024] : First cut of _SPcallback. 2006-04-04 21:26 gb * [r4023] : Catch frames save foreign sp from register, not from tcr slot. 2006-04-04 21:25 gb * [r4022] : Uncomment the body of interrupt_handler. 2006-04-04 21:20 gb * [r4021] : *** empty log message *** 2006-04-04 21:18 gb * [r4020] : %LFUN-INFO & friends: do (FUNCTION-TO-FUNCTION-VECTOR). 2006-04-04 21:18 gb * [r4019] : check-nargs: bad comparison on max when both min and max. 2006-04-04 21:18 gb * [r4018] : spurious checkin; no real change 2006-04-04 05:20 gb * [r4017] : *** empty log message *** 2006-04-04 05:01 gb * [r4016] : Make purify() a nop, for now. 2006-04-04 04:43 gb * [r4015] : Maybe (check for x8664 Linux is untested). 2006-04-04 04:42 gb * [r4014] : Constants are word-size (not CPU-variant) -specific. 2006-04-04 04:42 gb * [r4013] : 512GB MAXIMUM_MAPPABLE_MEMORY on x8664/Linux as well. 2006-04-04 04:41 gb * [r4012] : Remove X86 conditionalization. Maybe merge x86-gc.c and ppc-gc.c back into a single file someday (or at least put truly shared code there),but for now this is simpler. 2006-04-04 04:38 gb * [r4011] : A little more infrastructure, so that we can actually call the GC. 2006-04-04 04:37 gb * [r4010] : Define UUO, trap opcodes. 2006-04-04 04:36 gb * [r4009] : It's ... alive. 2006-04-04 04:36 gb * [r4008] : Use new vector bounds traps. How about boundp checks ? 2006-04-04 04:35 gb * [r4007] : New uuo definitions. 2006-04-04 04:35 gb * [r4006] : DO-GC-AREAS,which is one of those macro things, moved here. Uses archmacro AREA-SUCC (defined here) to avoid TARGET-ARCH-CASE lossage. 2006-04-04 04:34 gb * [r4005] : DO-GC-AREAS,which is one of those macro things, moved to macros.lisp. 2006-04-04 04:33 gb * [r4004] : Zero out all of the write-lock's writer slot, not just the low half. 2006-04-04 04:32 gb * [r4003] : TRUE, FALSE : cmplr vs cmplri. Typical rookie PPC assembler mistake ... 2006-04-04 04:31 gb * [r4002] : %FREEBYTES looked like a job for %GET-NATURAL. 2006-04-04 04:31 gb * [r4001] : Use newfangled UUOs. 2006-04-04 04:31 gb * [r4000] : AREA-SUCC archmacro. 2006-04-04 04:30 gb * [r3999] : Add UUO-ERROR-UNBOUND, UUO-ERROR-SLOT-UNBOUND, UUO-ERROR-EEP-UNRESOLVED. 2006-04-02 07:20 gb * [r3998] : *** empty log message *** 2006-04-02 06:36 gb * [r3997] : xpMMXreg, sigaltstack stuff. 2006-04-02 06:35 gb * [r3996] : take signals on an alternate signal stack, run the handler on the foreign stack (this is all intended to keep signal handlers from splatting sigcontexts all over the tagged lisp stack.) 2006-04-02 06:33 gb * [r3995] : mmx register indices. 2006-04-02 06:33 gb * [r3994] : switch_to_foreign_stack(). 2006-04-02 06:32 gb * [r3993] : register_cstack_holding_area_lock() takes a natural stack size. 2006-04-02 06:32 gb * [r3992] : register_cstack_holding_area_lock() takes a natural stack size, sets up alternate signal stack if USE_SIGALTSTACK. 2006-04-02 06:31 gb * [r3991] : USE_SIGALTSTACK if X8664. (Might someday depend on platform.) 2006-04-01 06:20 gb * [r3990] : *** empty log message *** 2006-04-01 06:15 gb * [r3989] : gc.o -> x86-gc.o 2006-04-01 06:00 gb * [r3988] : New file; currently, an exact copy of ppc-gc.c, but they'll diverge some soon. 2006-04-01 05:56 gb * [r3987] : gc.o -> ppc-gc.o 2006-04-01 05:51 gb * [r3986] : Need prototype for heap_area_containing(). heap_node_area_containing(), which doesn't exist ... not so much. 2006-04-01 05:50 gb * [r3985] : Some functions moved here from gc.c before it was renamed/split. 2006-04-01 05:48 gb * [r3984] : rename/split 2006-04-01 05:46 gb * [r3983] : renamed 2006-03-31 08:20 gb * [r3982] : *** empty log message *** 2006-03-31 07:54 gb * [r3981] : Progvsave, progvrestore seem to work. Not sure about the tail end of the multiple-value case of THROW, but at least it's there. 2006-03-31 07:53 gb * [r3980] : still need work, but more stuff (including socket syscalls) are defined now. 2006-03-31 07:52 gb * [r3979] : Bump. 2006-03-31 07:52 gb * [r3978] : Conditionalize lots for x8664. 2006-03-31 07:51 gb * [r3977] : No debug trap in %CFP-LFUN. 2006-03-31 07:50 gb * [r3976] : Simplify *FEATURES* init a little. Try to ensure that :linuxx8664-target is there if it should be. 2006-03-31 07:46 gb * [r3975] : Define MAX-8-BIT-IVECTOR-SUBTAG, too. 2006-03-30 11:20 gb * [r3974] : *** empty log message *** 2006-03-30 10:23 gb * [r3973] : ELEMENT-SUBTYPE-TYPE for x8664. 2006-03-30 10:22 gb * [r3972] : Load some x8664 files, 2006-03-30 10:21 gb * [r3971] : Comments and formatting. 2006-03-30 10:20 gb * [r3970] : Some x8664 changes. 2006-03-30 10:20 gb * [r3969] : *** empty log message *** 2006-03-30 10:20 gb * [r3968] : Use REPLACE-FUNCTION-CODE another time or two. Set class's INSTANCE.HASH slots to class name, to make classes easier to identify in kernel debugger. (Might not want to keept this, but could set/clear a bit there.) 2006-03-30 10:18 gb * [r3967] : Comments and formatting changes only. 2006-03-30 10:17 gb * [r3966] : Start to move some stuff into platform-dependent files. 2006-03-30 10:17 gb * [r3965] : Conditionalize. 2006-03-30 10:16 gb * [r3964] : 64-bit conditionalization, and x8664 stuff. 2006-03-30 10:15 gb * [r3963] : 64-bit array type on all (both) 64-bit platforms. 2006-03-30 10:13 gb * [r3962] : TRUE and FALSE now in LAP. 2006-03-30 10:13 gb * [r3961] : Load "new" x8664 files. 2006-03-30 10:12 gb * [r3960] : Conditionalize some for x8664. 2006-03-30 10:12 gb * [r3959] : Some stuff moved here from l1-lisp-threads. 2006-03-30 10:11 gb * [r3958] : X8664-related changes. 2006-03-30 10:11 gb * [r3957] : Setup host-ftd based on compile-time target. Yow, recursion. 2006-03-30 10:10 gb * [r3956] : LET-GLOBALLY deals with symptr/symvector stuff. 2006-03-30 10:10 gb * [r3955] : Fix Linux MACHINE-INSTANCE. Move DISASSEMBLE and interface here. 2006-03-30 10:09 gb * [r3954] : Conditionalize for x8664; dump native functions via fasl-clfun. 2006-03-30 10:08 gb * [r3953] : Needed symptr/symvector stuff. 2006-03-30 10:08 gb * [r3952] : Add new x86 level-1 stuff; remove some truly ancient MCLisms. 2006-03-30 10:06 gb * [r3951] : Some fixes and additions, but still needs lots of work. 2006-03-30 10:06 gb * [r3950] : If instance.hash of an instance isn't a fixnum, treat it as a class name. (Hard to otherwise recognize classes from the kernel.) 2006-03-30 10:04 gb * [r3949] : rename poweropen_callback ->callback. 2006-03-30 10:04 gb * [r3948] : Lots of bug fixes. Multiple-value THROW still needs work, and CALLBACK still needs to be written (both sides). 2006-03-30 10:02 gb * [r3947] : No FPSCR here, but an MXCSR. 2006-03-30 10:01 gb * [r3946] : Error call in misc_set_invalid involves 4 args. 2006-03-30 10:00 gb * [r3945] : Larger GC defaults if 64-bit, not just PPC64. Insist on Linux kernel 2.6 if x8664 linux. 2006-03-30 09:59 gb * [r3944] : Missing #\% in format string. 2006-03-30 09:56 gb * [r3943] : More stuff. 2006-03-30 09:55 gb * [r3942] : Nee files. 2006-03-30 09:54 gb * [r3941] : Remove debug traps from things that're believed to be working. 2006-03-30 09:54 gb * [r3940] : CLASS-OF changes/fixes. Move #'TRUE/#'FALSE here. 2006-03-30 09:53 gb * [r3939] : %SYMPTR->SYMBOL: check for nilsym, not NIL. 2006-03-30 09:52 gb * [r3938] : %FIXNUM-TRUNCATE: return results in the right order. %IASL: shift the right register. 2006-03-30 09:51 gb * [r3937] : Fix %FUDGE-HEAP-POINTER; define INTERRUPT-LEVEL. 2006-03-30 09:51 gb * [r3936] : %get-mxcsr stores the mxcsr in memory so that we can load it into a fixnum. 2006-03-30 09:50 gb * [r3935] : %CFP-LFUN: try to recognize multiple-value returning calls, return NIL if tra offset is 0. 2006-03-30 09:49 gb * [r3934] : Remove debug traps; use registers sanely in the SETF case. 2006-03-30 09:48 gb * [r3933] : Implement TRUE and FALSE in LAP. 2006-03-30 09:48 gb * [r3932] : Change the name of a constant that'll get clobbered when this is cloned anyway. (Is that clear ?) 2006-03-30 09:47 gb * [r3931] : %TYPE-OF fixes for x8664. We haven't had interpreted functions for a few years. 2006-03-30 09:46 gb * [r3930] : %SET-DOUBLE-FLOAT-EXP: endianness, use symbolic constant for high word. 2006-03-30 09:45 gb * [r3929] : SUBTAG-BYTES for X8664. 2006-03-30 09:45 gb * [r3928] : Lots of bug fixes. 2006-03-30 09:44 gb * [r3927] : REQUIRE what we need. Get CATCH-FRAME definition right. .SPpoweropen-callback -> .SPcallback. 2006-03-30 09:43 gb * [r3926] : Lots of changes and bugfixes. 2006-03-30 09:42 gb * [r3925] : Avid path in REQUIRE; create native functions. 2006-03-30 09:42 gb * [r3924] : Run natively: provide X8664-XDISASSEMBLE. 2006-03-30 09:41 gb * [r3923] : Assemble things (uuos) with a "pseudo" ModRM byte, which allows encoding two 4-bit x8664 registers in a single byte. 2006-03-30 09:40 gb * [r3922] : Need to explicitly (REQUIRE "ARCH"). 2006-03-30 09:39 gb * [r3921] : Define high and low 32-bit parts of a DOUBLE-FLOAT. 2006-03-30 09:39 gb * [r3920] : DISASSEMBLE and FUNCTION-FOR-DISASSEMBLY moved elsewhere. 2006-03-26 04:20 gb * [r3919] : *** empty log message *** 2006-03-26 03:47 gb * [r3918] : Fix x8664 lisp register display. 2006-03-26 03:42 gb * [r3917] : Fix export of CCL::GETENV and CCL::SETENV. 2006-03-26 03:37 gb * [r3916] : Dozens of bug fixes. Do uvref (not yet uvset) via jump table. 2006-03-26 03:36 gb * [r3915] : Do arithmetic correctly in Cons macro. Split up Misc_Alloc. 2006-03-26 03:34 gb * [r3914] : Tag juggling. 2006-03-26 03:34 gb * [r3913] : Identify SLOT-VECTORs in kernel debugger. 2006-03-26 03:33 gb * [r3912] : Add some error numbers for uvref/uvset failures that should never happen. 2006-03-26 03:32 gb * [r3911] : Change x8664 lisp register display. 2006-03-26 03:32 gb * [r3910] : IMPURIFY: maybe grow the dynamic area, but don't shrink it. 2006-03-26 03:31 gb * [r3909] : Symbol address may be longer than 32 bits. 2006-03-26 03:31 gb * [r3908] : sgbuf has been lisp_thread for the past few years. 2006-03-26 03:30 gb * [r3907] : Do uvref/uvset via jump tables. (Fewer branches, more redundancy.) 2006-03-26 03:29 gb * [r3906] : Identify SLOT-VECTORs. Print TRAs. 2006-03-26 03:28 gb * [r3905] : Skip over lexpr nonsense. 2006-03-26 03:28 gb * [r3904] : More GDB macros. 2006-03-26 03:27 gb * [r3903] : %GF-DISPATCH-TABLE-REF uses %SVREF. 2006-03-26 03:26 gb * [r3902] : MACHINE-VERSION caches the result per-session; looks in /proc/cpuinfo for something machine-dependent on Linux, passes 64-bit stuff on DarwinPPC64. 2006-03-26 03:25 gb * [r3901] : LFUN-BITS-KNOWN-FUNCTIONS uses arch macros to access function vector. 2006-03-26 03:23 gb * [r3900] : Bump. 2006-03-26 03:23 gb * [r3899] : No more interpreted functions. 2006-03-26 03:23 gb * [r3898] : Mostly comment and formatting changes. 2006-03-26 03:22 gb * [r3897] : Fix typos in x86 function cloning; comment/formatting changs. 2006-03-26 03:21 gb * [r3896] : Lots of formatting changes; try to use DEFGLOBAL, since special variable lookup is ~10 instructions or so. 2006-03-26 03:20 gb * [r3895] : HOST-PLATFORM, PLATFORM-DESCRIPTION: generalize, a little. 2006-03-26 03:20 gb * [r3894] : *** empty log message *** 2006-03-26 03:19 gb * [r3893] : LFUN-KEYVECT access function vector, if that's not the function. NREVERSE checks for NIL, mostly so I wouldn't have to single-step through that case in GDB. 2006-03-26 03:18 gb * [r3892] : Shift in the correct direction, sign-extend (or don't) as apropriate, etc. Still some debug traps on things that haven't been called yet. 2006-03-26 03:17 gb * [r3891] : CLASS-OF fixes. 2006-03-26 03:17 gb * [r3890] : If the recursive list case of EQUAL fails, fail; don't fall into the uvector case ... 2006-03-26 03:16 gb * [r3889] : Fenceposts in fixnum integer-length. 2006-03-26 03:15 gb * [r3888] : Strip-tag-to-fixnum without shifts. 2006-03-26 03:15 gb * [r3887] : Stuff here seems to work, so far ... 2006-03-26 03:14 gb * [r3886] : debug traps on slot-lookup LAP prototypes. 2006-03-26 03:12 gb * [r3885] : rearrange %type-of for tag changes. 2006-03-26 03:12 gb * [r3884] : No need to specify an initial value 0 when stack-consing a bignum in %BIGNUM-RANDOM. Add a fixnum decl or two. 2006-03-26 03:11 gb * [r3883] : Symbol aren't (necessarily) uvectors. X86 breakpoints if we hash any symbols ... 2006-03-26 03:09 gb * [r3882] : Um, most of this is supposed to be #+64-bit-target, not #+ppc64-target. 2006-03-26 03:09 gb * [r3881] : SET-CLOSURE-FORWARD-REFERENCE: functions aren't uvectors. Lots of other changes/fixes. 2006-03-26 03:08 gb * [r3880] : Tag reorganization. MISC-SYMBOL-OFFSET, MISC-FUNCTION-OFFSET. FUNCTION-TO-FUNCTION-VECTOR, FUNCTION-VECTOR-TO-FUNCTION arch macros have something to do here. 2006-03-26 03:06 gb * [r3879] : Some operators can push results directly. Lots of bug fixes. 2006-03-26 03:05 gb * [r3878] : CMPB: accept signed/unsigned immediate. 2006-03-26 03:04 gb * [r3877] : Arch macros for FUNCTION-TO-FUNCTION-VECTOR and FUNCTION-VECTOR-TO-FUNCTION. (NOPs.) 2006-03-26 03:03 gb * [r3876] : MACPTRP. 2006-03-26 03:03 gb * [r3875] : COMPILE: don't check for INTERPRETED-FUNCTION, unless/until we have them again (in some form.) 2006-03-20 17:20 gb * [r3874] : *** empty log message *** 2006-03-20 17:05 gwking * [r3873] : Exporting getenv and setenv from CCL 2006-03-20 04:20 gb * [r3872] : *** empty log message *** 2006-03-20 03:30 gb * [r3871] : Old-style destructuring, VALUES. 2006-03-20 03:29 gb * [r3870] : Define %nargs_b. 2006-03-20 03:29 gb * [r3869] : Define registers/bitfields used in old-style destructuring subprims. 2006-03-20 03:28 gb * [r3868] : $XNOFINFUNCTION. 2006-03-20 03:27 gb * [r3867] : Conditionalize out bit, bitfield accessors until they work ... 2006-03-20 03:27 gb * [r3866] : Conditionalize GET-REGISTER-VALUE/SET-REGISTER-VALUE catch frame access. 2006-03-20 03:26 gb * [r3865] : Conditionalize, mostly for whether the implementation has separate code-vector and function objects or not. (ppc vs x86, for now.) 2006-03-20 03:25 gb * [r3864] : %FIXNUM-GCD. Slower then the one that counts bits, but doesn't need as many registers. 2006-03-20 03:24 gb * [r3863] : Do the lexpr/methd-context apply stuff. %X86-CLONE-FUNCTION, REPLACE-FUNCTION-CODE. Wouldn't hurt to do some typechecking in some of the lap code here. 2006-03-20 03:22 gb * [r3862] : Yet. 2006-03-20 03:22 gb * [r3861] : REPLACE-FUNCTION-CODE abstraction. 2006-03-20 03:22 gb * [r3860] : UNSET-FIN-TRAMPOLINE doesn't reference constants. (This could neve have worked.) 2006-03-20 03:21 gb * [r3859] : Replace error format string 9. 2006-03-20 03:20 gb * [r3858] : &lexpr fixes, old-style destructuring. 2006-03-20 03:20 gb * [r3857] : *** empty log message *** 2006-03-20 03:20 gb * [r3856] : Add the xtra slot to lisp-frame. yield syscall, gf-code-size. 2006-03-20 03:19 gb * [r3855] : Fix some lexpr confusion, mv-call recursive problems. 2006-03-20 03:18 gb * [r3854] : SAVE-FRAME-VARIABLE-ARG-COUNT, CALL-SUBPRIM. 2006-03-20 03:17 gb * [r3853] : Support :CODE-SIZE and :ORG directives. Change :CONSTANTS to :FIXED-CONSTANTS. 2006-03-20 03:16 gb * [r3852] : Can use MAKE-ARRAY rather than MAKE-SEQUENCE. 2006-03-18 07:20 gb * [r3851] : *** empty log message *** 2006-03-18 06:38 gb * [r3850] : Special case references to NIL in package htabs. Leave deleted markers behind for symbols that we haven't seen references to; leave empty (0) words empty. 2006-03-18 06:36 gb * [r3849] : Add a few lisp debugging things. 2006-03-18 06:36 gb * [r3848] : Recognize multiple-value returning frames and find the real tra underneath ret1valn. 2006-03-18 06:35 gb * [r3847] : "gen" tcalls had the same bug as the known-stack-args cases did several days ago. 2006-03-18 06:34 gb * [r3846] : New do_funcall(). 2006-03-18 06:33 gb * [r3845] : When we field an alloc trap, make sure that both the allocptr register and the save_allocptr slot in the TCR are updated and consistent with each other. 2006-03-18 06:32 gb * [r3844] : A lisp frame may have two return addresses (one for the multiple-value mechanism and the real one.) 2006-03-18 06:31 gb * [r3843] : If there's such a thing as fulltag_symbol, try to use it. 2006-03-18 06:29 gb * [r3842] : Symptr/symvector deconflation. 2006-03-18 06:28 gb * [r3841] : Add %LISP-LOWBYTE-REF; could be open-coded (and may once have been.) 2006-03-18 06:27 gb * [r3840] : Some SYMPTR->SYMVECTOR usage. Define %FASL-READ-SIGNED-LONG; use it in a few places. Assume that scalar integers are stored big-endian; construct native-endian integers from them. 2006-03-18 06:26 gb * [r3839] : Lots of SYMPTR->SYMVECTOR usage. 2006-03-18 06:25 gb * [r3838] : Do %TYPE-OF for x8664. 2006-03-18 06:24 gb * [r3837] : Use SYMPTR->SYMVECTOR where necessary, so that we can %SVREF around in symbols. A little light reorganization, to make it easier to add debugging breakpoints for things like FUNCTION-NAME. 2006-03-18 06:22 gb * [r3836] : Kinder, gentler CHECK-MISC-BOUND. (What was I thinking ?) More cleanup to FUNCALL and variants. MAKE-TSP-CONS: fix braindamage, use only one register. (Need to write down pc-luser-xp constraints for TSP allocation, but I think that this meets the current artist's conception of them.) %SYMBOL->SYMPTR does what it does on other platforms (e.g., map NIL to a proxy symbol and/or insist on a real symbol.) %SYMPTR->SYMVECTOR maps something tagged as a symbol to the underlying uvector by playing with tag bits, and %SYMVECTOR->SYMPTR does the inverse. There were two definitions of %SCHARCODE in the file; the later one (that took effect) was completely wrong. 2006-03-18 06:20 gb * [r3835] : *** empty log message *** 2006-03-18 06:05 gb * [r3834] : Add HONS fields to AREA definition. SYMVECTOR->SYMPTR and its inverse don't do much, but it's not worth spending a function call to do a little. 2006-03-18 06:03 gb * [r3833] : Don't sign-extend the $LFBITS-NONAME-BIT; doing so had to do with 29-bit fixnums on an old 68K implementation. Fix parenthesization in %i-, so that we branch in all cases that need to. Implement %SYMPTR->SYMVECTOR and %SYMVECTOR->SYMPTR, which change something tagged X8664::FULLTAG-SYMBOL into the underlying uvector and back again. 2006-03-18 05:59 gb * [r3832] : Change the sign of TRA displacements, to match what the compiler expects and what some other things do. When checking to see if an ASSUMED-SHORT[-CONDITIONAL]-BRANCH can use a 1-byte branch displacement, get the sign of that displacment right. (Doing it the wrong way was encoding a forward branch to a label 128 bytes away in one byte, which made it a backward branch 128 bytes away ...) 2006-03-18 05:55 gb * [r3831] : Use NTH-IMMEDIATE to access COMBINED-METHOD slots. 2006-03-18 05:54 gb * [r3830] : Comments clarify that scalar integers are always dumped big-endian. (They have to be read/written a byte at a time anyway; there's no clear advantage in writing them in native endianness.) 2006-03-18 05:52 gb * [r3829] : SYMPTR->SYMVECTOR and SYMVECTOR->SYMBOL are arch-specfic (NOP on PPC, inlined tag-bits stuff on x8664.) 2006-03-18 05:51 gb * [r3828] : Comment formatting. 2006-03-18 05:50 gb * [r3827] : Star to conditionalize, a little. 2006-03-18 05:48 gb * [r3826] : Conditionalize. 2006-03-18 05:46 gb * [r3825] : %SYMPTR->SYMVECTOR and %SYMVECTOR->SYMPTR shold never be called on PPC, but implement them as idenity operations, just in case. 2006-03-18 05:44 gb * [r3824] : Worth inlining %SYMPTR->SYMVECTOR on x8664 (should never happen on PPC, but treat it as identity function there.) 2006-03-16 09:21 gb * [r3823] : *** empty log message *** 2006-03-16 09:03 gb * [r3822] : Spell LINUX correctly. 2006-03-16 09:00 gb * [r3821] : We were inlining NTH-VALUE incorrectly, but were always inlining it. Use debug traps to catch some cases that're NYI. 2006-03-16 08:58 gb * [r3820] : X8664 register-printing changes. 2006-03-16 08:58 gb * [r3819] : Treat VEC64/VEC128 like large integers, for now. 2006-03-16 08:57 gb * [r3818] : gf accessors use NTH-IMMEDIATE. 2006-03-16 08:56 gb * [r3817] : Conditionaliz on target word size. 2006-03-16 08:56 gb * [r3816] : Need to be aware of cases where host/target endianness differ. Byte-swap 16/32-bit constant ivectors; note that other cases may also need byte-swapping. 2006-03-16 08:54 gb * [r3815] : NTH-IMMEDIATE is arch-specific; so is SET-NTH-IMMEDIATE, which is its SETF inverse. 2006-03-16 08:53 gb * [r3814] : So far, Darwin's the only platform that uses signed chars. (For a good laugh, note that the PPC has an unsigned "load byte" instruction, but it takes two instructions to treat char as signed. This is, of course, much less significant than all of the other things that are wrong about Darwin ...) 2006-03-16 08:51 gb * [r3813] : RISC-LAP is PPC-specific (and maybe specific to other RISC architectures). X86-DISASSEMBLE should be built for x8664. 2006-03-16 08:49 gb * [r3812] : Some conditinalization. 2006-03-16 08:49 gb * [r3811] : Ignore fasl extensions yet to come. 2006-03-16 08:46 gb * [r3810] : Bump. 2006-03-16 08:46 gb * [r3809] : Some conditionalization; need more, and need to handle wrong-endianness. 2006-03-16 08:44 gb * [r3808] : %KERNEL-GLOBAL-OFFSET-FORM: probably should lose this. It's a function, and therefore has some trouble accessing the right (macroexpand-time) backend. 2006-03-16 08:43 gb * [r3807] : Set *.FASL-PATHAME* via #.'ed reference to target backend. 2006-03-16 08:42 gb * [r3806] : Conditionalize a little. 2006-03-16 08:42 gb * [r3805] : Use #. to set OS package name fron backend/ftd. 2006-03-16 08:41 gb * [r3804] : %SYMPTR->SYMBOL is often called with real (tagged) symbols; make it a nop in that case. 2006-03-16 08:40 gb * [r3803] : %SET-NTH-IMMEDIATE. 2006-03-16 08:40 gb * [r3802] : Start some of this. Still need to decide whether to use lexprs; better decide soon. 2006-03-16 08:38 gb * [r3801] : TRAP-UNLESS-CHARACTER. Float memory accessors/setters for the constant-offset case. NTH-VALUE was wrong; now simpler. MAKE-TSP-VCELL. Binding subprims. Yet another FUNCALL scheme. 2006-03-16 08:36 gb * [r3800] : MAKE-SYMBOL returns the symbol pointer. Maybe conditionalize this for X8664. 2006-03-16 08:35 gb * [r3799] : Hopefully, the "split" 64-bit lseek is at best LinuxPPC32-specific. 2006-03-16 08:34 gb * [r3798] : Use %ra0 for the inner function when making a closure, to avoid conflicts with other temps. No need to reference memory on self-calls (and good reasons not to ...) X862-UNBOXED-INTEGER-ARG-TO-REG handles 64-bit regs. 2006-03-16 08:31 gb * [r3797] : EXTRACT-TYPECODE: use a GENSYMed label. 2006-03-16 08:30 gb * [r3796] : New stuff for "pre-allocated" constants. 2006-03-16 08:29 gb * [r3795] : Didn't have the relatively recent STACK-SIZE. 2006-03-16 08:29 gb * [r3794] : Change the %GET-KERNEL-GLOBAL[-PTR] archmacros, so that they'll work with cross-compilation. Fixnum offsets aren't supported; they weren't used, this isn't public, and it wasn't well-defined anyway. Add %NTH-IMMEDIATE. 2006-03-16 08:28 gb * [r3793] : Change the %GET-KERNEL-GLOBAL[-PTR] archmacros, so that they'll work with cross-compilation. Fixnum offsets aren't supported; they weren't used, this isn't public, and it wasn't well-defined anyway. 2006-03-14 21:20 gb * [r3792] : *** empty log message *** 2006-03-14 21:08 gb * [r3791] : remove 2006-03-14 21:02 gb * [r3790] : Always use physical bin, l1-fasls relative pathnames. 2006-03-14 21:02 gb * [r3789] : Simpler translations. 2006-03-14 20:54 gb * [r3788] : removed 2006-03-14 20:32 gb * [r3787] : new file. 2006-03-14 20:24 gb * [r3786] : Step 1: don't hardwire the target fasl pathname in a big #+ nest, get it from the target backend. (Hope that works with macrolet.) 2006-03-14 20:22 gb * [r3785] : _SPstack_misc_alloc fixes. _SPsyscall: just return negative return values, don't negate them. 2006-03-14 20:20 gb * [r3784] : Hide the incomplete definitions. This needs work; should use real foreign types. 2006-03-14 20:20 gb * [r3783] : *** empty log message *** 2006-03-14 20:19 gb * [r3782] : %init-gvector: pop the last element first. 2006-03-14 20:19 gb * [r3781] : :linuxx8664-target on features, too. 2006-03-14 20:18 gb * [r3780] : x862-allocate-initialized-gvector: works better if the gvector is actually allocated ... 2006-03-14 10:20 gb * [r3779] : *** empty log message *** 2006-03-14 10:11 gb * [r3778] : Remove a comment that described the mid-1990s behavior of MacsBug. 2006-03-14 10:10 gb * [r3777] : In X86-LAP-EQUATE-FORM, check for and disallow bindings that match defined X86 register names. 2006-03-14 10:09 gb * [r3776] : Some UUOs shouldn't terminate a disassembly block. 2006-03-14 10:09 gb * [r3775] : Whoops. 2006-03-14 10:01 gb * [r3774] : Fixes in _SPcall_closure, _SPstkgvector. 2006-03-14 10:00 gb * [r3773] : Handle uuo_debug_trap. 2006-03-14 10:00 gb * [r3772] : Iflags is REG_EFL on Linux. 2006-03-14 09:59 gb * [r3771] : Make (R)egisters command work on x8664. 2006-03-14 09:59 gb * [r3770] : BACKEND-SYSCALL-DEFINITIONS uses whatever bits are in the backend's BACKEND-PLATFORM-SYSCALL-MASK. 2006-03-14 09:57 gb * [r3769] : New file. Woefully incomplete, and may need word-size fixes. 2006-03-14 09:57 gb * [r3768] : Set the mask per the new (new) scheme. 2006-03-14 09:56 gb * [r3767] : Fix a few bugs in heap-walking code. 2006-03-14 09:56 gb * [r3766] : %NTH-IMMEDIATE needed to box; both it and %FUNCTION-CODE-WORDS needed to avoid naming parameters FN. 2006-03-14 09:55 gb * [r3765] : Use the right set of syscall definitions for Linux. 2006-03-14 09:54 gb * [r3764] : Add a fixnum declaration. 2006-03-14 09:54 gb * [r3763] : Include the right set of syscall defs for Linux. 2006-03-14 09:53 gb * [r3762] : Need to use %symptr->symbol for x86; harmless on PPC. 2006-03-14 09:53 gb * [r3761] : SETUP-VCELL-ALLOCATION needs to bias the size by fulltag-misc. 2006-03-14 09:52 gb * [r3760] : At least for Linux/X86/64, need to look at os/cpu/word-size to deterine applicable syscall defs. 2006-03-14 09:51 gb * [r3759] : All PPC (32/64) backends can use syscall defs for OS/CPU, regardless of word size. 2006-03-14 09:50 gb * [r3758] : Add PLATFORM-SYSCALL-MASK to backend struct. 2006-03-14 09:21 gb * [r3757] : *** empty log message *** 2006-03-14 09:10 gb * [r3756] : renamed 2006-03-13 10:20 gb * [r3755] : *** empty log message *** 2006-03-13 09:33 gb * [r3754] : Bump. 2006-03-13 09:24 gb * [r3753] : AREA constants now in CCL package. 2006-03-13 09:23 gb * [r3752] : ret1valn: no LEAVE here. nthrowvalues: fix several bugs. 2006-03-13 09:22 gb * [r3751] : TSP_Alloc_Var: zero the dnode at the new tsp, as well. 2006-03-13 09:21 gb * [r3750] : Define AREA constants here (in CCL package.) 2006-03-13 09:21 gb * [r3749] : *** empty log message *** 2006-03-13 09:20 gb * [r3748] : Move DO-CONSING-AREAS here. AREA-CODE is an arch macro. 2006-03-13 09:20 gb * [r3747] : Move DO-CONSING-AREAS to macros.lisp. 2006-03-13 09:19 gb * [r3746] : Walk static, dynamic areas. 2006-03-13 09:19 gb * [r3745] : Comments. 2006-03-13 09:18 gb * [r3744] : %MAP-AREAS call might have to deal with symbol/symbol-vector distinction. 2006-03-13 09:18 gb * [r3743] : AREA constants are in CCL package, not PPC. %MAP-LFUNS might have to deal with function/function-vector distinction. 2006-03-13 09:17 gb * [r3742] : AREA constants are in CCL package, not PPC. 2006-03-13 09:17 gb * [r3741] : %GROW-HASH-TABLE is always called inside WITHOUT-INTERRUPTS, so no need to do it again. 2006-03-13 09:16 gb * [r3740] : Push at most one stack frame on function entry ... 2006-03-13 09:14 gb * [r3739] : AREA-CODE archmacro. 2006-03-13 09:14 gb * [r3738] : AREA constants moved to CCL package. 2006-03-12 07:20 gb * [r3737] : *** empty log message *** 2006-03-12 06:54 gb * [r3736] : Don't change handling of SIGUSR1 or SIGUSR2; I don't think that we use them under Linux. 2006-03-12 06:52 gb * [r3735] : builtin_eql was extracting lisptags and comparing fulltags, and thus returning false when it shouldn't have. nthrow1value was making too small of a TSP frame around the call to the cleanup code, clobbering TSP links in some cases. It was also failing to pop catch frames. stkgvector was unimplented. _car order-of-arguments problems in (at least) spreadargz. 2006-03-12 06:47 gb * [r3734] : Push a 0 and make the initial lisp %rbp point to it, to mark the end of the line. 2006-03-12 06:47 gb * [r3733] : Sketchy kernel-debugger backtrace. 2006-03-12 06:46 gb * [r3732] : A lisp_frame struct, for kernel backtrace. 2006-03-12 06:46 gb * [r3731] : Try (temporily) disabling Lisp's SIGINT handler. There seem to be problems using a keyboard interrupt to awaken GDB on the version of x96-64 Linux I'm using. This didn't fix those problems completely, but we won't need to handle SIGINT in the lisp until more code works. 2006-03-12 06:43 gb * [r3730] : I may have parenthesized nodeheader_tag_p and immheader_tag_p; several versions of GCC generated code that looked wrong, but since it's C it's hard to tell. Splitting the macros up (into "mask definition" and "mask testing" seems to generate the right code (a bit test if optimization's cranked up.) 2006-03-12 06:41 gb * [r3729] : Fix a few typos (fixnum vs unboxed constants, etc.) %ATOMIC-DECF-PTR-IF-POSITIVE should return what we would have (might have) stored in the word accessed by the pointer (in particular, it should return a negative value if there are no other threads waiting on a recursive lock.) 2006-03-12 06:38 gb * [r3728] : Fix file name in comment. 2006-03-12 06:38 gb * [r3727] : ENSURE-RESERVED-FRAME: test was backwards. 2006-03-12 06:37 gb * [r3726] : Don't use FN as arg name in LFUN-VECTOR, since x8664::fn is a constant (symbolic register name.) Add missing close paren. 2006-03-12 06:36 gb * [r3725] : Ensure stack fencepost in simple &optional arg case. Do the variable-arg-count case if complex &optional, no &rest/&key, and some (possibly unsupplied) optionals may or may not have been pushed by the caller. 2006-03-11 11:20 gb * [r3724] : *** empty log message *** 2006-03-11 11:11 gb * [r3723] : Fix conditionalization: we don't define PPC32, but it might be clearer if we did. 2006-03-11 10:28 gb * [r3722] : Bump. 2006-03-11 10:20 gb * [r3721] : New mechanism, mostly to make cross-compilation easier. 2006-03-11 10:20 gb * [r3720] : *** empty log message *** 2006-03-11 10:20 gb * [r3719] : Extra paren; %NTH-IMMEDIATE and function/lfun-vector stuff, CLOSURE-FUNCTION here until that can be abstracted away sanely. 2006-03-11 10:18 gb * [r3718] : %STORE-IMMEDIATE-CONDITIONAL needs to discard the frame that the caller reserved. 2006-03-11 10:17 gb * [r3717] : Comment out the PPC LAP (e.g., everything in the file, still). 2006-03-11 10:15 gb * [r3716] : CLOSURE-FUNCTION here. (Abstract away NTH-IMMEDIATE ?) 2006-03-11 10:14 gb * [r3715] : Indentation. 2006-03-11 10:13 gb * [r3714] : Move CLOSURE-FUNCTION to arch-specific dir. Use LFUN-VECTOR, which should hopefully compile away on PPC but needs to be there on X8664. 2006-03-11 10:11 gb * [r3713] : Now that subprim addresses are higher, there's another significant byte to deal with in the subprim jump in %INIT-CLOSURE. Use a temporary - and a smaller instruction sequence - in MISC-ELEMENT-COUNT-FIXNUM, so that src and dest can overlap. (More instructions, but fewer bytes.) 2006-03-11 10:09 gb * [r3712] : LFUN-VECTOR and LFUN-VECTOR-LFUN are arch-specific macros. 2006-03-11 10:07 gb * [r3711] : Reformat a comment. 2006-03-11 05:20 gb * [r3710] : *** empty log message *** 2006-03-11 05:06 gb * [r3709] : Target-specific static-space (NIL representation, mostly) init functions. Move image-base-address up a few terabytes. Undefined function and macro trampolines are vectors with an executable fixnum in the 0th word. 2006-03-11 05:05 gb * [r3708] : Target-specific static-space (NIL representation, mostly) init functions. 2006-03-11 05:04 gb * [r3707] : Lots of changes, to deal with symbol/function tagging issues, x86 clfuns, etc. 2006-03-11 05:03 gb * [r3706] : Endianness, word-size conditionalization stuff. 2006-03-11 05:02 gb * [r3705] : C files depend on lisptypes.h. 2006-03-11 05:01 gb * [r3704] : Don't try to link into low memory (no real reason to.) Do try to put spjump at a fixed/known/page-aligned address. 2006-03-11 04:59 gb * [r3703] : Recognize some x8664 tags. 2006-03-11 04:59 gb * [r3702] : Use Rforeign_sp; get valence right on exit. 2006-03-11 04:58 gb * [r3701] : Rename ffcall/syscall. 2006-03-11 04:58 gb * [r3700] : Lots of fixes/changes (including ffcall/syscall). 2006-03-11 04:56 gb * [r3699] : TSP_Alloc_Size is an immediate operand in TSP_Alloc_Fixed(). Fix allocptr/allocbase comparison when consing. Misc_Alloc_Fixed: m4/gas $ conflation. 2006-03-11 04:54 gb * [r3698] : Extra parens around xpGPR expansion. 2006-03-11 04:54 gb * [r3697] : Actually handle something (uuo_alloc()), since we can't do much without it. 2006-03-11 04:53 gb * [r3696] : Call the foreign sp register Rforeign_sp, to avoid m4 lossage. nrs starts at #x2020, because NIL is between there and #x2000. 2006-03-11 04:52 gb * [r3695] : Iallocptr for trap handler. tcr was missing foreign_sp definition. 2006-03-11 04:51 gb * [r3694] : get_tcr() change for HAVE_TLS case. 2006-03-11 04:51 gb * [r3693] : Put a sane value in the FPSCR before calling back to lisp on an exception. 2006-03-11 04:50 gb * [r3692] : Use address of bad_funcall, not "value". Disable EGC on x8664. 2006-03-11 04:49 gb * [r3691] : formatting change only. 2006-03-11 04:49 gb * [r3690] : Layout of lisp-globals/nrs on x8664 not identical to ppc64 (NIL representation differs.) 2006-03-11 04:48 gb * [r3689] : Lisp register stuff for x8664. 2006-03-11 04:48 gb * [r3688] : Fix some conditionalization. 2006-03-11 04:47 gb * [r3687] : Larger IMAGE-BASE-ADDRESS for X8664 Linux. 2006-03-11 04:46 gb * [r3686] : Try to make things a little more portable, but really need to keep things in the target-ftd. Soon. 2006-03-11 04:45 gb * [r3685] : Don't reference TARGET::LEAST-SIGNIFICANT-BIT in ASH; express limits in terms of TARGET::NBITS-IN-WORD. 2006-03-11 04:45 gb * [r3684] : Add some fixnum decls to %GET-CSTRING, so that single-stepping through it in GDB is faster. 2006-03-11 04:44 gb * [r3683] : %YIELD here, so all level-0 syscalls are in one file. 2006-03-11 04:43 gb * [r3682] : Comment indentation. 2006-03-11 04:43 gb * [r3681] : We generally want foreign symbol addresses to be boxed integers on X8664, and can't assume that they're fixnum-aligned. 2006-03-11 04:41 gb * [r3680] : Well, it's a start ... 2006-03-11 04:41 gb * [r3679] : %KERNEL-IMPORT boxes what may not have been a fixnum. 2006-03-11 04:40 gb * [r3678] : Fix typo. STORE-NODE-CONDITIONAL pops stack arg, discards stack frame reserved by caller before jumping into kernel. 2006-03-11 04:39 gb * [r3677] : Missing SET-MXCSR-CONTROL was hidden in duplicate SET-MSCSR-STATUS. May want to always set/get the control and rounding modes. 2006-03-11 04:37 gb * [r3676] : Comment out the missing lexpr/apply stuff, so that the rest of the file compiles. 2006-03-11 04:36 gb * [r3675] : %INIT-MISC wants/needs to be in LAP because of write-barrier issues. 2006-03-11 04:35 gb * [r3674] : It's OK (in fact necessary) to use FP0 as a temporary. 0.0 is in FPZERO. 2006-03-11 04:34 gb * [r3673] : Spit out a fasl op while writing CLFUNs. 2006-03-11 04:33 gb * [r3672] : Don't CLEAR-FPU-EXCEPTIONS, at least without knowing why. 2006-03-11 04:32 gb * [r3671] : SETcc sets the byte to 1 or 0, so we need to negate. Other bug fixes and additions. 2006-03-11 04:31 gb * [r3670] : Fix a typo in X8664-MISC-BYTE-SIZE. Move subprims base up to #x410000, to make it easier on the linker. 2006-03-11 04:29 gb * [r3669] : Lots-o-fixes; FFI stuff. When generating code involving a :REG8 operand, be careful to pass the 8-bit reg entry to LAP. 2006-03-11 04:27 gb * [r3668] : X86 CMOV instructions reference the source operand (possibly generating a memory fault) even if the destination operand isn't written to. Who knew ? 2006-03-11 04:24 gb * [r3667] : Recognize newer UUOs. 2006-03-11 04:24 gb * [r3666] : Use the #x80 bit in an instruction's REX prefix to indicate when the prefix needs to be emitted (even if the #x7 bits are clear). 2006-03-11 04:23 gb * [r3665] : A few UUOs; try to ensure that we'll generate (at least) a #x40 REX prefix if a new-style byte register is used as an operand. (e.g., we generally don't want to refer to the traditional things like %dh, and do want to refer to the newer things like %sil.) 2006-03-11 04:20 gb * [r3664] : rename *X86-NILREG-RELATIVE-SYMBOLS* -> *X86-NIL-RELATIVE-SYMBOLS*. 2006-03-11 04:20 gb * [r3663] : *** empty log message *** 2006-03-11 04:19 gb * [r3662] : %SINGLE-TO-DOUBLE, %DOUBLE-TO-SINGLE, to avoid consing in some simple cases. 2006-03-11 04:17 gb * [r3661] : Rename *PPC-NILREG-RELATIVE-SYMBOLS* -> *PPC-NIL-RELATIVE-SYMBOLS* (NIL hasn't been in a register in a few years.) 2006-03-11 04:15 gb * [r3660] : On 64-bit targets, SINGLE-FLOATs are EQL iff EQ. Try to recognize a few ways of coercing SINGLE-FLOATs to DOUBLE-FLOATs and vice versa, so that we have some chance of avoiding consing in some cases. 2006-03-11 04:13 gb * [r3659] : Add %SINGLE-TO-DOUBLE, %DOUBLE-TO-SINGLE. Replace unused EMBEDDED-CALL with SYSCALL. 2006-03-11 04:12 gb * [r3658] : Just use SYSCALL operator to mark x8664 syscalls. 2006-03-11 04:12 gb * [r3657] : Change patterns. 2006-03-07 14:20 gb * [r3656] : *** empty log message *** 2006-03-07 13:49 gb * [r3655] : Assume that fpzero (not fp0) contains 0.0. Get a few more things working, or at least written. 2006-03-07 13:49 gb * [r3654] : Assume that fpzero (not fp0) contains 0.0. 2006-03-07 13:48 gb * [r3653] : Define fpzero = xmm15. Pad the (misnamed) fpscr field in the TCR. 2006-03-06 14:20 gb * [r3652] : *** empty log message *** 2006-03-06 13:37 gb * [r3651] : Initial, crude support for cross-dumping "concatenated" lisp functions. 2006-03-06 13:36 gb * [r3650] : Fix comment indentation in a function that the x8664 backend was choking on. 2006-03-06 13:36 gb * [r3649] : use fpzero when a zero is needed, not fp0. 2006-03-06 13:35 gb * [r3648] : xmm15 is fpzero (contains 0.0[de]0). (xmm0 is used for ff-call.) 2006-03-06 13:35 gb * [r3647] : X862-COMPILE: don't reserve quad for null function name, even if debug info is present. 2006-03-06 13:34 gb * [r3646] : RELAX-FRAG-LIST: account for (possibly non-zero) length of :ALIGN frags. CROSS-CREATE-X86-FUNCTION: don't emit function name if it's nil, even if debug-info is present. 2006-03-06 13:30 gb * [r3645] : Add a declaration to suppress compiler warning on forward reference. 2006-03-06 11:20 gb * [r3644] : *** empty log message *** 2006-03-06 10:27 gb * [r3643] : Fix parens. 2006-03-06 10:20 gb * [r3642] : *** empty log message *** 2006-03-06 09:59 gb * [r3641] : Remove many uses of TARGET-ARCH-CASE (thus removing many uses of ppc/ppc32/ppc64 package symbols.) Some uses remain. 2006-03-06 09:57 gb * [r3640] : Define TARGET-ARCH-CASE in terms of CASE, not ECASE. 2006-03-06 02:20 gb * [r3639] : *** empty log message *** 2006-03-06 02:12 gb * [r3638] : typecode, lisptag are in ccl package. 2006-03-06 02:01 gb * [r3637] : Bump; some of this is hard to bootstrap. 2006-03-06 02:00 gb * [r3636] : Move some %kernel-global nonsense here. 2006-03-06 02:00 gb * [r3635] : Define macros in ... macros.lisp, of all places. 2006-03-06 01:59 gb * [r3634] : write some code in lisp (temporarily), clean up some typos so that the file compiles. 2006-03-06 01:58 gb * [r3633] : More parameters (mostly for xloader) in target-arch struct. 2006-03-06 01:52 gb * [r3632] : Get target parameters from (backend-target-arch *target-backend*)). 2006-03-05 13:20 gb * [r3631] : *** empty log message *** 2006-03-05 13:17 gb * [r3630] : Closer ... 2006-03-05 13:17 gb * [r3629] : Arch macros for hashing macros. 2006-03-05 13:16 gb * [r3628] : Arch macros for kernel globals. 2006-03-05 13:16 gb * [r3627] : Check *interrupt-level* value, not (interrupt-level). 2006-03-05 13:15 gb * [r3626] : Make *class-table* a global. 2006-03-05 13:15 gb * [r3625] : use temp2.b (= cl) for variable shifts. 2006-03-05 13:14 gb * [r3624] : No code-vectors on x8664. 2006-03-05 13:14 gb * [r3623] : Conditionalization: 64-bit-target implies no (need for) destructive single-float ops. 2006-03-05 13:13 gb * [r3622] : Remove some unused stuff. 2006-03-05 13:13 gb * [r3621] : array-element-type for x8664. 2006-03-05 13:12 gb * [r3620] : More stuff. 2006-03-05 13:12 gb * [r3619] : (Mostly) try to fix fixnum subtraction. 2006-03-05 13:11 gb * [r3618] : TRAP-UNLESS-FIXNUM, PUSH-ARGREGS, some syntax fixes. 2006-03-05 13:10 gb * [r3617] : More arch macros. 2006-03-05 13:10 gb * [r3616] : DECLARE-ARCH-SPECIFIC-MACRO returns macro name. 2006-03-05 13:09 gb * [r3615] : Warn for things that don't work yet; other things compile. 2006-03-05 13:08 gb * [r3614] : Compiles. 2006-03-05 13:08 gb * [r3613] : Fix typos/new syntax. 2006-03-05 02:20 gb * [r3612] : *** empty log message *** 2006-03-05 02:13 gb * [r3611] : Some conditionalization. 2006-03-05 02:12 gb * [r3610] : More stuff. Need to review float arithmetic. 2006-03-05 02:11 gb * [r3609] : Comment indentation. 2006-03-05 02:11 gb * [r3608] : Fix a few typos; needs more work. 2006-03-05 01:30 gb * [r3607] : Can use TARGET-WORD-SIZE-CASE instead of TARGET-ARCH-CASE. 2006-03-05 01:29 gb * [r3606] : New arch-specific macros. (Still need to do something about DEFCALLBACK and related functions.) 2006-03-05 01:29 gb * [r3605] : New arch-specific macros. 2006-03-05 01:28 gb * [r3604] : Keep unboxed float constants in the function (as aligned words following "real" instructions and preceding boxed constants.) 2006-03-05 01:26 gb * [r3603] : Remove some remaining uses of TARGET-ARCH-CASE by using arch-specific macro implementations. 2006-03-05 01:25 gb * [r3602] : Mechanism for arch-specific macros. 2006-03-05 01:24 gb * [r3601] : New constructor, so that we can keep track of defined target-archs. Hash table in each arch structure, for arch-dependent macros. 2006-03-04 03:20 gb * [r3600] : *** empty log message *** 2006-03-04 02:38 gb * [r3599] : Will need a lot more work. 2006-03-04 02:38 gb * [r3598] : xx8664xfasload. 2006-03-04 02:38 gb * [r3597] : *x8664-xload-modules*. 2006-03-04 02:37 gb * [r3596] : ignore lx64fsl files. 2006-03-04 02:37 gb * [r3595] : extra close paren. 2006-03-04 02:36 gb * [r3594] : package prefix on MXCSR-CONTROL-BYTE. 2006-03-04 02:36 gb * [r3593] : Don't call local variable "shift". 2006-03-04 02:20 gb * [r3592] : *** empty log message *** 2006-03-04 02:13 gb * [r3591] : Ignore lx64fsl files. 2006-03-04 02:11 gb * [r3590] : A few more things. 2006-03-04 02:11 gb * [r3589] : Conserve imm regs more in inc-ptr 2006-03-04 02:10 gb * [r3588] : Remove (duplicate, bad) definitions of int-to-double. 2006-03-04 02:10 gb * [r3587] : A few more UUOs. 2006-03-04 02:09 gb * [r3586] : Use (old, unused) ff-call operator to mark x8664 ff-calls. 2006-03-03 09:21 gb * [r3585] : *** empty log message *** 2006-03-03 09:09 gb * [r3584] : fix typo. 2006-03-03 09:09 gb * [r3583] : rcontext is %fs, and (with that change) tcr-bias is 0. 2006-03-03 09:08 gb * [r3582] : move xcompile-level-0, etc, away from here. 2006-03-03 09:07 gb * [r3581] : move xcompile-level-0, etc, here. 2006-03-03 09:06 gb * [r3580] : New file; still needs work. 2006-03-03 08:20 gb * [r3579] : *** empty log message *** 2006-03-03 08:15 gb * [r3578] : Um,function name is 'amd64_set_gsbase'. 2006-03-03 08:05 gb * [r3577] : Assert that we have thread-local storage by passing -DHAVE_TLS. 2006-03-03 08:04 gb * [r3576] : new_tcr sets up extra segment register if x8664 and HAVE_TLS. 2006-03-03 08:03 gb * [r3575] : tcr.single_float_convert is a struct; tag in most significant word. 2006-03-03 08:03 gb * [r3574] : tcr bias is 0. rcontext = %gs, at least for Linux/FreeBSD. 2006-03-03 06:20 gb * [r3573] : *** empty log message *** 2006-03-03 06:16 gb * [r3572] : add big-endian field to target-arch. 2006-03-03 06:16 gb * [r3571] : Big endian. 2006-03-03 06:15 gb * [r3570] : &lexpr stuff, mostly. 2006-03-03 06:14 gb * [r3569] : We're not big-endian. 2006-03-03 06:13 gb * [r3568] : Comment formatting. 2006-03-03 06:13 gb * [r3567] : Conditionalization changes. 2006-03-03 06:11 gb * [r3566] : Change thread creation/activation, so that new_tcr() call happens on the new thread. 2006-03-03 05:51 gb * [r3565] : Fix new_tcr() prototype. Why did this compile in FreeBSD/x86-64 ? 2006-03-03 05:28 gb * [r3564] : Depend on the linker script. Use some make variables to make changing/ ignoring that script easier. 2006-03-03 05:27 gb * [r3563] : Conditionalize some for FreeBSD. 2006-03-03 05:27 gb * [r3562] : FreeBSD can also map (at least) 512GB. Not sure how to get real name of executable in FreeBSd, so use argv[0] for now. CPUID instruction returns cache block size in ebx, not eax. 2006-03-03 05:25 gb * [r3561] : Output file wasn't executable if subprims were mapped to low memory; map them to a fixed address (0x410000). Will need to change subprims base in arch definition, and things (bad_funcall) that may depend on it. 2006-03-03 02:20 gb * [r3560] : *** empty log message *** 2006-03-03 02:19 gb * [r3559] : Try to avoid use of %temp0 (which may contain next_method_context); use %temp1 instead. 2006-03-03 02:18 gb * [r3558] : next_method_context is temp0, not temp1 (according to lisp). 2006-03-01 20:20 gb * [r3557] : *** empty log message *** 2006-03-01 19:42 gb * [r3556] : One lousy missing #\)! All this fuss over one lousy missing #\) ! 2006-03-01 19:39 gb * [r3555] : Define $fasl-clfun. 2006-03-01 19:38 gb * [r3554] : Rename %FIXNUM-SET-U32 -> %FIXNUM-SET-NATURAL. 2006-03-01 19:37 gb * [r3553] : Translate most of this, except for the lexpr stuff. 2006-03-01 19:37 gb * [r3552] : Delete a few things that seem to be totally unused. Rename %FIXNUM-SET-U32 -> %FIXNUM-SET-NATURAL. 2006-03-01 19:36 gb * [r3551] : $fasl-clfun. 2006-03-01 19:35 gb * [r3550] : %ilogbitp, inline multiplication. 2006-03-01 19:32 gb * [r3549] : %define-x86-lap-function: "function creator" isn't a parameter, so signature matches what the compiler expects. 2006-03-01 19:31 gb * [r3548] : Insert register operand correctly in immeditate bt/btc/btr/bts. Some imul variants (at least the 64/32/16-bit versions.) 2006-03-01 19:20 gb * [r3547] : *** empty log message *** 2006-03-01 19:19 gb * [r3546] : Don't require SPARC-LAP in nx-x86-lap-function ... 2006-03-01 03:20 gb * [r3545] : *** empty log message *** 2006-03-01 02:46 gb * [r3544] : Always build/ensure that a frame exists before calling a subprim. Handle &key, other/harder &optional cases. Todo: &lexpr, %ilogbitp, ?? 2006-03-01 02:44 gb * [r3543] : &optional defaulting and supplied-p handling; keywords subprims. 2006-03-01 02:43 gb * [r3542] : Signal IMPOSSIBLE-NUMBER conditions if parsing an potential number yields an ARITHMETIC-ERROR. 2006-03-01 02:43 gb * [r3541] : Define IMPOSSIBLE-NUMBER as a subtype of READER-ERROR. 2006-03-01 02:42 gb * [r3540] : If we'd get overflow or underflow when creating a float and the corresponding exception is enabled, generate the appropriate condition rather than CERRORing unconditionally. 2006-03-01 02:40 gb * [r3539] : Use target-word-size-case in %ilogbitp. 2006-03-01 02:37 gb * [r3538] : stkconslist, stkconslist_star, mkstackv, fix in heap_cons_rest_arg, simple_keywords, keyword_args, keyword_bind, fix in stack_cons_rest_arg, spreadargz, tfuncallgen, tfuncallslide,tfuncallvsp,tcallsymgen, tcallsymslide, tcallsymvsp, tcallnfngen, tcallnfnslide, tcallnfnvsp. todo: ffcall, syscall, callback, lexprs. 2006-03-01 02:33 gb * [r3537] : compare_reg_to_nil, discard_temp_frame. 2006-03-01 02:32 gb * [r3536] : function_header_offset, function_data_offset 2006-03-01 02:32 gb * [r3535] : TCR flags are a natural. 2006-03-01 02:31 gb * [r3534] : TCR flags are unsigned. 2006-03-01 02:31 gb * [r3533] : TCR flags. 2006-03-01 02:31 gb * [r3532] : Fix comment (have never used temp0 for closure data). 2006-02-27 21:20 gb * [r3531] : *** empty log message *** 2006-02-27 21:14 gb * [r3530] : Hadn't saved the buffer; ppc2-%aset2 changes were incomplete. 2006-02-27 07:20 gb * [r3529] : *** empty log message *** 2006-02-27 06:30 gb * [r3528] : Suppress compiler warning. 2006-02-27 06:06 gb * [r3527] : *** empty log message *** 2006-02-27 05:51 gb * [r3526] : Update. 2006-02-27 05:51 gb * [r3525] : Exercise more control over the stack sizes of listener threads (especially the initial listener.) (The --thread-stack-size argument affects the size of the initial listener's stacks only.) 2006-02-27 05:49 gb * [r3524] : Fix the usage message. 2006-02-27 05:20 gb * [r3523] : *** empty log message *** 2006-02-27 04:58 gb * [r3522] : Split --stack-size and --thread-stack-size args, to avoid rlimit problems (and because they're really separate issues.) 2006-02-27 04:23 gb * [r3521] : Remove (most) support for tracing callbacks. (Having the functionality isn't a bad idea, but the way that it was implemented made the code hard to read and maintain. Maybe redo this.) 2006-02-27 04:21 gb * [r3520] : Can't reserve 512GB on LinuxPPC64; try 256GB instead. 2006-02-27 04:20 gb * [r3519] : *** empty log message *** 2006-02-27 03:59 gb * [r3518] : Bump. 2006-02-27 03:38 gb * [r3517] : Restore COPT setting, clobbered on last (accidental) checkin. 2006-02-27 03:35 gb * [r3516] : atomic_and 2006-02-27 03:35 gb * [r3515] : more mappable memory on 64-bit platforms. Set stack-size global from command line arg 2006-02-27 03:35 gb * [r3514] : check_range() takes a headers_allowed arg (not allowed on vsp). Some debugging fprintfs. 2006-02-27 03:35 gb * [r3513] : bug fixes 2006-02-27 03:35 gb * [r3512] : simpler suspend ack, other changes, heavy, unreadable conditionalization 2006-02-27 03:35 gb * [r3511] : fixes in pc_luser_xp. Defer exceptions if we don't have the lock on Darwin 2006-02-27 03:35 gb * [r3510] : change defs of C write-barrier exports 2006-02-27 03:35 gb * [r3509] : tcr flags are natural 2006-02-27 03:35 gb * [r3508] : stack-size global 2006-02-27 03:35 gb * [r3507] : Prototype atomic_and(), atomic_ior() here. Use them to set/clear TCR flag bits. 2006-02-27 03:34 gb * [r3506] : Print trace level in trace output. 2006-02-27 03:34 gb * [r3505] : Add :reload-arguments arg to REBUILD-CCL. 2006-02-27 03:33 gb * [r3504] : Adjust default stack sizes based on --stack-size argument. 2006-02-27 03:32 gb * [r3503] : ROOM: don't report hons-space size if it's 0. 2006-02-27 03:32 gb * [r3502] : Use #+64-bit-target instead of #+ppc64-target in a few places. 2006-02-27 03:31 gb * [r3501] : Still needs work, but not quite so brain-dead. 2006-02-27 03:30 gb * [r3500] : A few bug fixes. 2006-02-27 03:30 gb * [r3499] : Fix typos in unwind-protect; use the same general scheme for PROGV. 2006-02-27 03:29 gb * [r3498] : In PPC2-%ASET2, typename is quoted in acode. 2006-02-27 03:28 gb * [r3497] : Replace BAD-CURRENT-VS global with STACK-SIZE. 2006-02-16 23:20 gb * [r3496] : *** empty log message *** 2006-02-16 22:49 gb * [r3495] : Wait for acknowledgement after suspend_tcr. Probably needs some testing, but this may be the right approach. 2006-02-16 22:48 gb * [r3494] : catch_exception_raise maintains TCR_FLAG_BIT_PENDING_EXCEPTION. (There's also some half-finished Mach-based code here for doing interrupts via suspend/resume, but it's probably a bad idea.) 2006-02-16 22:45 gb * [r3493] : Define a few more tcr flags. 2006-02-15 07:20 gb * [r3492] : *** empty log message *** 2006-02-15 06:56 gb * [r3491] : Define LWSYNC. Less sure how to tell if it's available. 2006-02-15 06:51 gb * [r3490] : Define GET-INTERNAL-REAL-TIME here; define GET-TICK-COUNT in terms of it (less likely to cons.) Define %TCR-INTERRUPT, so the lisp doesn't know how the guts of PROCESS-INTERRUPT are implemented (and so the kernel does.) 2006-02-15 06:49 gb * [r3489] : Define GET-INTERNAL-REAL-TIME earlier in the load sequence. 2006-02-15 06:48 gb * [r3488] : Replace unused condemn-area kernel global with raise-thread-interrupt. 2006-02-15 06:47 gb * [r3487] : Export raise_thread_interrupt. 2006-02-15 06:47 gb * [r3486] : Define raise_thread_interrupt. 2006-02-14 11:20 gb * [r3485] : *** empty log message *** 2006-02-14 10:23 gb * [r3484] : Try using Mach suspend/resume again. 2006-02-14 10:20 gb * [r3483] : *** empty log message *** 2006-02-14 10:20 gb * [r3482] : default &optionals, unwind-protect: push %ra0 "manually". 2006-02-14 10:17 gb * [r3481] : New inline REF-SYMBOL-VALUE: don't use a 64-bit trap on ppc32. 2006-02-14 06:20 gb * [r3480] : *** empty log message *** 2006-02-14 05:29 gb * [r3479] : Conditionalize references to subtag_code_vector in purify(),impurify(); code_vectors don't exist on X8664, and purify()/impurify() will have to do something different there. 2006-02-14 04:20 gb * [r3478] : *** empty log message *** 2006-02-14 04:01 gb * [r3477] : Closure & vcell stuff, some logical ops, tail funcall with no stack args. Funcall inline. 2006-02-14 03:58 gb * [r3476] : Closure & vcell stuff, some logical ops, tail funcall. 2006-02-14 03:57 gb * [r3475] : Add bad-funcall kernel global for x86 funcall. 2006-02-14 03:53 gb * [r3474] : Lots-o-stuff (misc_set, etc.) 2006-02-14 03:52 gb * [r3473] : Pad jump table to page boundary, so bad_funcall() has a known address. 2006-02-14 03:51 gb * [r3472] : new do_funcall(). 2006-02-14 03:51 gb * [r3471] : Define tag_single_float. 2006-02-14 03:50 gb * [r3470] : BAD_FUNCALL, for x86 funcall. 2006-02-14 03:50 gb * [r3469] : Setup BAD_FUNCALL, for x86 funcall. 2006-02-14 03:49 gb * [r3468] : Don't generate REX prefix on uuo_error_vector_bound unless we need one. 2006-02-13 22:20 gb * [r3467] : *** empty log message *** 2006-02-13 21:32 gb * [r3466] : last change broke purify on ppc too, back out of it 2006-02-13 07:20 gb * [r3465] : *** empty log message *** 2006-02-13 07:18 gb * [r3464] : More stuff. 2006-02-13 07:15 gb * [r3463] : use rcmp aliases for comparisons when testing for testing for other than equality. 2006-02-13 07:14 gb * [r3462] : rcmp aliases for all "cmp[bwlq]" instructions. 2006-02-13 07:13 gb * [r3461] : Always open an undo-frame for %stack-block, even if the "zero" flag is variable. 2006-02-13 07:11 gb * [r3460] : More stuff. Getting there ... 2006-02-13 07:11 gb * [r3459] : aligned_bignum_size 2006-02-13 07:10 gb * [r3458] : target fixnum limits. 2006-02-13 06:20 gb * [r3457] : *** empty log message *** 2006-02-13 06:08 gb * [r3456] : Fix %set-z-flag-if-u64-fits-in-fixnum. 2006-02-13 05:20 gb * [r3455] : *** empty log message *** 2006-02-13 04:42 gb * [r3454] : Every day, more stuff. 2006-02-13 04:41 gb * [r3453] : Add a few things, work around $ conflation. 2006-02-13 04:40 gb * [r3452] : add uuo_error_reg_not_fixnum. 2006-02-13 04:39 gb * [r3451] : define tstack_alloc_limit. 2006-02-12 23:20 gb * [r3450] : *** empty log message *** 2006-02-12 22:58 gb * [r3449] : NEW-NUMTOKEN: don't allow leading sign on denominator when trying to parse a ratio. 2006-02-12 03:20 gb * [r3448] : *** empty log message *** 2006-02-12 02:56 gb * [r3447] : start_lisp, toplevel_loop. 2006-02-12 02:55 gb * [r3446] : funcall does so. 2006-02-12 02:55 gb * [r3445] : Lose code_vector subtag; renumber things so that subtag_double_float has lower value. 2006-02-12 02:54 gb * [r3444] : Rewrite comment "clarifying" what toplevel_loop does. 2006-02-12 02:54 gb * [r3443] : define TCR states. 2006-02-12 02:53 gb * [r3442] : fix do_funcall(). 2006-02-12 02:53 gb * [r3441] : Don't reference subtag_code_vector. 2006-02-12 02:53 gb * [r3440] : Don't reference subtag_code_vector in purify/etc on x8664. 2006-02-12 02:52 gb * [r3439] : Setf inverses for unsigned memory accessors. 2006-02-12 02:51 gb * [r3438] : eval-redefs for unsigned memory setters. 2006-02-12 02:51 gb * [r3437] : Start to conditionalize for #+x8664-target. 2006-02-12 02:51 gb * [r3436] : REQUIRE-type primitives, more stuff, bug fixes. 2006-02-12 02:50 gb * [r3435] : No code-vector. Renumber things, so that SUBTAG-DOUBLE-FLOAT has a smaller value. 2006-02-12 02:50 gb * [r3434] : Use fp1 (arbitrarily), not fp0 when setting double-float vector elements. 2006-02-12 02:49 gb * [r3433] : Tab out further when printing instruction address. 2006-02-12 02:48 gb * [r3432] : Fix error trap in REQUIRE-REAL. 2006-02-12 02:48 gb * [r3431] : Memory-setting primitives come in signed and unsigned flavors. 2006-02-11 11:21 gb * [r3430] : *** empty log message *** 2006-02-11 10:34 gb * [r3429] : _SPmakes64: more svelte, less buggy. 2006-02-11 10:32 gb * [r3428] : Bump. 2006-02-11 10:32 gb * [r3427] : Lots-o-stuff: most vector access & assignment, etc. (Some cases need more testing, since PPC code this is based on assumed more imm temp registers and it's easy to run out when there are only 2 of them.) 2006-02-11 10:30 gb * [r3426] : Cache the foreign SP in %mm5. 2006-02-11 10:29 gb * [r3425] : Match use of dedicate shift-count register (%cl) via :%SHIFT. It's 2006. Why do processors have dedicated shift-count registers ? 2006-02-11 10:28 gb * [r3424] : Don't call UVREF to access value of single-float vector, since single-floats are immediate on 64-bit platforms. 2006-02-11 10:26 gb * [r3423] : CHAR-EQUAL. 2006-02-11 10:26 gb * [r3422] : ACODE-INTEGER-CONSTANT-P, which may have been a bad idea. 2006-02-10 21:20 gb * [r3421] : *** empty log message *** 2006-02-10 21:07 gb * [r3420] : More stuff, especially for vector reference. 2006-02-10 21:06 gb * [r3419] : Propagate some ppc backend fixes; get vref working. 2006-02-10 21:05 gb * [r3418] : setc is an alias for setb. 2006-02-10 21:04 gb * [r3417] : :SINGLE-FLOAT-VECTOR is a 32-bit ivector type. :SINGLE-FLOAT isn't. 2006-02-10 21:04 gb * [r3416] : :SINGLE-FLOAT-VECTOR is a 32-bit ivector type. 2006-02-10 07:20 gb * [r3415] : *** empty log message *** 2006-02-10 06:55 gb * [r3414] : Add a :FULL option to REBUILD-CCL. 2006-02-10 06:50 gb * [r3413] : Fix calculation of static_dnodes_for_area (was unmapping the wrong memory when dynamic area shrunk.) Maybe be verbose about GC activity. 2006-02-10 06:48 gb * [r3412] : Disable mach_suspend_tcr again. Mach sucks. 2006-02-10 06:47 gb * [r3411] : Define constant bits for GC verbosity. 2006-02-10 06:46 gb * [r3410] : For thing that actually macroexpand %SVSET and %SET-SCHAR, don't evaluate the new value inside the OPTIMIZE declarations. (COMPILE-FILE does this macroexpansion; COMPILE does its own handling of these SETF inverses.) 2006-02-10 06:44 gb * [r3409] : Export GC-VERBOSE, GC-VERBOSE-P. 2006-02-10 06:44 gb * [r3408] : Add GC-VERBOSE, GC-VERBOSE-P. 2006-02-10 06:43 gb * [r3407] : MISC-NODE-SET without EGC checks. Open-code special reference. 2006-02-10 06:42 gb * [r3406] : Avoid EGC write-barrier checks when storing some types of things in node vectors. Try to do special variable lookup inline (helps a little). 2006-02-10 06:40 gb * [r3405] : vcell, fcell refs are loop-weighted. 2006-02-10 06:38 gb * [r3404] : Most of throw, nthrow. 2006-02-06 06:20 gb * [r3403] : *** empty log message *** 2006-02-06 05:20 gb * [r3402] : suspend_other_threads() and resume_other_threads() take a Boolean "for_gc" arg. lisp_suspend_other_threads() and lisp_resume_other_threads() defined here, pass false to the underlying function. 2006-02-06 05:20 gb * [r3401] : *** empty log message *** 2006-02-06 05:19 gb * [r3400] : suspend_other_threads() and resume_other_threads() take a Boolean "for_gc" arg. Pass false here. 2006-02-06 05:18 gb * [r3399] : suspend_other_threads() and resume_other_threads() take a Boolean "for_gc" arg. Pass true here. 2006-02-06 05:18 gb * [r3398] : suspend_other_threads() and resume_other_threads() take a Boolean "for_gc" arg. 2006-02-06 05:17 gb * [r3397] : provide lisp_suspend_other_threads() and "lisp_resume_other_threads() to lisp. 2006-02-03 14:20 gb * [r3396] : *** empty log message *** 2006-02-03 14:03 gb * [r3395] : Mach exception lock is a pthread_mutex_t *; lock it around suspend/resume. Enable Mach suspend/suspend resume code, on the assumption that it (finally) works reliably. 2006-02-03 14:02 gb * [r3394] : Mach exception lock is a pthread_mutex_t *; don't lock it reentrantly. Define a "demux" routine for exception messages, though this probably wasn't necessary. allocptr_displacement needs to be signed_natural, not just an int (for ppc64.) Don't call the (fake) sigcontext a "linux sigcontext struct"; that's just confusing. (Rename variables "lss" -> "pseudosigcontext".) In mach_suspend_tcr(), use thread_abort_safely(), recognizing that it can have transient failures (including exception messages.) Keep trying until pending syscalls are aborted. Not all of this is really necessary; the old Mach lock_set either wasn't getting created correctly, or Mach lock_sets just don't work. Need to ensure that thread_suspension uses the same lock as the exception thread does. 2006-02-03 13:53 gb * [r3393] : More Mach exception stuff here. Mach exception lock is a pthread_mutex_t *. 2006-02-03 13:52 gb * [r3392] : Casts in vm_copy call. (Check these on ppc64 ?). 2006-02-01 13:20 gb * [r3391] : *** empty log message *** 2006-02-01 12:21 gb * [r3390] : Get tra right; tsp allocation/dnode_align stuff. 2006-02-01 12:20 gb * [r3389] : *** empty log message *** 2006-02-01 12:19 gb * [r3388] : comment format 2006-02-01 11:45 gb * [r3387] : On return from ff-call, ensure that allocptr/allocbase have harmless values before asserting TCR_STATE_LISP. 2006-02-01 00:20 gb * [r3386] : *** empty log message *** 2006-02-01 00:07 gb * [r3385] : Tried a few (conditionalized) things to make resume_tcr more reliable; currently, using the tcr->resume semaphore seems more reliable than sigsuspend when there are lots of active threads. 2006-02-01 00:05 gb * [r3384] : Zero_TSP_Frame and Zero_TSP_Frame_nz: don't zero the frame.type on PPC64. 2006-02-01 00:04 gb * [r3383] : Fix conditinal definition of ExceptionInformation 2006-01-31 22:20 gb * [r3382] : *** empty log message *** 2006-01-31 21:39 gb * [r3381] : remove extra blank line. 2006-01-31 21:38 gb * [r3380] : UNBOX-S64: get error right. 2006-01-31 21:37 gb * [r3379] : In PPC2-VREF, in the :SIGNED-BYTE-64 case where the index is constant, load into the temporary S64-REG, not RZERO! 2006-01-31 03:20 gb * [r3378] : *** empty log message *** 2006-01-31 02:48 gb * [r3377] : If using "modern" GAS, quote expressions in recursive macro calls. 2006-01-31 02:47 gb * [r3376] : __endfn undefines __func_name 2006-01-30 14:20 gb * [r3375] : *** empty log message *** 2006-01-30 13:59 gb * [r3374] : More stuff. 2006-01-30 13:59 gb * [r3373] : Move a comment. 2006-01-30 10:20 gb * [r3372] : *** empty log message *** 2006-01-30 09:39 gb * [r3371] : Conditionalize for FreeBSD/x86-64. 2006-01-30 09:39 gb * [r3370] : New files. FreeBSD make (and gmake) both want to rebuild all .s files, but this otherwise seems to work. 2006-01-30 04:20 gb * [r3369] : *** empty log message *** 2006-01-30 03:40 gb * [r3368] : May not even compile at the moment, but (a) my AMD64 laptop's disk is acting flaky and (b) a new AMD64 X2 dual-core box has a weird Windows-centric motherboard and I'm having trouble getting Linux running on it. Any questions ? 2006-01-28 23:20 gb * [r3367] : *** empty log message *** 2006-01-28 22:37 gb * [r3366] : Check in some stuff that's not ready yet; HW problems on x86-64 box. 2006-01-28 10:20 gb * [r3365] : *** empty log message *** 2006-01-28 09:52 gb * [r3364] : Some consing, stack-consing, multiple-value things. Try to limit the number of vinsns that actually do heap allocation, to make pc-lusering simpler (or possible ...). Assume that tsp & next-tsp are in registers. (MMX registers, but beggars can't be choosers.) 2006-01-28 09:49 gb * [r3363] : Don't enumerate/give pet names to MMX registers. TCR changes. 2006-01-28 09:38 gb * [r3362] : More stuff works; still need to do: - unwind-protect cleanup context stuff - hairy lambda-list stuff - more consing, especially stack-consing - ffi - lots of "simple" primitives, memory reference stuff. 2006-01-28 09:35 gb * [r3361] : No more * for indirect jmp (there's enough syntax already, without that.) VECTOR-LENGTH: use 2 args, more carefully. 2006-01-28 09:34 gb * [r3360] : Stop block on UUO. Handle more label-arithmetic cases, new UUO. 2006-01-28 09:33 gb * [r3359] : Simplify pseudo-op operands, too. 2006-01-28 09:33 gb * [r3358] : Fix mmx->mmx MOVQ. Add MOVAPD, a few more UUOs. 2006-01-28 09:31 gb * [r3357] : More binding stuff: unbinding, *interrupt-level* binding. 2006-01-28 09:30 gb * [r3356] : Arg order comment. 2006-01-28 09:29 gb * [r3355] : tsp, next_tsp in MMX regs. 2006-01-28 09:29 gb * [r3354] : interrupt_now() 2006-01-28 04:20 gb * [r3353] : *** empty log message *** 2006-01-28 03:52 gb * [r3352] : Only about an hour was wasted, but it somehow seems longer. 2006-01-28 03:39 gb * [r3351] : Splitting ppc-spentry.s into several smaller files files wasn't a bad idea, but the fact that Apple's 64-bit linker is badly broken (can't handle branches to external symbols, has difficulty with forward references, etc.) makes it impractical. (Yes, bugs in Apple's 64-bit linker have been reported, many months ago.) 2006-01-28 03:20 gb * [r3350] : *** empty log message *** 2006-01-28 03:05 gb * [r3349] : Split out ffi subprims to ppc-spffi.s 2006-01-28 02:49 gb * [r3348] : Move binding-related subprims to ppc-spbind.o 2006-01-28 02:48 gb * [r3347] : Add a comment. 2006-01-28 02:35 gb * [r3346] : ppc-spbuiltin.o 2006-01-28 02:32 gb * [r3345] : Oh, now I remember: there is no "subi.", which explains the use of "subic.". 2006-01-28 02:28 gb * [r3344] : Start to split up ppc-spentry.s into several smaller files. Will need makefile changes as well; won't compile cleanly for a while. 2006-01-28 02:27 gb * [r3343] : xuuo, other small changes. 2006-01-28 02:27 gb * [r3342] : Binding stuff. Will soon split up this file and PPC equivalent. 2006-01-28 02:26 gb * [r3341] : Cons arg order. More changes/reorganization to come. 2006-01-28 02:25 gb * [r3340] : Get registers consistent, again. 2006-01-27 11:20 gb * [r3339] : *** empty log message *** 2006-01-27 11:01 gb * [r3338] : Compile on x8664. 2006-01-27 10:20 gb * [r3337] : *** empty log message *** 2006-01-27 09:48 gb * [r3336] : Start to define some of what's in LAP. 2006-01-27 09:47 gb * [r3335] : call_closure. 2006-01-27 09:47 gb * [r3334] : Flesh out some more. 2006-01-27 09:47 gb * [r3333] : Add nargregs. 2006-01-27 07:20 gb * [r3332] : *** empty log message *** 2006-01-27 06:40 gb * [r3331] : Compile on x8664. 2006-01-27 06:40 gb * [r3330] : Use subi. in SPfitvals; no need for subic. 2006-01-27 06:39 gb * [r3329] : Use symbolic (GNU_SOURCE) names for register indices. 2006-01-27 06:38 gb * [r3328] : Lisp_frame, nargs extensions. 2006-01-27 06:37 gb * [r3327] : Change arg order in macros, to match LAP. Some fixes in do_funcall; need to update UUO defs. 2006-01-27 06:36 gb * [r3326] : Macro arg order, some multiple-value stuff, stub out unused things. 2006-01-27 06:35 gb * [r3325] : Flesh this out. 2006-01-27 01:20 gb * [r3324] : *** empty log message *** 2006-01-27 00:30 gb * [r3323] : Didn't I check this in already ? Fix CLRHASH locking. 2006-01-26 02:20 gb * [r3322] : *** empty log message *** 2006-01-26 01:53 gb * [r3321] : Ellipses to indicate that kernel build may take a little while, and delay after successful build to increase the chance that the message reporting success is seen before it scrolls off. 2006-01-26 01:51 gb * [r3320] : Update, describing (some) recent changes. 2006-01-26 01:20 gb * [r3319] : *** empty log message *** 2006-01-26 01:13 gb * [r3318] : Bump kernel abi version. 2006-01-26 01:13 gb * [r3317] : Progress messages around kernel build. 2006-01-26 01:03 gb * [r3316] : Finish bumping fasl version. 2006-01-26 00:57 gb * [r3315] : Next step in bumping FASL version. 2006-01-26 00:52 gb * [r3314] : Start bumping fasl version. 2006-01-26 00:45 gb * [r3313] : export allocation-quantum stuff. 2006-01-26 00:45 gb * [r3312] : new_heap_segment tries to use per-thread allocation quantum. 2006-01-26 00:44 gb * [r3311] : Some more allocation-quantum stuff. 2006-01-26 00:44 gb * [r3310] : %get-natural, %get-signed-natural. 2006-01-26 00:43 gb * [r3309] : Bump. 2006-01-25 22:20 gb * [r3308] : *** empty log message *** 2006-01-25 22:16 gb * [r3307] : Start to maintain per-process allocation quantum. Still more work to do here. 2006-01-25 22:14 gb * [r3306] : initialize tcr's log2_allocation_quantum from lisp_global(DEFAULT_ALLOCATION_QUANTUM) 2006-01-25 22:12 gb * [r3305] : set lisp_global(DEFAULT_ALLOCATION_QUANTUM) 2006-01-25 22:12 gb * [r3304] : OLDinterrupt-level -> log2-allocation-quantum. 2006-01-25 22:10 gb * [r3303] : OLDinterrupt-level -> default-allocation-quantum. 2006-01-25 22:09 gb * [r3302] : Constants again. 2006-01-25 10:20 gb * [r3301] : *** empty log message *** 2006-01-25 10:15 gb * [r3300] : Don't use SIZEOF_HEADERS, since this seems to cause confusion with ld 2.16.91/gcc 4.1.0 on FC5. Define _SDA_BASE_, since some versions of crt1.o reference it. (We don't use a small data area, so it'd be nice if crt1.o stopped referencing it. Perhaps there's an option that controls the C runtime startup file ?) 2006-01-25 10:10 gb * [r3299] : Add an instruction (a NOP) here. This seems to be necessary to force alignment to work correctly when the linker script omits SIZEOF_HEADERS. 2006-01-25 01:20 gb * [r3298] : *** empty log message *** 2006-01-25 00:45 gb * [r3297] : Several changes, not all of them finished. Make the primitive which does timed semaphore waits use milliseconds (not nanoseconds), and do the multiplication in the kernel. 2006-01-25 00:43 gb * [r3296] : A few more things. 2006-01-25 00:43 gb * [r3295] : Don't (canonically) call temp0.b "shift". 2006-01-24 21:20 gb * [r3294] : *** empty log message *** 2006-01-24 20:24 gb * [r3293] : In %nanosleep, check if -both- remaining seconds/remaining nanos are 0, not -either-. 2006-01-24 20:22 gb * [r3292] : Some FP stuff. 2006-01-24 20:20 gb * [r3291] : Kinder, gentler seconds-to-nanoseconds conversion in SLEEP. 2006-01-24 20:20 gb * [r3290] : *** empty log message *** 2006-01-24 20:18 gb * [r3289] : A few float-related (and 2-address vs 3-address -related) changes. 2006-01-24 20:17 gb * [r3288] : Go through tcr.single-float-convert when boxing/unboxing immediate SINGLE-FLOATs. 2006-01-24 20:17 gb * [r3287] : AVAILABLE-FP-TEMP: don't assume 32 bits in "available" mask. 2006-01-24 10:20 gb * [r3286] : *** empty log message *** 2006-01-24 09:32 gb * [r3285] : Change the values of those parameters. 2006-01-24 09:21 gb * [r3284] : *** empty log message *** 2006-01-24 09:13 gb * [r3283] : Make some bitfields used in fixnums which (sometimes) describe physical machine registers variables. This is a little tricky to bootstrap; making them variables (instead of constants) is a step towards making it easier to change their values without fouling up the running compiler. (They'll change back to constants after the "value" bitfield gets wide enough to accomodate interesting architectures. Not sure about the Itanium, but x8664 needs more bits for byte/word/long/quad reg variants and for mmx/xmm/x87/seg registers, and making this field wider seems better than other alternatives.) 2006-01-24 08:20 gb * [r3282] : *** empty log message *** 2006-01-24 08:04 gb * [r3281] : Constants for single-float tags. single-float-convert in the tcr, so we don't have to push and pop so much. 2006-01-23 09:20 gb * [r3280] : *** empty log message *** 2006-01-23 08:53 gb * [r3279] : bump. 2006-01-23 08:31 gb * [r3278] : Insert :%mmx registers. Insert them in the right place in MOVD/MOVQ. 2006-01-23 08:30 gb * [r3277] : Recognize :%mmx, :%xmm registers. 2006-01-23 08:29 gb * [r3276] : New (hard-wired) registers to recognize. op-indire: we're 64-bit. 2006-01-23 08:28 gb * [r3275] : Relaxation of :ALIGN frags : get diff right. 2006-01-23 08:27 gb * [r3274] : More stuff works. Need a less-ugly way of dealing with the fact that x8864 save regs, x8664 arg regs aren't assigned contiguous identifiers. 2006-01-23 08:26 gb * [r3273] : Make a few symbolic register names more canonical. 2006-01-23 08:25 gb * [r3272] : Fixnum addition, overflow handling, etc. 2006-01-23 08:24 gb * [r3271] : Move %CONS-FAKE-STACK-FRAME elsewhere. 2006-01-23 08:22 gb * [r3270] : TIMED-WAIT-ON-SEMAPHORE: use GET-INTERNAL-REAL-TIME, accept an optional semaphore-notification. 2006-01-23 08:21 gb * [r3269] : Move the %CONS-FAKE-STACK-FRAME macro here. When consing up a fake stack from from an exception context, stick the context in the frame's (new) xp slot. 2006-01-23 08:20 gb * [r3268] : *** empty log message *** 2006-01-23 08:20 gb * [r3267] : Warn that much of this is PPC-specific. When looking for a saved register value, look in the xp of any fake-stack-frame we encounter. 2006-01-23 08:18 gb * [r3266] : Add %fake-stack-frame-xp. (This causes some constants/accessors to be redefined; (COMPILE-CCL T) - and continuing through redefinitions - is suggested.) 2006-01-21 19:20 gb * [r3265] : *** empty log message *** 2006-01-21 18:48 gb * [r3264] : Fix LARGEST-PRIME-LESS-THAN-OR-EQUAL-TO in the case where the arg is a prime. In HONS, don't wander around looking for new honses in old tables. 2006-01-21 02:20 gb * [r3263] : *** empty log message *** 2006-01-21 01:57 gb * [r3262] : bump. 2006-01-21 01:56 gb * [r3261] : No more ERROR-HEADER. 2006-01-21 01:50 gb * [r3260] : Don't use ERROR-HEADER, since error messages are now formatted differently. 2006-01-21 01:49 gb * [r3259] : *MAIN-LISTENER-PROCESS-NAME* hasn't been correct in years, isn't used now. 2006-01-21 01:48 gb * [r3258] : Call %break-in-frame in response to a keyboard interrupt. 2006-01-21 01:47 gb * [r3257] : Some stack-walking changes, so error message contexts are often more meaningful. 2006-01-21 01:45 gb * [r3256] : Define FLATTEN-METHOD-LAMBDA-LIST, which removes initforms from the lambda list (making it more suitable for use as a gf lambda list). 2006-01-21 01:43 gb * [r3255] : Change error reporting in CHECK-DEFMETHOD-CONGRUENCY. 2006-01-21 01:41 gb * [r3254] : Lose extra comma. 2006-01-19 20:20 gb * [r3253] : *** empty log message *** 2006-01-19 20:13 gb * [r3252] : Check for null expt in PARSE-FLOAT. 2006-01-19 07:20 gb * [r3251] : *** empty log message *** 2006-01-19 06:31 gb * [r3250] : More stuff works. 2006-01-19 06:30 gb * [r3249] : Swap arg_z and temp0, on the theory that references to it may be able to avoid REX prefixes sometimes. 2006-01-19 06:29 gb * [r3248] : Get some more stuff working. 2006-01-19 06:29 gb * [r3247] : Use new UUOs for type - er for tag checking. 2006-01-19 06:28 gb * [r3246] : Get LAP constants in the right order. Indentation fixes, other subtle changes too hard to recognize. 2006-01-19 06:27 gb * [r3245] : Recognize new UUOs; print suffixes more often. 2006-01-19 06:26 gb * [r3244] : Several fixes, still more uuos. 2006-01-19 06:25 gb * [r3243] : Fix a constraint. 2006-01-19 06:24 gb * [r3242] : Use TARGET-WORD-SIZE-CASE, 'cause that's what we're testing for, 2006-01-19 06:23 gb * [r3241] : Introduce a :sets-cc vinsn attribute, which might (or might not) help to avoid some extra testing of condition codes. 2006-01-19 06:22 gb * [r3240] : Recognize s32 constants. (Note that this is for comparisons, and includes some non-fixnum cases.) 2006-01-19 06:20 gb * [r3239] : *** empty log message *** 2006-01-19 06:20 gb * [r3238] : A few horribly-named macros for register targeting, handling u8-targeted primitives (for tag checking.) 2006-01-19 04:20 gb * [r3237] : *** empty log message *** 2006-01-19 03:29 gb * [r3236] : Register renumbering. New consing/uuo scheme. (Looked at other strategies, including a spin-lock on a global heap with yield()ing. The per-thread memory pool scheme seems to win. Big. So we have to know how to interrupt it.) 2006-01-19 03:26 gb * [r3235] : Register renumbering. 2006-01-18 20:20 gb * [r3234] : *** empty log message *** 2006-01-18 20:07 gb * [r3233] : Add missing colon after label. 2006-01-18 12:20 gb * [r3232] : *** empty log message *** 2006-01-18 11:43 gb * [r3231] : some changes; may not compile 2006-01-17 12:20 gb * [r3230] : *** empty log message *** 2006-01-17 11:41 gb * [r3229] : More bumping; enough for today. Needs work, but I need sleep. 2006-01-17 11:40 gb * [r3228] : Lose target-arch-case. 2006-01-17 11:20 gb * [r3227] : *** empty log message *** 2006-01-17 11:04 gb * [r3226] : Use the new slots for type info. 2006-01-17 10:45 gb * [r3225] : Lose target-arch-case. 2006-01-17 10:44 gb * [r3224] : Backend vector type stuff here; lose target-arch-case. 2006-01-17 10:43 gb * [r3223] : More slots for arch-specific type info. 2006-01-17 10:42 gb * [r3222] : Can use target-word-size-case here. 2006-01-17 10:41 gb * [r3221] : Move vector typecode stuff to frontend and rename it. 2006-01-17 10:20 gb * [r3220] : *** empty log message *** 2006-01-17 09:56 gb * [r3219] : Bump; new binaries soon. 2006-01-17 09:55 gb * [r3218] : Start to use alternatives to TARGET-ARCH-CASE. 2006-01-17 09:54 gb * [r3217] : A little more. 2006-01-17 09:54 gb * [r3216] : set the new fields (unbound/slot-unbound tokens) 2006-01-17 09:29 gb * [r3215] : target-word-size-case 2006-01-17 09:28 gb * [r3214] : Add some fields to target-arch structure. 2006-01-17 08:20 gb * [r3213] : *** empty log message *** 2006-01-17 07:26 gb * [r3212] : Split things up to simplify vinsn expansion. 2006-01-17 07:25 gb * [r3211] : Vinsn template parsing. 2006-01-17 07:25 gb * [r3210] : Recognize more uuos, syntax changes. 2006-01-17 07:25 gb * [r3209] : Split some thing up for vinsn-expansion. Nuke branches to next instruction, which the compiler casually generates. 2006-01-17 07:22 gb * [r3208] : No "absolute" jump syntax (* ea). Use :^ in label-offset expressions. 2006-01-17 07:21 gb * [r3207] : Still several compilation warnings, but some things work. 2006-01-17 07:21 gb * [r3206] : Start to flesh out. Note use of (keyword) "overqualifiers" (tm). 2006-01-17 07:20 gb * [r3205] : *** empty log message *** 2006-01-17 07:19 gb * [r3204] : typo in EQL. 2006-01-17 07:18 gb * [r3203] : x86/x8664 compiler stuff. 2006-01-17 07:18 gb * [r3202] : add x8664-vinsns. ppc-reg was renamed long ago. 2006-01-17 07:17 gb * [r3201] : Exactly one cr field. 2006-01-17 07:16 gb * [r3200] : Use target::node-size (not 4!) in xp-argument-list. 2006-01-17 07:16 gb * [r3199] : node-size = word-size-in-bytes 2006-01-17 07:15 gb * [r3198] : rename s64-const -> s64const, since it's used now. 2006-01-13 17:20 gb * [r3197] : *** empty log message *** 2006-01-13 16:22 gb * [r3196] : $undo-x86-c-frame. 2006-01-13 16:21 gb * [r3195] : Add x862. 2006-01-13 16:21 gb * [r3194] : New syntax, where operand types are heavily qualified (the idea here is to make it possible to match opcodes exactly at %define-vinsn time.) 2006-01-13 16:20 gb * [r3193] : *** empty log message *** 2006-01-13 16:19 gb * [r3192] : Don't use #', since the functions are moving targets. 2006-01-13 16:18 gb * [r3191] : Well, it compiles (with lots of warnings) and needs some work ... 2006-01-13 16:17 gb * [r3190] : Define and use MATCH-TEMPLATE-TYPE. 2006-01-13 16:16 gb * [r3189] : Don't automatically label the first instruction (it gets labeled on backward branch.) 2006-01-13 16:15 gb * [r3188] : First cut at some vinsn-defining things. 2006-01-13 16:14 gb * [r3187] : some alu ops want :imm8s for 8-bit immediates, not :imm8. 2006-01-13 16:14 gb * [r3186] : condition-code constants. 2006-01-13 16:13 gb * [r3185] : Use renamed operators. 2006-01-13 16:13 gb * [r3184] : Fix some typos, I think .. 2006-01-13 16:12 gb * [r3183] : Rename some operators that had PPC in their names to distinguish themselves from the 68K versions, 10-12 years ago. 2006-01-12 19:20 gb * [r3182] : *** empty log message *** 2006-01-12 18:30 gb * [r3181] : ppc2-%i+,%i-,%ineg: when oveflow checking out-of-line, don't use <-, but copy arg_z to the (temporary) target register. When targeting "natural" args in poweropen ff-calls, account for the possibility that the target isn't what we'd allocated for it. 2006-01-12 09:20 gb * [r3180] : *** empty log message *** 2006-01-12 08:33 gb * [r3179] : Fix DEFINE-PPC-VINSN. 2006-01-12 08:32 gb * [r3178] : Bump again. Maybe a little difficult to bootstrap; new binaries soon if so. 2006-01-12 08:25 gb * [r3177] : Bump. Maybe a little difficult to bootstrap; new binaries soon if so. 2006-01-12 08:25 gb * [r3176] : Most (if not all) things that have been translated to x8664 assembler assemble (and, according to the dissassembler, do so correctly.) Maybe about 80% coverage; still a few hard cases. 2006-01-12 08:23 gb * [r3175] : VOID-ALLOCPTR looks like -16 on PPC64. 2006-01-12 08:22 gb * [r3174] : Reference new stub vinsn functions. 2006-01-12 08:21 gb * [r3173] : Stub vinsn functions. 2006-01-12 08:21 gb * [r3172] : DEFINE-X8664-VINSN. 2006-01-12 08:20 gb * [r3171] : *** empty log message *** 2006-01-12 08:19 gb * [r3170] : Change the calling sequence: no %nfn, new %ra0. 2006-01-12 08:18 gb * [r3169] : No explicit "@entry" label; ^-expressions implicitly reference it. 2006-01-12 08:16 gb * [r3168] : More bug fixes. 2006-01-12 08:12 gb * [r3167] : Use DEFINE-PPC-VINSN to define vinsns. 2006-01-12 08:11 gb * [r3166] : Define a DEFINE-VINSN function for PPC backends here. 2006-01-12 08:10 gb * [r3165] : Pass all args to the backend's define-vinsn function, including the backend 2006-01-12 07:55 gb * [r3164] : Remove ancient commented-out code. 2006-01-12 07:55 gb * [r3163] : %DEFINE-VINSN uses the backend's define-vinsn function. 2006-01-12 07:54 gb * [r3162] : BACKENDs now have a DEFINE-VINSN slot. 2006-01-11 04:20 gb * [r3161] : *** empty log message *** 2006-01-11 03:37 gb * [r3160] : testq matches :reg64 not :reg32. Imagine that. 2006-01-11 03:36 gb * [r3159] : Define (minimal) lisp-frame. 2006-01-11 03:36 gb * [r3158] : Moved some float stuff from x86-numbers to x86-float. Get most of this translated (except for %FIXNUM-GCD.) 2006-01-11 03:35 gb * [r3157] : Moved some float stuff from x86-numbers to x86-float. 2006-01-11 03:35 gb * [r3156] : Moved some float stuff from ppc-numbers to ppc-float. 2006-01-11 03:34 gb * [r3155] : Moved some trampoline stuff from x86-def to x86-clos. 2006-01-11 03:33 gb * [r3154] : Moved some trampoline stuff from ppc-def to ppc-clos. 2006-01-11 03:32 gb * [r3153] : x86-disassemble: source needed extension. 2006-01-11 03:32 gb * [r3152] : Indentation in NEEDS-COMPILE-P. 2006-01-10 13:20 gb * [r3151] : *** empty log message *** 2006-01-10 12:57 gb * [r3150] : more stack-frame stuff, %car/%cdr, constant refs, call/jump-symbol. 2006-01-10 12:50 gb * [r3149] : Use EVAL-WHEN, reorder things so that it compiles (still a warning or two, maybe.) Add UUO-STACK-OVERFLOW. 2006-01-10 12:50 gb * [r3148] : initial checkin 2006-01-10 12:46 gb * [r3147] : Recognize uuo-stack-overflow. When finding pc-relative refs in operands, the constant ref is a memory operand, not an immediate operand. 2006-01-10 12:38 gb * [r3146] : Rename a few registers. 2006-01-10 12:37 gb * [r3145] : x8664 LAP. 2006-01-10 12:37 gb * [r3144] : Add a 'q' suffix. 2006-01-09 13:20 gb * [r3143] : *** empty log message *** 2006-01-09 12:46 gb * [r3142] : Fixes: register-based shifts (#xd3 with different /n in the modrm byte) had the wrong table index. x8664-special (basically, movslq) table and handling of it was wrong. 2006-01-09 12:44 gb * [r3141] : Fix handling of index, scale when base register present in INSERT-MEMORY. More fixes to opcode templates, including movslq and mul. 2006-01-09 12:41 gb * [r3140] : Fix EVAL-WHEN stuff. 2006-01-09 12:29 gb * [r3139] : common for 32-bit targets 2006-01-09 12:27 gb * [r3138] : moved and renamed 2006-01-09 12:24 gb * [r3137] : common for 64-bit targets 2006-01-09 12:22 gb * [r3136] : renamed and moved 2006-01-09 12:20 gb * [r3135] : *** empty log message *** 2006-01-09 12:01 gb * [r3134] : new file 2006-01-09 11:53 gb * [r3133] : UNBOX-FIXNUM does arithmetic shift. 2006-01-09 11:52 gb * [r3132] : imm1.b is dl, not dx. 2006-01-09 08:20 gb * [r3131] : *** empty log message *** 2006-01-09 07:40 gb * [r3130] : Still some PPC code here. 2006-01-09 07:37 gb * [r3129] : No PPC code. No x8632 code, either. 2006-01-09 07:36 gb * [r3128] : Some conditional traps. CMP-TO-NIL for x8664 (where NIL has its own tag.) 2006-01-09 07:33 gb * [r3127] : Try to handle NIL, T in expressions. Move template-matching to x86-asm (and therefore x86::). 2006-01-09 07:31 gb * [r3126] : Handle UUOs. 2006-01-09 07:31 gb * [r3125] : More bug fixes. 2006-01-08 10:20 gb * [r3124] : *** empty log message *** 2006-01-08 09:46 gb * [r3123] : New file. No PPC or X8632 code here. 2006-01-08 09:45 gb * [r3122] : Don't use ^ for simple label references. 2006-01-08 09:44 gb * [r3121] : Use .pt and .pn for branch prediction. Don't use ^ syntax for simple label references. 2006-01-08 09:42 gb * [r3120] : Some CMOVxx templates put operands in the wrong places. 2006-01-08 09:41 gb * [r3119] : Compulsive comment formatting. 2006-01-08 03:20 gb * [r3118] : *** empty log message *** 2006-01-08 03:08 gb * [r3117] : No ppc code. (No 32-bit x86 code, either; will need some, eventually.) 2006-01-08 03:07 gb * [r3116] : conditionalize on target word size, which is what most/all of the #+ppc32-target/#+ppc64-target stuff was concerned with. (In turn, the word size is trying to say "are SINGLE-FLOATs immediate ?" and "do we have to drop into lap to deal with 32-bit words efficiently ?") 2006-01-08 03:05 gb * [r3115] : rcmp, to prevent (or delay the onset of) madness. Use .pt and .pn for static branch prediction. Add int-to-single. 2006-01-08 03:03 gb * [r3114] : Flesh out some more, fix some bugs. 2006-01-08 03:03 gb * [r3113] : Define MXCSR stuff. 2006-01-06 16:20 gb * [r3112] : *** empty log message *** 2006-01-06 15:32 gb * [r3111] : More of the new scheme. Still need to flesh out the opcode descriptions (mostly SSE* stuff and other newer instructions.) Should try to sanity check agains old (GAS) template data. Somehow. 2006-01-06 15:30 gb * [r3110] : Still some leftover code from the old scheme, but what's there of the new scheme seems to work. 2006-01-06 15:29 gb * [r3109] : Some bug fixes: op-imreg -really- isn't op-reg, sign-extend when getting a signed-byte 32 ... 2006-01-05 15:20 gb * [r3108] : *** empty log message *** 2006-01-05 14:55 gb * [r3107] : Remove 68k-MCL comment. 2006-01-05 14:54 gb * [r3106] : Start moving away from high-level "instruction templates" and towards more detailed "opcode templates". The latter are intended to make it easier to generate code from (e.g.) the compiler. (This is currently about half there. The assembler doesn't work at the moment, and the "opcode templates" are missing some things that might be used in compiled/lap code and their definitions may contain typos. It's worth doing, in that it'll allow the compiler to generate machine code while bypassing much of the old implementation's parse- time side effects and special-cases.) 2006-01-05 14:48 gb * [r3105] : Some stuff for entering/leaving functions. 2006-01-05 14:47 gb * [r3104] : %unreadable: don't assume that STREAM is referenced in format-string arg. 2006-01-05 14:46 gb * [r3103] : A little conditionalization. 2006-01-05 14:46 gb * [r3102] : A little more code and conditionalization. 2006-01-02 04:20 gb * [r3101] : *** empty log message *** 2006-01-02 03:23 gb * [r3100] : Bump. 2006-01-02 03:20 gb * [r3099] : *** empty log message *** 2006-01-02 03:20 gb * [r3098] : More changes. Enough for now. 2006-01-02 03:04 gb * [r3097] : Init hons space to #, don't make this appear unconditional. 2006-01-02 01:20 gb * [r3096] : *** empty log message *** 2006-01-02 00:28 gb * [r3095] : No cmpd here. 2006-01-02 00:20 gb * [r3094] : *** empty log message *** 2006-01-01 23:59 gb * [r3093] : Correct last comment: real change was revival of HONS-FREE-MARKER; it and HONS-DELETED-MARKER are macros. 2006-01-01 23:58 gb * [r3092] : Changed, yet again. Some intentionally bad hashing here. 2006-01-01 23:56 gb * [r3091] : _SPbuiltin_eql: as far as it goes, doesn't have to be conditionalized for ppc32/ppc64. Should really do it all here; the only (slightly) hard cases are ratio & complex. 2006-01-01 23:54 gb * [r3090] : bzero new static dnodes in grow_hons_area. Skip over them when counting immediate bytes before purify. (Need to do more skipping in purify/impurify.) 2006-01-01 23:52 gb * [r3089] : Changed, yet again. Some intentionally bad hashing here. 2005-12-31 21:20 gb * [r3088] : *** empty log message *** 2005-12-31 20:34 gb * [r3087] : Don't write back "for obscure GC reasons". 2005-12-31 19:20 gb * [r3086] : *** empty log message *** 2005-12-31 19:04 gb * [r3085] : Don't touch pages when committing them. 2005-12-31 19:04 gb * [r3084] : Don't write to new hons cells in grow_hons_area(); let application code fault pages in. 2005-12-31 19:03 gb * [r3083] : Don't use "free markers", but do lookup indices in the bitmap. 2005-12-31 19:02 gb * [r3082] : add :OPENMCL-HASH-CONSING to *FEATURES*. 2005-12-31 19:01 gb * [r3081] : New file. 2005-12-31 19:01 gb * [r3080] : Add TEST-BIT-AT-INDEX. 2005-12-31 15:20 gb * [r3079] : *** empty log message *** 2005-12-31 14:35 gb * [r3078] : Handle faults in touch_page(), by forcing it to return false. 2005-12-31 14:34 gb * [r3077] : try to touch_all_pages() after committing memory, to defeat overcommit. 2005-12-31 14:26 gb * [r3076] : touch_page. 2005-12-31 11:21 gb * [r3075] : *** empty log message *** 2005-12-31 11:12 gb * [r3074] : new file 2005-12-31 10:37 gb * [r3073] : Rename "rebuild-openmcl" to "rebuild-ccl" and export it. 2005-12-31 10:20 gb * [r3072] : *** empty log message *** 2005-12-31 10:08 gb * [r3071] : :linuxppc64 standard-boot-image-name is ppc-boot64, not ppc64-boot. 2005-12-31 09:55 gb * [r3070] : Lose the (old) metering_control. 2005-12-31 09:54 gb * [r3069] : Don't reference metering_control. 2005-12-31 09:37 gb * [r3068] : when shrinking HONS area, remember that delta_in_bytes is negative. 2005-12-31 09:37 gb * [r3067] : Lose old metering stuff. Don't enter debugger on mmap failure in grow_dynamic_area 2005-12-31 09:35 gb * [r3066] : Add a newline to an error message. Retain DELETED_STATIC_PAIRS in saved images. 2005-12-31 09:34 gb * [r3065] : Experimental REBUILD-CCL function. 2005-12-31 09:33 gb * [r3064] : double-float-abs!. Not sure what to do about NaNs, etc. 2005-12-31 09:32 gb * [r3063] : Keep bumping. 2005-12-31 09:32 gb * [r3062] : Try to bum the 32-bit EQUAL a bit. The ppc64 version needs work, too. 2005-12-31 09:30 gb * [r3061] : nargs is a 16-bit register on x86. 2005-12-31 09:30 gb * [r3060] : define log2_nbits_in_word for x8664. 2005-12-31 09:29 gb * [r3059] : AREA_CSTACK is uninteresting to GC on x86. no "nukable" pointers (to HONS space) in AREA_CSTACK on ppc, either. 2005-12-31 09:28 gb * [r3058] : Ignore *x86cl*. 2005-12-30 09:20 gb * [r3057] : *** empty log message *** 2005-12-30 09:11 gb * [r3056] : hons-area-size is now in openmcl-hons. 2005-12-30 08:20 gb * [r3055] : *** empty log message *** 2005-12-30 08:11 gb * [r3054] : Missing package prefix. 2005-12-30 08:11 gb * [r3053] : CVS mishap. 2005-12-30 08:10 gb * [r3052] : Missing package prefix. 2005-12-30 08:02 gb * [r3051] : Put things in their own package (OPENMCL-HONS:), flesh out documentation/doc strings a bit. (more). 2005-12-30 03:20 gb * [r3050] : *** empty log message *** 2005-12-30 02:44 gb * [r3049] : Make :Q work again. 2005-12-29 12:20 gb * [r3048] : *** empty log message *** 2005-12-29 11:48 gb * [r3047] : Deal with shrinking the hons area. 2005-12-29 11:48 gb * [r3046] : Handle cleanup of each tcr's gc_context in gc_like_from_xp, not in the GC proper. 2005-12-29 11:47 gb * [r3045] : shrink_dynamic_area by a natural, not just an unsigned. When creating the tenured area, move the static_used bitvector there from the active dynamic area. 2005-12-29 11:45 gb * [r3044] : Deal with the hons-space bitmap (saved in the image, after the dynamic section contents). It's an error if we're on the wrong platform. 2005-12-29 11:36 gb * [r3043] : Bump ABI_VERSION_MIN. 2005-12-29 11:34 gb * [r3042] : %USEDBYTES and ROOM account for hons space. 2005-12-29 00:20 gb * [r3041] : *** empty log message *** 2005-12-28 23:34 gb * [r3040] : write (target) platform info to header. TODO: endianness of host/target, page size assumptions. 2005-12-28 23:32 gb * [r3039] : rename $fasl-arch to $fasl-platform. 2005-12-28 23:31 gb * [r3038] : bump max/current abi version, min to soon follow. store target "platform" in header.flags, regardless of word size. check flags when loading; should be fatal error after bootstrapping. 2005-12-28 23:29 gb * [r3037] : use new PLATFORM constants to set kernel_global(HOST_PLATFORM). 2005-12-28 23:29 gb * [r3036] : New PLATFORM constants. 2005-12-28 23:28 gb * [r3035] : $fasl-arch -> $fasl-platform, get it from backend-target-platform. 2005-12-28 23:28 gb * [r3034] : Bump; new binaries soon. 2005-12-28 23:27 gb * [r3033] : Rename $fasl-arch to $fasl-platform. 2005-12-28 23:27 gb * [r3032] : Referring to area-staticlib is soooo 2005, but these constants don't need to be in the ppc package in 2006. 2005-12-28 23:26 gb * [r3031] : backend-target-architecture -> backend-target-platform; use new constants. 2005-12-28 23:25 gb * [r3030] : Fix package in attribute line. 2005-12-28 23:24 gb * [r3029] : Platform constants; backend-target-architecture -> backend-target-platform. 2005-12-28 17:20 gb * [r3028] : *** empty log message *** 2005-12-28 16:49 gb * [r3027] : Don't set the (was disk-size, now static-dnodes) field of section header to non-zero value. 2005-12-27 21:20 gb * [r3026] : *** empty log message *** 2005-12-27 21:19 gb * [r3025] : ADVISE-GLOBAL-DEF: don't generate DYNAMIC-EXTENT declarations, since they're too easy to violate. 2005-12-27 13:20 gb * [r3024] : *** empty log message *** 2005-12-27 12:29 gb * [r3023] : Bumped (yesterday ...). 2005-12-27 12:28 gb * [r3022] : hons-size stuff moved to HASH-CONS. 2005-12-27 12:26 gb * [r3021] : Absolutely wrong and useless, but at least it's checked in. 2005-12-27 12:25 gb * [r3020] : Use the 32-bit ROTATE-HASH-CODE on PPC64, for now. 2005-12-27 12:25 gb * [r3019] : Constants for bitvector operations. 2005-12-27 12:20 gb * [r3018] : Bitvector stuff (used exactly once.) 2005-12-27 12:20 gb * [r3017] : *** empty log message *** 2005-12-27 12:19 gb * [r3016] : Reformat a comment. 2005-12-27 12:18 gb * [r3015] : New file. 2005-12-27 12:18 gb * [r3014] : Add HASH-CONS. 2005-12-27 12:17 gb * [r3013] : Don't use disk_size field of section header. 2005-12-27 12:16 gb * [r3012] : replace "disk_size" field in section header with "static_dnodes". 2005-12-27 12:15 gb * [r3011] : Keep static_dnodes, static_used in tenured_area (never active), use static_dnodes_for_area(). forward_and_resolve_static_references(): only call (paraoia) if not EGC. Retain usedbits set from markbits. adjust_pointers_in_dynamic_area(): handles static dnodes if necessary. adjust_all_pointers: don't adjust area start/end, since grow_dynamic_area() does all that. Caller adjusts active area active field. Add an mremap() for Darwin. 2005-12-27 11:57 gb * [r3010] : hardlimit field is unused on heap areas. 2005-12-27 11:56 gb * [r3009] : box_fixnum, unbox_fixnum: signed_natural, not int. 2005-12-26 11:20 gb * [r3008] : *** empty log message *** 2005-12-26 10:39 gb * [r3007] : -D_GNU_SOURCE 2005-12-26 08:20 gb * [r3006] : *** empty log message *** 2005-12-26 08:18 gb * [r3005] : Lots of stuff of allocating hons area & adjusting other addresses. Tenured_area (and aliases) may now have static prefix, so be careful when relocating/forwarding (and use new "dynamic" GC pointers.) 2005-12-26 08:16 gb * [r3004] : More GC variables, for relocation of dynamic dnodes only. Extra param on some GC-like functions. Add GC_TRAP_FUNCTION_SET_HONS_AREA_SIZE 2005-12-26 08:15 gb * [r3003] : say "log2_page_size", not "12". 2005-12-26 08:14 gb * [r3002] : gc_like_from_xp functions take an extra signed_natural parameter. 2005-12-26 08:13 gb * [r3001] : Declare page_size, log2_page_size. 2005-12-26 08:12 gb * [r3000] : bad_go_tag_counter is now "deleted_static_pairs". 2005-12-26 08:12 gb * [r2999] : use "log2_page_size" instead of "12". Ahem. Map lots-o-memory on linuxppc64, too. 2005-12-26 08:10 gb * [r2998] : define log2_nbits_in_word. 2005-12-26 08:09 gb * [r2997] : define page_size, log2_page_size. 2005-12-26 08:08 gb * [r2996] : use "log2_page_size" instead of "12". 2005-12-26 08:07 gb * [r2995] : Use new name for "managed static" areas. New static_dnodes, static_used fields in areas. 2005-12-26 08:06 gb * [r2994] : Use new name for "managed static" areas. 2005-12-26 08:05 gb * [r2993] : Use symbolic constants for gc traps. Defin SET-CONS-AREA-SIZE. 2005-12-26 08:05 gb * [r2992] : Use symbolic constants for gc traps. 2005-12-26 08:04 gb * [r2991] : bad-go-tag-counter -> deleted-static-pairs, use "managed-static" area coded instead of "staticlib". (Define area codes, which were missing.) 2005-12-26 08:03 gb * [r2990] : area struct: it's been "dnodes" (not "dwords") for a while now. Add static-dnodes, static-used fields. 2005-12-26 08:01 gb * [r2989] : bad-go-tag-counter -> deleted-static-pairs, use "managed-static" area coded instead of "staticlib" 2005-12-26 07:54 gb * [r2988] : Define GC trap function constants, including new GC-TRAP-FUNCTION-SET-HONS-AREA-SIZE. 2005-12-24 02:20 gb * [r2987] : *** empty log message *** 2005-12-24 01:23 gb * [r2986] : Output of :FORM observes *BACKTRACE-PRINT-LEVEL* and *BACKTRACE-PRINT-LENGTH*. 2005-12-23 22:20 gb * [r2985] : *** empty log message *** 2005-12-23 21:50 gb * [r2984] : Add :form and :function backtrace commands. 2005-12-23 00:20 gb * [r2983] : *** empty log message *** 2005-12-22 23:58 gb * [r2982] : Try to show something that looks like a function call (with required arguments and an indication of other arguments) in backtrace. Use *BACKTRACE-PRINT-LEVEL* and *BACKTRACE-PRINT-LENGTH* to constrain printing of arguments and frame contents. 2005-12-22 18:20 gb * [r2981] : *** empty log message *** 2005-12-22 17:34 gb * [r2980] : Being compilable is good. 2005-12-22 17:20 gb * [r2979] : *** empty log message *** 2005-12-22 17:13 gb * [r2978] : Start trying to separate x86-64 from x86-32 stuff, a little. 2005-12-22 17:12 gb * [r2977] : "symbolic registers" map standard reg names (not x86-reg-entrys) to lisp reg names. 2005-12-22 17:11 gb * [r2976] : Tweak target features, target-arch id. 2005-12-22 17:10 gb * [r2975] : Update a little more. 2005-12-22 17:09 gb * [r2974] : new 2005-12-21 15:20 gb * [r2973] : *** empty log message *** 2005-12-21 14:38 gb * [r2972] : new files 2005-12-21 14:20 gb * [r2971] : *** empty log message *** 2005-12-21 14:01 gb * [r2970] : A few small changes, mostly to see if they'll disassemble. 2005-12-21 13:58 gb * [r2969] : Lose UNPARSE-OPERAND methods. EVAL-WHEN cleanup. 2005-12-21 13:57 gb * [r2968] : Don't assume an x86-target. 2005-12-21 13:57 gb * [r2967] : Lose the PRINT-OBJECT method, now that there's a mostly-working disassembler. 2005-12-21 13:56 gb * [r2966] : use EVAL-WHEN, so it at least compiles ... 2005-12-21 13:53 gb * [r2965] : Some X86 stuff. 2005-12-21 13:52 gb * [r2964] : Compulsive comment formatting. 2005-12-21 13:52 gb * [r2963] : *PPC-BACKEND* is nil if not on a PPC. 2005-12-20 11:20 gb * [r2962] : *** empty log message *** 2005-12-20 10:25 gb * [r2961] : Fix a typo. 2005-12-20 10:24 gb * [r2960] : Maintain symbolic register names, for disassembler. 2005-12-20 10:24 gb * [r2959] : New! Improved! Almost usable! 2005-12-17 10:20 gb * [r2958] : *** empty log message *** 2005-12-17 09:50 gb * [r2957] : Not quite finished, but not -too- far away. 2005-12-17 09:49 gb * [r2956] : Cloned the PPC version and renamed a few things; needs thought/work. 2005-12-17 09:46 gb * [r2955] : X86-SEGMENT-REGISTER. 2005-12-17 09:45 gb * [r2954] : LOOKUP-X86-REGISTER accepts strings (again.) 2005-12-16 09:20 gb * [r2953] : *** empty log message *** 2005-12-16 09:15 gb * [r2952] : Some changes to the TCR. 2005-12-16 08:48 gb * [r2951] : New file; might be more-or-less right. (Might want to make some of the bitmasks a bit smaller.) 2005-12-16 08:39 gb * [r2950] : Mostly just copied from PPC versions; just to exercise X86 LAP/LAPMACROS for now. 2005-12-16 08:37 gb * [r2949] : Registers can be referenced via small integer as well as by name. 2005-12-16 08:37 gb * [r2948] : First draft, may not even compile yet. 2005-12-16 08:36 gb * [r2947] : More lap macros. 2005-12-16 08:35 gb * [r2946] : Enough stuff to support cross-compiled DEFX86LAPFUNCTION. 2005-12-16 08:35 gb * [r2945] : Start defining some UUOs (int n, n=192-255). Define more registers in *x8664-register-entries*. 2005-12-14 08:20 gb * [r2944] : *** empty log message *** 2005-12-14 08:16 gb * [r2943] : Add :clozure to features (Hey! Marketing!) Other changes to *features*. 2005-12-14 08:15 gb * [r2942] : use _l suffix for 32-bit regs. nargs is 16 bits wide, not 32. Fix comment in tag_list definition. TCR contains linear_end field as well as linear field. 2005-12-14 08:12 gb * [r2941] : Try to identify the thread associated with each stack. 2005-12-14 08:11 gb * [r2940] : WAIT-ON-SEMAPHORE takes WHOSTATE arg. 2005-12-14 08:10 gb * [r2939] : Lots-o-changes. Move to the CCL package. 2005-12-14 08:10 gb * [r2938] : Lots-o-changes. 2005-12-14 08:09 gb * [r2937] : Kernel globals and nil-relative-symbols. 2005-12-14 08:08 gb * [r2936] : Copy process PSN to CS-AREA.OWNER of TCR, so that kernel backtrace can identify stacks by PSN. 2005-12-14 08:07 gb * [r2935] : gotta start somewhere. 2005-12-14 08:05 gb * [r2934] : Comment. 2005-12-14 08:04 gb * [r2933] : Indentation. 2005-12-13 09:20 gb * [r2932] : *** empty log message *** 2005-12-13 08:37 gb * [r2931] : New scheme for dealing with alignment/span-dependent instructions (looks suspiciously like GAS's). Lots of other changes and fixes, handle :ALIGN pseudoop (and :TRA, for tagged return addresses), and :long/:byte etc. A simple test case: ? (load "ccl:compiler;X86;x86-lap") ? (x86::x86-lap '((xchg (% rdi) (% rsi)) (lea (@ (- (^ foo) #xf) (% rsi)) (% rsi)) (jmp (* (% rdi))) (:tra foo))) should generate output that looks like: frag at #x8 04 00 00 00 00 00 00 00 48 87 FE 48 8D B6 15 00 00 00 FF E7 CC CC CC CC frag at #x20 EB FF FF FF CC CC CC CC frag at #x28 which seems correct. Still need to size immediates and emit them, and need syntax/support for lisp constants. 2005-12-13 08:20 gb * [r2930] : Invert the order of templates that used to use the :D bit, to match the order that GAS encounters them. (There are multiple ways to encode instructions that take a (register or memory address/register) and also (register/register or memory address) when both args are registers.) Revert to using short branches and short conditional branches, for new span-dependent-instruction scheme. Vector of seg-entries, ordered by reg-num. 2005-12-10 08:20 gb * [r2929] : *** empty log message *** 2005-12-10 07:49 gb * [r2928] : Constants delimited by +. Code emission starting to work; need scheme for span-dependent instructions (including relative displacements in memory operands), forcing alignment, etc. 2005-12-10 07:20 gb * [r2927] : *** empty log message *** 2005-12-10 06:51 gb * [r2926] : Constants delimited by +. Get rid of the :D modifier; define some templates redundantly (two-arg instructions that can take (reg, memory | reg) or (memory | reg, reg)). 2005-12-08 07:20 gb * [r2925] : *** empty log message *** 2005-12-08 06:22 gb * [r2924] : Start to think about code generation: code fragments, etc. Some things (modrm-byte/sib-byte generation) still unfinished (commented out), but want to get this checked in. 2005-12-08 06:21 gb * [r2923] : Define and use :label operand type. 2005-12-04 13:20 gb * [r2922] : *** empty log message *** 2005-12-04 12:45 gb * [r2921] : Update a little. 2005-12-04 12:42 gb * [r2920] : Time to bump. 2005-12-04 12:28 gb * [r2919] : In %FOREIGN-THREAD-INITIALIZE, don't assume that we're entered with no bindings in effect (the callback trampolines now bind a few specials, mostly having to do with tracing.) Likewise, any catch/unwind-protect frames that were made with no bindings should be fixed to point to the end of the phony binding chain that this function initializes. 2005-12-04 12:20 gb * [r2918] : *** empty log message *** 2005-12-04 12:14 gb * [r2917] : WORD_SIZE64 is never defined; WORD_SIZE is defined, and may have the value 64. (This was screwing up the 64-bit definition of the qnode type, and this screwing up 64-bit GC.) 2005-12-04 12:13 gb * [r2916] : Provide ANSI-style prototype for install_pmcl_exception_handlers(). 2005-12-04 12:12 gb * [r2915] : Fix definition of TCR_FLAG_BIT_PROPAGATE_EXCEPTION. 2005-12-04 12:12 gb * [r2914] : Don't set/test TCR_FLAG_BIT_ALT_SUSPEND, do set/test (1< unconditionally. 2005-11-15 09:32 gb * [r2882] : Depend on memory.o 2005-11-15 09:30 gb * [r2881] : Stubs for x86 exception stuff. 2005-11-15 09:29 gb * [r2880] : Migrate some stuff elsewhere; conditionalize a bit more. 2005-11-15 09:28 gb * [r2879] : Use CPUID to check x86 features, determine cache_block_size, etc. 2005-11-15 09:27 gb * [r2878] : get/put_vector_registers. Probably unused; referenced from the kernel imports table. 2005-11-15 09:25 gb * [r2877] : Conditionalize for x86, WORD_SIZE. 2005-11-15 09:24 gb * [r2876] : Move some code here from the (old) lisp-exceptions.c. 2005-11-15 09:24 gb * [r2875] : #define WORD_SIZE as 32 or 64, depending on preprocessor definitions. Stop using #ifdef WORD_SIZE64, here and elsewhere. Note that there are probably still many things that are conditionalized on #ifdef PPC64 when they're really concerned about 64/32-bit issues. (I've caught a few cases, but I'm sure that many still remain.) 2005-11-15 09:23 gb * [r2874] : Expect get/put_vector_registers (not necessarily AltiVec. (I don't think that this is called from Lisp code anymore.) 2005-11-15 09:22 gb * [r2873] : Use WORD_SIZE value to conditionalize. 2005-11-15 09:20 gb * [r2872] : Define IMAGE_BASE_ADDRESS for x86-64 Linux. 2005-11-15 09:20 gb * [r2871] : *** empty log message *** 2005-11-15 09:20 gb * [r2870] : Define a few externally referenced things. (Still a ways from being able to put code there, but the symbols need to be exported so that the kernel can link on x86-64 Linux. 2005-11-15 09:18 gb * [r2869] : Depend on xlbt.o, not plbt.o 2005-11-15 09:18 gb * [r2868] : Implement the kernel debugger's backtrace API, such as it is. (Still need to determine what the lisp stack looks like - and how it relates to the foreign stack - before any of this could work.) 2005-11-15 09:16 gb * [r2867] : #define WORD_SIZE as 32 or 64, depending on preprocessor definitions. Stop using #ifdef WORD_SIZE64, here and elsewhere. Note that there are probably still many things that are conditionalized on #ifdef PPC64 when they're really concerned about 64/32-bit issues. (I've caught a few cases, but I'm sure that many still remain.) 2005-11-15 09:14 gb * [r2866] : New file: functions related to memory allocation/zeroing/mapping/protecting/etc. (Most of this came from the old lisp-exceptions.c, but there are fairly generic memory-handling functions in other files that could reasonably migrate here.) Note that stuff having to do with lisp memory allocation (trapping into the kernel when the current thread's pool is too small, etc.) involves a lot of platform-dependent trap decoding and probably doesn't belong in this file. 2005-11-15 06:20 gb * [r2865] : *** empty log message *** 2005-11-15 06:15 gb * [r2864] : New (mostly empty) file. 2005-11-15 06:12 gb * [r2863] : PPC makefiles: rename lisp-exceptions.o to ppc-exceptions.o 2005-11-15 06:06 gb * [r2862] : renamed 2005-11-15 06:05 gb * [r2861] : rename 2005-11-15 06:04 gb * [r2860] : More cut-and-paste mishaps. 2005-11-15 05:59 gb * [r2859] : Fix (hopefully ...) botched PPC conditionalization. 2005-11-15 05:56 gb * [r2858] : Some X86 conditionalization. 2005-11-15 05:56 gb * [r2857] : Replace ppc_print.o with x86_print.o in DEBUGOBJ. 2005-11-15 05:55 gb * [r2856] : New file. 2005-11-14 11:20 gb * [r2855] : *** empty log message *** 2005-11-14 10:25 gb * [r2854] : CHEADERS include ppc-exceptions.h 2005-11-14 10:21 gb * [r2853] : siginfo_t debugging stuff. Only likely to be useful on Linux, or if signal-based exception handling's used. 2005-11-14 09:20 gb * [r2852] : *** empty log message *** 2005-11-14 09:10 gb * [r2851] : Start to make all debugger commands and lisp_Debugger() take a (possibly null) siginfo_t argument. Probably won't compile cleanly (yet). 2005-11-14 09:09 gb * [r2850] : Some X86[64] changes, mostly to the way that exception frames are accessed and in the ways that ivector sizes are determined. This compiles without warnings on X8664, but there need to be some (significant) semantic changes in the ways that the marker (especially the DWS marker) treats new types (TRAs, new symbol/function pointers.) 2005-11-14 09:03 gb * [r2849] : Split into platform-dependent subfiles. 2005-11-14 09:03 gb * [r2848] : Fix typos, design flaws ... 2005-11-14 09:01 gb * [r2847] : New files. 2005-11-14 09:00 gb * [r2846] : x86 variant of CS_OVERFLOW_FORCE_LIMIT 2005-11-14 04:20 gb * [r2845] : *** empty log message *** 2005-11-14 03:35 gb * [r2844] : Intrinsics are named __builtin_clz, __builtin_clzll. 2005-11-14 03:30 gb * [r2843] : Lots-o-changes, so some things are closer to compiling on x86-64. MAY HAVE BROKEN PPC/PPC64 COMPILATION; need to check changes in to check that. 2005-11-14 01:20 gb * [r2842] : *** empty log message *** 2005-11-14 00:50 gb * [r2841] : Bump. 2005-11-13 04:20 gb * [r2840] : *** empty log message *** 2005-11-13 03:35 gb * [r2839] : nuke this 2005-11-13 01:20 gb * [r2838] : *** empty log message *** 2005-11-13 01:17 gb * [r2837] : Tagging: complete, maybe usable. 2005-11-13 01:16 gb * [r2836] : A few more things, mostly just to check syntax. 2005-11-13 01:14 gb * [r2835] : Fix typo, tighten up a bit (cmov...). 2005-11-13 01:13 gb * [r2834] : _structf macro takes an optional displacement arg (defaults to -misc_bias). 2005-11-13 01:09 gb * [r2833] : update ppc-uuo.s dependency 2005-11-13 00:44 gb * [r2832] : Fix (several-days-old) typo in ppc64 code. 2005-11-13 00:20 gb * [r2831] : *** empty log message *** 2005-11-13 00:09 gb * [r2830] : Conditionalize-out (rather than comment-out) Steele's public-domain backquote implementation. It'd probably be worth switching to this (it's clearer and seems to work better); that might require some changes in the printer/pretty-printer to provide equivalent support, and probably requires some testing. 2005-11-13 00:06 gb * [r2829] : Dan Corkill's patch to allow readable printing of NaN's and infinities. (Bits of this were tweaked to require changes to GET-FPU-MODE; rebuild ppc-boot and build an image - or mouse on the new GET-FPU-MODE - before compiling these changes. 2005-11-13 00:02 gb * [r2828] : The Darwin PPC64 versions of XP-DOUBLE-FLOAT and XP-FPSCR-INFO were looking at a 32-bit machine context (and therefore returning garbage); the operands to arithmetic errors were usually reported incorrectly. Not yet sure of Linux PPC64. 2005-11-13 00:00 gb * [r2827] : GET-FPU-MODE takes an optional argument; if it's provided, it returns information about that attribute only. SET-FPU-MODE returns an unspecified non-NIL value, rather than calling GET-FPU-MODE with no arguments. Bootstrapping note: some higher-level code to be checked in depends on this incompatible change to GET-FPU-MODE. Do (xload-level-0) and create a new image before compiling those higher-level sources. 2005-11-10 07:20 gb * [r2826] : *** empty log message *** 2005-11-10 07:15 gb * [r2825] : Provide #, as a loadable extension. 2005-11-09 09:20 gb * [r2824] : *** empty log message *** 2005-11-09 08:24 gb * [r2823] : Detect (in a simple-minded, easily-fooled way ...) if the PPC version is running on something that doesn't appear to be a PPC. Limit the amount of reserved heap memory in that case (since an emulator seems to be in our way ...). Disable Mach exception handling (using POSIX signals instead) if we believe that we're running emulated. File bugs with Apple, because emulated POSIX signal handling's taken a step backward in that emulated environment. Not surprising. 2005-11-09 03:20 gb * [r2822] : *** empty log message *** 2005-11-09 02:43 gb * [r2821] : Flip %imm1 and %temp2: making %imm1 an alias for %rdx simplifies integer multiplication (which clobbers %rdx on this piece-o-crap.) Remove inconcistencies from tag assignments (so things assemble); still need to do a -real- tag assignment pass. 2005-11-09 02:41 gb * [r2820] : Add _builtin_function indices. 2005-11-09 02:40 gb * [r2819] : Try to make arg order of macros consistent with PPC (destination precedes source, registers precede memory operands, etc.) Some additions and fixes. 2005-11-09 02:38 gb * [r2818] : The little bit that's there assembles and looks reasonable. 2005-11-09 02:37 gb * [r2817] : Depend on x86-uuo.s (since "uuo.s" is -so- 11/07.) 2005-11-08 23:20 gb * [r2816] : *** empty log message *** 2005-11-08 22:53 gb * [r2815] : Use aligned_bignum_size() macro when allocating fixed-size bignums. (Some of the hardwired constants replaced by use of that macro may have been a bit too large, especially on PPC64.) Try to ensure that there's a one-to-one correspondence between the header used and the allocated size (e.g., in ppc32 _SPmakeu64, don't assume that 2-digit and 3-digit bignums have the same physical size.) 2005-11-08 22:47 gb * [r2814] : Hey! PPC changes! Define aligned_bignum_size macro, which returns the size in bytes needed for an N-digit bignum (including header and alignment padding.) 2005-11-08 12:20 gb * [r2813] : *** empty log message *** 2005-11-08 11:46 gb * [r2812] : Start to actually write some code; still need more infrastructure. On a 1 to 10 complexity scale, .SPbuiltin_plus is probably a 5 or 6; will need to be careful about register usage when consing bignums ('cause there aren't enough registers. Stop me if you've heard this one.) 2005-11-08 11:43 gb * [r2811] : Flesh out a bit. x86_64 tagging still needs to be fixed. 2005-11-08 11:40 gb * [r2810] : Include x86_uuo.s on X86. 2005-11-08 11:39 gb * [r2809] : Just getting started ... 2005-11-08 11:38 gb * [r2808] : Allow struct sizes to be implicitly padded. (Should maybe check if that's negative padding.) 2005-11-08 11:28 gb * [r2807] : Rename "uuo.s" to "ppc-uuo.s". 2005-11-08 11:26 gb * [r2806] : rename 2005-11-07 08:20 gb * [r2805] : *** empty log message *** 2005-11-07 08:15 gb * [r2804] : I might believe the register assignments here - at least provisionally - but the tagging will need work. (It's mostly pasted from ppc-constants64.s; the good news is that x86-64 isn't as constrained as ppc64 is.) Need to fit tagged return addresses into the scheme in some way that doesn't conflict with other primary tags. 4/12 ? Hybrid function objects ? Etc. Needs some thought. 2005-11-07 07:27 gb * [r2803] : Conditionally include x86 headers. 2005-11-07 07:24 gb * [r2802] : Hmm. x86 executables probably shouldn't have "ppc" in their names. Make up something else. 2005-11-07 04:20 gb * [r2801] : *** empty log message *** 2005-11-07 03:27 gb * [r2800] : 64-bit specific subprims files, since the 32-bit and 64-bit X86 are too dissimilar. 2005-11-07 03:26 gb * [r2799] : More new, empty files. 2005-11-07 03:20 gb * [r2798] : *** empty log message *** 2005-11-07 02:48 gb * [r2797] : x86-asmutils64.o 2005-11-07 02:47 gb * [r2796] : New file. If there's ever a 32-bit version, it'd likely be substantially different. (This is certainly different from the PPC version.) 2005-11-07 02:46 gb * [r2795] : New files; mostly empty, for now. 2005-11-07 02:39 gb * [r2794] : Tweak _emit_ELF_source_line_stab: use (m4 macro) __func_name instead of (as-defined) func_start; invert order of label and .stabn. This seems to generate good line number information on linuxx8664; need to check linuxppc, linuxppc64 as well. 2005-11-07 02:29 gb * [r2793] : Depend on ppc-macros.s, not macros.s. 2005-11-07 02:27 gb * [r2792] : Conditionally include for PPC, so far. 2005-11-07 02:20 gb * [r2791] : *** empty log message *** 2005-11-07 02:17 gb * [r2790] : renamed 2005-11-06 20:20 gb * [r2789] : *** empty log message *** 2005-11-06 19:21 gb * [r2788] : Fix other reference to asmutils.o in Makefiles. 2005-11-06 19:20 gb * [r2787] : *** empty log message *** 2005-11-06 19:19 gb * [r2786] : Fix other reference to "asmutils.o" in Makefiles. 2005-11-06 19:07 gb * [r2785] : Use .quad on X8664 as well. 2005-11-06 19:06 gb * [r2784] : Reference ppc-asmutils.[so] in PPC makefiles. 2005-11-06 19:02 gb * [r2783] : renamed 2005-11-06 07:20 gb * [r2782] : *** empty log message *** 2005-11-06 07:11 gb * [r2781] : artist's conception; still lots of PPC-specific code, still no x86-64-specific code ... 2005-11-06 07:05 gb * [r2780] : Update for platform-dependent SPOBJ. 2005-11-06 06:59 gb * [r2779] : Rename SPOBJ in Darwin makefiles. 2005-11-06 06:50 gb * [r2778] : rename 2005-11-06 05:20 gb * [r2777] : *** empty log message *** 2005-11-06 04:54 gb * [r2776] : Update dependencies on platform-dependent *-constants*.[hs] files. 2005-11-06 04:35 gb * [r2775] : Update dependencies on constants headers. 2005-11-06 04:32 gb * [r2774] : Inclde platform-dependent constants.s file. 2005-11-06 04:32 gb * [r2773] : rename 2005-11-06 04:30 gb * [r2772] : rename, include new files 2005-11-06 04:23 gb * [r2771] : Inclde platform-dependent constants.h file. 2005-11-06 04:22 gb * [r2770] : rename 2005-11-06 04:21 gb * [r2769] : Removed. 2005-11-06 04:20 gb * [r2768] : *** empty log message *** 2005-11-06 03:20 gb * [r2767] : *** empty log message *** 2005-11-06 02:34 gb * [r2766] : Try to get XCOMPILE-CCL back in synch. 2005-11-06 02:34 gb * [r2765] : Subprim name not constant in linuxppc64 MAKE-CALLBACK-TRAMPOLINE. (Conditionalize this a little more explicitly, though #+(and linux-target poweropen-target) is pretty much the same as #+(and linuxppc-target poweropen-target), since nothing else would use poweropen conventions.) 2005-11-06 02:20 gb * [r2764] : *** empty log message *** 2005-11-06 01:40 gb * [r2763] : Define subprimitive stuff in the backend. 2005-11-06 01:39 gb * [r2762] : Subprim names aren't constant anymore, so look them up at (compiler) runtime in the backend. Map between the names of builtin functions and the subprimitives that implement them via the backed. 2005-11-06 01:37 gb * [r2761] : *PPC-SUBPRIMS* is in the "PPC" package. 2005-11-06 01:33 gb * [r2760] : PPC subprim defs (common to ppc32 and ppc64) moved here. 2005-11-06 01:32 gb * [r2759] : Subprim names aren't constants anymore, so look them up when parsing vinsn templates. 2005-11-06 01:31 gb * [r2758] : Subprim defs are architecture and/or backend-specific. 2005-11-06 01:31 gb * [r2757] : PROVIDE's been back in the CL package for the last decade or so. 2005-11-06 01:30 gb * [r2756] : Bump to 051105. 2005-11-06 01:30 gb * [r2755] : Change the load order (subprims before arch stuff) a little. Load the target ppc??-arch explicitly, though some circular REQUIRES will load both ppc32-arch and ppc64-arch, currently. 2005-11-06 00:20 gb * [r2754] : *** empty log message *** 2005-11-05 23:23 gb * [r2753] : Start to isolate PPC subprims. 2005-11-05 23:20 gb * [r2752] : REVIVE-OBJC-CLASSES: don't be so dumb about reviving metaclass pointers (addresses may change.) 2005-11-05 17:20 gb * [r2751] : *** empty log message *** 2005-11-05 16:22 gb * [r2750] : Define PPC (via -DPPC) explicitly. 2005-11-05 16:20 gb * [r2749] : *** empty log message *** 2005-11-05 15:45 gb * [r2748] : directory rename 2005-11-05 15:33 gb * [r2747] : directory rename 2005-11-05 15:22 gb * [r2746] : directory renaming 2005-11-05 15:20 gb * [r2745] : *** empty log message *** 2005-11-05 15:14 gb * [r2744] : finish rename darwin -> darwinppc 2005-11-05 15:12 gb * [r2743] : more directory renaming 2005-11-05 15:07 gb * [r2742] : rename darwin directory to darwinppc 2005-11-05 15:05 gb * [r2741] : Awkward CVS directory rename 2005-11-05 15:05 gb * [r2740] : Awkward CVS directory rename. 2005-11-04 19:20 gb * [r2739] : *** empty log message *** 2005-11-04 18:34 gb * [r2738] : In #+ppc32 version of %%SET-SIGNED-LONGLONG, trap if the pointer's not a PPC32::SUBTAG-MACPTR (not a PPC64::SUBTAG-MACPTR). Sheesh. 2005-10-31 01:20 gb * [r2737] : *** empty log message *** 2005-10-31 00:58 gb * [r2736] : Reason to suspect that mach_suspend_tcr is still unreliable. 2005-10-30 07:20 gb * [r2735] : *** empty log message *** 2005-10-30 06:28 gb * [r2734] : Lots of conditionalization assumed that :MCL implies :OPENMCL. 2005-10-29 22:20 gb * [r2733] : *** empty log message *** 2005-10-29 21:58 gb * [r2732] : Bump to 051029b. 2005-10-29 21:57 gb * [r2731] : Remove :CCL*, :MCL, etc from *FEATURES*. 2005-10-29 21:57 gb * [r2730] : Change order of tests in %HASH-PROBE, so that test functions other than EQUAL/EQUALP don't have to compare keys to missing/deleted tokens. 2005-10-29 21:20 gb * [r2729] : *** empty log message *** 2005-10-29 20:40 gb * [r2728] : Confess to the day's problems ... 2005-10-29 20:39 gb * [r2727] : Bump. 2005-10-29 20:39 gb * [r2726] : Fix %RESTORE-OUTERMOST-BINDING. 2005-10-29 20:39 gb * [r2725] : The ASH compiler macro made 32-bit assumptions and doesn't do anything that the compiler doesn't do by itself. The LSH compiler macro made those same assumptions. LSH isn't really used enough to justify having a compiler macro, much less a buggy one ... 2005-10-28 13:20 gb * [r2724] : *** empty log message *** 2005-10-28 12:33 gb * [r2723] : PPC2-SYMBOL-VALUE-CELL calls ENSURE-BINDING-INDEX. 2005-10-28 12:20 gb * [r2722] : *** empty log message *** 2005-10-28 11:23 gb * [r2721] : Bump. 2005-10-28 11:23 gb * [r2720] : Yesterday's images didn't work on ppc32 on non-G5s. 2005-10-28 11:20 gb * [r2719] : *** empty log message *** 2005-10-28 11:17 gb * [r2718] : Don't use 64-bit instructions (tdgti) in a 32-bit lisp. 2005-10-27 09:20 gb * [r2717] : *** empty log message *** 2005-10-27 08:22 gb * [r2716] : Say something about LinuxPPC64 and about some bug fixes. 2005-10-27 07:20 gb * [r2715] : *** empty log message *** 2005-10-27 07:18 gb * [r2714] : Bump; claim to be a "pre-1.1" release. 2005-10-27 07:17 gb * [r2713] : Fix some typos. 2005-10-27 06:20 gb * [r2712] : *** empty log message *** 2005-10-27 06:19 gb * [r2711] : DEFINE-MODIFY-MACRO: don't try to special-case "place" being a simple (non-symbol-macro) symbol and expand into SETQ, since we don't know anything about side-effects in other arguments in the lambda list. INCF & DECF: -do- special case simple symbols as places, if the delta is also either a constant or a simple symbol; do the more general expansion otherwise (Bryan's code.) 2005-10-27 01:20 gb * [r2710] : *** empty log message *** 2005-10-27 00:40 gb * [r2709] : Some notes; hopefully, maintaining this file between releases will make it easier/quicker to prepare real release notes. 2005-10-27 00:20 gb * [r2708] : *** empty log message *** 2005-10-26 23:53 gb * [r2707] : maybe questionable INCF/fixnum stuff 2005-10-26 23:43 gb * [r2706] : avoid INCF/fixnum-arithmetic-sleaze issues in CDB-HASH. 2005-10-26 23:35 gb * [r2705] : Load l1-typesys & sysutils earlier, to make more of the type system available. 2005-10-26 23:33 gb * [r2704] : PROCESS-TERMINATION-SEMAPHORE stuff. 2005-10-26 21:20 gb * [r2703] : *** empty log message *** 2005-10-26 20:51 gb * [r2702] : Maintain tcr.unwinding during nthrow. 2005-10-26 20:50 gb * [r2701] : interrupt_handler() : defer interrupts if tcr.unwinding is set. 2005-10-26 20:50 gb * [r2700] : Replace the (unused) tcr.suspend-total-on-exception-entry field with tcr.unwinding. 2005-10-26 20:20 gb * [r2699] : *** empty log message *** 2005-10-26 19:59 gb * [r2698] : References to *INTERRUPT-LEVEL* are inlined. UNWIND-PROTECT plays with *INTERRUPT-LEVEL*, so that cleanup forms run with interrupts disabled. New scheme for special payback, so that restration of *INTERRUPT-LEVEL* polls for deferred interrupts. Remove some old (unused) interrupt handling stuff. 2005-10-26 19:51 gb * [r2697] : Enable interrupts before calling the first user toplevel function (not in RESTORE-LISP-POINTERS' cleanup form, since that's now ineffectual.) 2005-10-26 19:50 gb * [r2696] : Define & export WITH-INTERRUPTS-ENABLED. 2005-10-26 19:49 gb * [r2695] : Minor indentation change. 2005-10-26 19:48 gb * [r2694] : No such thing as :ADDRESS in vreg specs. Get words size right in REF-INTERRUPT-LEVEL. 2005-10-26 19:47 gb * [r2693] : No such thing as :ADDRESS in vreg specs. 2005-10-26 19:46 gb * [r2692] : Remove a few old interrupt-handling things. 2005-10-26 12:20 gb * [r2691] : *** empty log message *** 2005-10-26 11:51 gb * [r2690] : Define WITHOUT-INTERRUPTS and friends in terms binding CCL::*INTERRUPT-LEVEL*. 2005-10-26 11:20 gb * [r2689] : *** empty log message *** 2005-10-26 10:44 gb * [r2688] : Handling binding/unbinding of *interrupt-level* specially. 2005-10-26 10:43 gb * [r2687] : *INTERRUPT-LEVEL* binding, unbinding, and reference. 2005-10-26 10:43 gb * [r2686] : $undointerruptlevel. 2005-10-26 10:41 gb * [r2685] : TCR_INTERRUPT_LEVEL is signed ... 2005-10-26 09:20 gb * [r2684] : *** empty log message *** 2005-10-26 09:09 gb * [r2683] : Remove svar subprims; identify unused subprims; add support for binding/ unbinding *interrupt-level* and associated polling for pending interrupts. 2005-10-26 09:05 gb * [r2682] : Disavow all knowledge of svars. 2005-10-25 18:20 gb * [r2681] : *** empty log message *** 2005-10-25 17:23 gb * [r2680] : EVENT-POLL vinsn checks for pending interrupts. 2005-10-25 17:20 gb * [r2679] : *** empty log message *** 2005-10-25 17:02 gb * [r2678] : Arrgh. (Duplicate binding index #1, sometimes.) 2005-10-25 11:20 gb * [r2677] : *** empty log message *** 2005-10-25 11:04 gb * [r2676] : replace references to tcr.interrupt level with (effectively) thread-specific binding of CCL::*INTERRUPT-LEVEL* on the lisp side, as well. 2005-10-25 10:20 gb * [r2675] : *** empty log message *** 2005-10-25 09:40 gb * [r2674] : deprecate (and stop using, at least in the kernel) tcr.interrupt_level; use tcr->tlb_pointer[INTERRUPT_LEVEL_BINDING_INDEX] instead (which just happens to be the dynamic value of CCL::*INTERRUPT-LEVEL*.) 2005-10-25 09:20 gb * [r2673] : *** empty log message *** 2005-10-25 09:03 gb * [r2672] : Finish bumping FASL version. 2005-10-25 08:44 gb * [r2671] : tlb[INTERRUPT_LEVEL_BINDING_INDEX] == fixnum -1 in each new tcr. 2005-10-25 08:42 gb * [r2670] : INTERRUPT_LEVEL_BINDING_INDEX == (symbol-binding-index ccl::*interrupt-level*). (This means that the end of this bootstrapping cycle is in sight; perhaps after some food ...) 2005-10-25 08:39 gb * [r2669] : QUIT needs to force-output on ccl::*stdout*, as many people have mentioned. 2005-10-25 08:38 gb * [r2668] : Reserved binding index scheme, finally. 2005-10-25 07:20 gb * [r2667] : *** empty log message *** 2005-10-25 06:51 gb * [r2666] : Remove SVAR type/class. Use BINDING-INDEX-SYMBOL instead of INDEX-SVAR. 2005-10-25 06:20 gb * [r2665] : *** empty log message *** 2005-10-25 05:49 gb * [r2664] : Don't use svars in %RESET-OUTERMOST-BINDING (which looks wrong, btw.) 2005-10-25 05:48 gb * [r2663] : Don't use svars in binding, special reference/assignment. 2005-10-25 05:20 gb * [r2662] : *** empty log message *** 2005-10-25 05:04 gb * [r2661] : No more SVAR-CHECK-SYMBOL-LIST. 2005-10-25 04:52 gb * [r2660] : PROGVSAVE vinsn uses non-svar subprim. 2005-10-25 04:51 gb * [r2659] : Was out of synch with spjump.s. 2005-10-25 04:51 gb * [r2658] : Don't use svars in PROGV. 2005-10-25 04:20 gb * [r2657] : *** empty log message *** 2005-10-25 04:10 gb * [r2656] : When setting xload binding addresses, set symbol.binding-index cell (not svar.idx-cell ...) 2005-10-25 04:09 gb * [r2655] : Don't use svars at runtime. 2005-10-25 04:06 gb * [r2654] : Keep track of binding indices assigned in the cold load. (Could do this another way ...) 2005-10-25 04:05 gb * [r2653] : CHECK-SYMBOL-LIST (used in PROGV) calls ENSURE-BINDING-INDEX. Don't use svars in %SYM-VALUE, etc. Keep binding indices and svar indices in synch. Add SYMBOL-BINDING-INDEX accessor, for debugging. 2005-10-24 13:20 gb * [r2652] : *** empty log message *** 2005-10-24 13:00 gb * [r2651] : No need to clean up symbol.package-plist, which no longer exists. 2005-10-24 12:59 gb * [r2650] : New (non-svar) subprims for special binding/access. 2005-10-24 12:58 gb * [r2649] : New fasl ops: intern, create symbols and update their binding indices. 2005-10-24 12:58 gb * [r2648] : New binding-index stuff. May need to keep in synch with svar indices while weaning ourselves of svars; may need to do that weaning cold turkey. 2005-10-24 12:54 gb * [r2647] : .SPSVAR-UNBIND-TO is now just .SPUNBIND-TO. 2005-10-24 12:53 gb * [r2646] : SVAR-DPAYBACK is now just DPAYBACK. 2005-10-24 12:53 gb * [r2645] : Rename some unbinding vinsns, use new subprim names. 2005-10-24 12:52 gb * [r2644] : New subprims, to phase out svar-based special binding/access. (Note that unbinding works the same way for svars/symbol-binding-indices; svar-based unbinding routines were just renamed.) 2005-10-23 11:20 gb * [r2643] : *** empty log message *** 2005-10-23 10:44 gb * [r2642] : New low-level plist stuff. 2005-10-23 10:44 gb * [r2641] : Accept a slightly newer fasl version (not finished bootstrapping yet.) 2005-10-23 10:43 gb * [r2640] : Depend on constants64.? 2005-10-23 10:42 gb * [r2639] : Move symbol-plist to new symbol slot; simplify access to symbol package, type-predicate. 2005-10-23 10:41 gb * [r2638] : Remove old symbol.package-plist constants. 2005-10-23 10:40 gb * [r2637] : Open-code %SYMBOL->SYMPTR. 2005-10-23 08:20 gb * [r2636] : *** empty log message *** 2005-10-23 07:42 gb * [r2635] : New symbol slots on ppc64, too. 2005-10-23 07:41 gb * [r2634] : Remove the crude bootstrapping hack (at least the most recent one ...) 2005-10-23 06:20 gb * [r2633] : *** empty log message *** 2005-10-23 05:49 gb * [r2632] : Start new symbol stuff. More to be done here. 2005-10-23 05:49 gb * [r2631] : ABI version change. 2005-10-23 05:49 gb * [r2630] : Depend on constants32.? 2005-10-23 05:49 gb * [r2629] : Symbol package isn't where it used to be. 2005-10-23 05:48 gb * [r2628] : New "abi" version, for symbol size/format changes. 2005-10-23 05:47 gb * [r2627] : Anticipate the symbol-package/plist breakup. 2005-10-23 05:47 gb * [r2626] : New symbol slots. 2005-10-23 05:46 gb * [r2625] : Add a very ugly bootstrapping hack, and a warning reminding that it should be removed. 2005-10-23 05:45 gb * [r2624] : Initialize new symbol slots in MAKE-SYMBOL. 2005-10-23 05:44 gb * [r2623] : Add a couple of new fields to symbols; this is intended to help get rid of svar objects for thread-specific special bindings, and to make accessing a symbol's plist a little faster/simpler. Changing the size of a symbol introduces some bootstrapping issues; the bleeding-edge sources may not build until that bootstrapping process is complete (on all platforms); there will be new binaries then (probably sometime on October 23.) 2005-10-23 01:20 gb * [r2622] : *** empty log message *** 2005-10-23 00:58 gb * [r2621] : newblocktag/newgotag have been obsolete for a few years now. 2005-10-23 00:57 gb * [r2620] : A little kernel-global cleanup. 2005-10-23 00:20 gb * [r2619] : *** empty log message *** 2005-10-23 00:17 gb * [r2618] : Add UNWIND-PROTECT-DISABLE-INTERRUPTS-DURING-CLEANUP, at least for the time being ... 2005-10-23 00:15 gb * [r2617] : An EOF in the REPL isn't "transient" if --batch is in effect. 2005-10-23 00:15 gb * [r2616] : Still a ways from reality, but a little closer. 2005-10-23 00:14 gb * [r2615] : go-tag-counter & block-tag counter have been obsolete for a few years. 2005-10-18 23:20 gb * [r2614] : *** empty log message *** 2005-10-18 23:03 gb * [r2613] : THREAD-INTERRUPT: checked in unsaved buffer, bad version always returned NIL. 2005-10-18 05:20 gb * [r2612] : *** empty log message *** 2005-10-18 04:52 gb * [r2611] : In xNewThread, try to detect stack/thread initialization failures. SEM_WAIT_FOREVER for new thread activation. 2005-10-18 04:51 gb * [r2610] : Don't maintain tcr.suspend_total_on_exception_entry. 2005-10-18 04:50 gb * [r2609] : SEMAPHORE-NOTIFICATION stuff. 2005-10-18 04:49 gb * [r2608] : PROCESS-KILL-ISSUED, SEMAPHORE-NOTIFICATION-STATUS, CLEAR-SEMAPHORE-NOTIFICATION-STATUS, SEMAPHORE-NOTIFICATION, MAKE-SEMAPHORE-NOTIFICATION. 2005-10-18 04:47 gb * [r2607] : Bump. 2005-10-18 04:46 gb * [r2606] : WAIT-ON-SEMAPHORE can accept a SEMAPHORE-NOTIFICATION object and set its status atomically. 2005-10-18 04:46 gb * [r2605] : SEMAPHORE-NOTIFICATION stuff. PROCESS-INTERRIPT returns T on success. Add PROCESS-KILL-ISSUED and SETF of it; make PROCESS-KILL set PROCESS-KILL-ISSUED. 2005-10-18 04:37 gb * [r2604] : INIT-THREAD-FROM-TCR, NEW-TCR: maybe no TCR, maybe no stack areas if out of memory/ out of thread resources. THREAD-INTERRUPT returns T if a thread exists and a signal is successfully delivered. 2005-10-18 04:35 gb * [r2603] : SEMAPHORE-NOTIFICATION objects. 2005-10-15 22:20 gb * [r2602] : *** empty log message *** 2005-10-15 22:14 gb * [r2601] : On PPC64, 64-bit memory accesses (accesses to signed/unsigned 64-bit integers or to pointers) with a constant 16-bit offset can only use LD/STD if the low 2 bits of that offset are 0. Something like: (%get-ptr p 3) would have elicited a warning from the assembler that it's clearing those low 2 bits for you; that's maybe better than generating the wrong instruction, but it doesn't lead to "the pointer 3 bytes from p" being returned, either ... 2005-10-06 19:20 gb * [r2600] : *** empty log message *** 2005-10-06 18:53 gb * [r2599] : Bump: "post-051006". 2005-10-06 18:27 gb * [r2598] : DEFCALLBACK-BODY: ppc64 changes: store all (64 of 'em) bits of an integer return value. Store float return values 2 natural words (16 bytes) into the argument area. 2005-10-05 18:20 gb * [r2597] : *** empty log message *** 2005-10-05 18:17 gb * [r2596] : Fix typo. 2005-09-30 19:20 gb * [r2595] : *** empty log message *** 2005-09-30 18:51 gb * [r2594] : Flesh out a bit more. 2005-09-30 10:20 gb * [r2593] : *** empty log message *** 2005-09-30 09:31 gb * [r2592] : new file 2005-09-27 16:44 gb * [r2591] : *** empty log message *** 2005-09-27 16:20 bryan * [r2590] : import latest asdf-install from cclan. 2005-09-26 15:20 gb * [r2589] : *** empty log message *** 2005-09-26 14:57 gb * [r2588] : In DEFCALLBACK-BODY, coerce SINGLE-FLOAT result to DOUBLE-FLOAT, store in argument area via (SETF (%GET-DOUBLE-FLOAT ,stack-ptr ,offset)). (E.g., un-break what got broken a month or so ago.) 2005-09-24 22:20 gb * [r2587] : *** empty log message *** 2005-09-24 22:05 gb * [r2586] : Bump. 2005-09-24 21:53 gb * [r2585] : Bump ABI_VERSION_MIN. 2005-09-24 21:52 gb * [r2584] : Thread-private hash tables store/check *CURRENT-PROCESS* instead of (%CURRENT-TCR), since TCRs get recycled. 2005-09-24 11:20 gb * [r2583] : *** empty log message *** 2005-09-24 11:03 gb * [r2582] : Clear hash-table-vector caches if the hash-table is weak (on key or value.) Could be smarter about this (e.g., treat them as weak references), but don't want things to stay around just because they're referenced from the cache. 2005-09-24 10:43 gb * [r2581] : #include , so that this compiles under Jaguar. 2005-09-23 05:20 gb * [r2580] : *** empty log message *** 2005-09-23 04:39 gb * [r2579] : Bump, 1.0-rc1. 2005-09-23 04:39 gb * [r2578] : Back out of change to DECF, 'til after 1.0. 2005-09-21 11:20 gb * [r2577] : *** empty log message *** 2005-09-21 10:37 gb * [r2576] : START-HASH-TABLE-ITERATOR: use LOCK-HASH-TABLE instead of manipulating the exclusion lock directly. FINISH-HASH-TABLE-ITERATOR: use UNLOCK-HASH-TABLE. 2005-09-21 09:20 gb * [r2575] : *** empty log message *** 2005-09-21 08:57 gb * [r2574] : In _SPset_hash_key, ensure that a local label "2" is defined (so that "(bne 2f)" branches to the right place.) 2005-09-14 05:20 gb * [r2573] : *** empty log message *** 2005-09-14 04:37 gb * [r2572] : Bump; call it a "test" release. 2005-09-14 04:36 gb * [r2571] : ABI-VERSION 1003. 2005-09-14 04:30 gb * [r2570] : %WITH-RECURSIVE-LOCK-PTR uses a LOCK-ACQUISITION. 2005-09-14 04:29 gb * [r2569] : Locking primitives accept optional LOCK-ACQUISITION objects. Accesors (get, clear) for LOCK-ACQUISITION objects; PRINT-OBJECT method. 2005-09-14 04:27 gb * [r2568] : Accept and update a LOCK-ACQUISITION object when locking (or trying to). 2005-09-14 04:26 gb * [r2567] : PUTHASH: don't complain about mapping conflicts unless we try to remove something and it's not at the current iteration index. 2005-09-14 04:20 gb * [r2566] : Export LOCK-ACQUISITION, MAKE-LOCK-ACQUISITION, LOCK-ACQUISITION-STATUS, and CLEAR-LOCK-ACQUISITION-STATUS. 2005-09-14 03:20 gb * [r2565] : *** empty log message *** 2005-09-14 03:12 gb * [r2564] : Accessor, constructor for LOCK-ACQUISITION. 2005-09-14 03:11 gb * [r2563] : BUILT-IN-CLASS for LOCK-ACQUISITION. 2005-09-14 03:08 gb * [r2562] : Mention CPU-COUNT. 2005-09-14 02:20 gb * [r2561] : *** empty log message *** 2005-09-14 01:32 gb * [r2560] : On PPC64, save_application() writes the image and section headers after writing the section contents, maintains the offset from the end of those headers to the start of section data in the (split) 64-bit section_data_offset field(s) of the header. (This should allow very large sections to be written; the offset from the trailer at the end of the file to the header has to fit in 32 bits, and - if the header preceded the section data - would have constained the entire image to about 2GB.) 2005-09-14 01:28 gb * [r2559] : high half of PPC64 section_data_offset field is signed, low half unsigned. 2005-09-13 22:20 gb * [r2558] : *** empty log message *** 2005-09-13 22:16 gb * [r2557] : Split section_data_offset in half, to deal with alignment issues. 2005-09-13 22:01 gb * [r2556] : If gc is inhibited on entry, ensure that GC_INHIBIT_COUNT looks negative. (%UNLOCK-GC-LOCK should increment it in this case, and do a gc trap if it goes from -1 to 0.) 2005-09-13 21:59 gb * [r2555] : On ppc64, zero (for now) fh.section_data_offset; zeroing "pad" array only happens on ppc32. 2005-09-13 21:58 gb * [r2554] : Fix newly-introduced typo. 2005-09-13 21:53 gb * [r2553] : Move the "strange reference to exp()" from current_stack_pointer() - which is probably inlined - to flush_cache_lines(). Make it a data reference, and make it word-size dependent. 2005-09-13 21:50 gb * [r2552] : exportfn generates and exports .foo as well as foo when #+(and LINUX PPC64). 2005-09-13 21:48 gb * [r2551] : Some older toolchains don't recognize DATA_SEGMENT_RELPRO_END (and, frankly, neither do I.) Comment out the reference, and see if it's important on platforms that -do- reference it. 2005-09-13 21:36 gb * [r2550] : If reloading a hash table from a fasl file, use the GENERAL-HASH-FIND[-FOR-PUT] functions instead of anonymous lambdas. 2005-09-13 21:33 gb * [r2549] : No compression in SAVE-APPLICATION. 2005-09-13 21:33 gb * [r2548] : No compression in SAVE-APPLICATION, so lose GC_TRAP_FLAG_COMPRESS. 2005-09-13 03:20 gb * [r2547] : *** empty log message *** 2005-09-13 03:14 gb * [r2546] : In %NEEDS-REHASHING: both (GC-COUNT) and (NHASH.GC-COUNT hash) are fixnums. When creating a hash table, use new GENERAL-HASH-FIND and GENERAL-HASH-FIND-FOR-PUT, rather than anonymous lambdas, for the non-EQ/non-EQL cases. PUTHASH's iteration check looks at all pending iterators, uses hash-table-test to compare keys. REMHASH wasn't checking. (Sorry, I thought that I'd checked this in already.) In %GROW-HASH-TABLE, remove a PROGN and a comment; remove "can't happen" check of nhash.lock, which should be mostly unused. Only process a key from the vector if it's neither FREE-HASH-KEY-MARKER nor DELETED-HASH-KEY-MARKER (code was just checking for (EQ KEY (%UNBOUND-MARKER)), where (%UNBOUND-MARKER) == FREE-HASH-KEY-MARKER.) This *might* cause keys to become lost (leading to duplicate key errors), or might not. %REHASH doesn't clear cache key, since DO-REHASH always clears it. 2005-09-11 04:20 gb * [r2545] : *** empty log message *** 2005-09-11 03:47 gb * [r2544] : Bump. 2005-09-11 03:36 gb * [r2543] : Fix other typos. 2005-09-11 03:34 gb * [r2542] : Fix typo. 2005-09-11 03:30 gb * [r2541] : CCL::TOPLEVEL isn't too well-defined outside of the initial process; let it do PROCESS-RESET in that case. 2005-09-11 03:27 gb * [r2540] : %SYM-VALUE, %SET-SYM-VALUE: don't create an SVAR; only call %FIND-SVAR if not constant/global. The idea of pre-allocating a reserved svar-idx for *INTERRUPT-LEVEL* doesn't work too well, since nothing references that svar (and since the hash table's weak). Post-1.0. 2005-09-09 05:20 gb * [r2539] : *** empty log message *** 2005-09-09 04:43 gb * [r2538] : In PUTHASH, increment (nhash.count hash) when overwriting a deleted entry. Clobber cached value too when starting to DO-REHASH. When DO-REHASH encounters a deleted entry, change it to free; among other things, this keeps it from getting counted more than once, and doing that could cause the nhash.vector.deleted-count to go negative. (This seems to be what was causing problems with the weak hash tables that're used to map symbols to SVAR objects used by the special variable binding mechanism: otherwise unreferenced keys (symbols) had their entries removed by the GC; DO-REHASH saw the deleted entries when the hash table was rehashed after GC and counted them multiple times, causing the deleted count to go negative and the hash-table's size to be miscalcuated. It'd eventually become completely full, which isn't ever supposed to happen, and the failure involves a SEGFAULT.) I believe this was causing the problems that Bryan reported encountering while compiling SBCL. 2005-09-08 20:20 gb * [r2537] : *** empty log message *** 2005-09-08 19:39 bryan * [r2536] : csubtypep: we shouldn't answer definitively about all cases when type1 is *wild-type* or type2 is *empty-type*. don't special case and let invoke-type-method deal with them. 2005-09-08 10:20 gb * [r2535] : *** empty log message *** 2005-09-08 09:46 gb * [r2534] : APPLICABLE-RESTART-P: don't bother searching for null conditions (shouldn't have found any, in any case.) FIND-RESTART and COMPUTE-RESTARTS call APPLICABLE-RESTART-P to invoke any restart test function even when condtion is NIL. INVOKE-RESTART-INTERACTIVELY defined in terms of FIND-RESTART, INVOKE-RESTART (just so that there'd be less to go wrong if the internals of %ACTIVE-RESTART change.) 2005-09-08 08:20 gb * [r2533] : *** empty log message *** 2005-09-08 08:12 bryan * [r2532] : bump. 2005-09-08 03:20 gb * [r2531] : *** empty log message *** 2005-09-08 02:47 bryan * [r2530] : make-member-ctype: if doublep and not singlep, use *ctype-of-double-float-0*. 2005-09-08 01:20 gb * [r2529] : *** empty log message *** 2005-09-08 00:38 bryan * [r2528] : array :simple-intersection type-method tries to be more smarter about element-types. (patch ported from SBCL) specialized-element-type of (array nil) is now nil (as opposed to bit). 2005-09-07 20:20 gb * [r2527] : *** empty log message *** 2005-09-07 20:18 bryan * [r2526] : number :simple-union should also check the case when a rational type is adjacent to an integer type. (patch ported from SBCL) union (integer 0 0) (rational (0)) ==> (rational 0) 2005-09-07 19:20 gb * [r2525] : *** empty log message *** 2005-09-07 18:33 bryan * [r2524] : CONS :simple-union type-method should do a little more work for "harder to compute" CONS types especially involving negation. (patch corresponds to SBCL src/code/late-type.lisp v1.100) 2005-09-07 12:20 gb * [r2523] : *** empty log message *** 2005-09-07 12:15 gb * [r2522] : First stab at enumerating new features for 1.0. 2005-09-07 08:20 gb * [r2521] : *** empty log message *** 2005-09-07 08:00 bryan * [r2520] : bump. 2005-09-07 07:51 bryan * [r2519] : type-intersection of numeric types with same numeric lower bound (one exclusive, one not) would return a more inclusive type. type-intersection '(short-float (0.0)) '(short-float 0.0) => '(short-float (0.0)) ;; not '(short-float 0.0) 2005-09-06 23:20 gb * [r2518] : *** empty log message *** 2005-09-06 23:15 gb * [r2517] : Bump. 2005-09-06 12:20 gb * [r2516] : *** empty log message *** 2005-09-06 12:00 gb * [r2515] : Handle GC_TRAP_FUNCTION_IMMEDIATE_GC, by trying to invoke the GC ... immediately. Should maybe try to determine if it's necessary, since there are race conditions. 2005-09-06 11:58 gb * [r2514] : Define GC_TRAP_FUNCTION_IMMEDIATE_GC. 2005-09-06 11:57 gb * [r2513] : MAKE-HASH-TABLE ... :SHARED NIL creates a hash table that's effectively always locked by the calling thread; creation, GET/PUT/REMHASH use new locking primitives. Simplify PUTHASH a little; don't maintain NHASH.PUTHASH-COUNT, handle the replace/free/overwrite-deleted-entry cases separately. 2005-09-06 11:52 gb * [r2512] : %LOCK-GC-LOCK either increments or decrements the gc inhibition lock. %UNLOCK-GC-LOCK does the inverse, and tries to invoke the GC if it increments from -1 to 0. (The idea is that if the GC tries to run when the inhibition value is non-zero, the GC will ensure that that non-zero value is negative. That doesn't work yet; it's easy, but it's another case where the kernel and image need to be synched.) 2005-09-06 11:48 gb * [r2511] : MAKE-LOAD-FORM for HASH-TABLEs handles "private" (thread-owned) hash-tables. (Since we can't exactly dump a thread, they become owned by whatever thread loads them.) FASL locking/unlocking of hash tables handles the thread-owned cases. 2005-09-06 11:46 gb * [r2510] : Replace unused NHASH.LOCKED-ADDITIONS cell with NHASH.OWNER. 2005-09-06 11:44 gb * [r2509] : Add :OPENMCL-PRIVATE-HASH-TABLES to *FEATURES*. 2005-09-06 06:20 gb * [r2508] : *** empty log message *** 2005-09-06 05:54 bryan * [r2507] : decf: define as separate macro to honor order of arg evaluation per ansi 5.1.3. (a similar macro is needed for incf, but it breaks bootstrapping.) 2005-09-06 02:20 gb * [r2506] : *** empty log message *** 2005-09-06 01:45 gb * [r2505] : Update MAKE-LOAD-FORM, fasl-related locking for the new order (and/or for native threads, which happened a few years ago, depending on your point of view.) Try to remove outright lies from the comments. 2005-09-06 01:20 gb * [r2504] : *** empty log message *** 2005-09-06 00:55 gb * [r2503] : "b" is for "better ..." 2005-09-06 00:46 gb * [r2502] : hash-find functions: FIRST-DELETED-INDEX is always a fixnum or NIL, never T. Return FIRST-DELETED-INDEX if a secondary probe finds a free entry and FIRST-DELETED-INDEX is non-nil. (Always return an index.) (In some cases, we were returning T instead of an index from a probe functions. Some callers errored trying to do (1+ T), some callers had aggressive fixnum/optimize declarations that caused 1 to be added to T and the result used to access the hash vector. It's OK if the declarations are aggressive, unless they're wrong, of course. 2005-09-05 20:20 gb * [r2501] : *** empty log message *** 2005-09-05 20:02 bryan * [r2500] : (defclass file-stream): rename original-name slot to actual-filename. stream-original-name, (setf stream-original-name): renamed to stream-actual-filename, (setf stream-actual-filename). make-file-stream, close: use renamed functions. make-file-stream: if filename is relative, use (full-pathname filename). we don't call full-pathname if the filename is absolute to avoid clobbering a logical-pathname with its non-logical counterpart. (maybe full-pathname should do that for us?) make-file-stream, close: use ccl::unix-rename instead of cl:rename-file to avoid merge-pathnames issues when filename has no type. 2005-09-05 14:20 gb * [r2499] : *** empty log message *** 2005-09-05 13:55 gb * [r2498] : Bump. 2005-09-05 13:55 gb * [r2497] : In REMHASH: if the key being removed happens to also be in the cache, set the vector entry to DELETED-HASH-KEY-MARKER. 2005-09-05 13:53 gb * [r2496] : Define slot_unbound for ppc64. 2005-09-05 08:20 gb * [r2495] : *** empty log message *** 2005-09-05 07:34 gb * [r2494] : Bump ABI_VERSION_*, for hash-table related changes. 2005-09-05 07:33 gb * [r2493] : Bump *image-abi-version*. 2005-09-05 07:20 gb * [r2492] : *** empty log message *** 2005-09-05 07:10 gb * [r2491] : Lots of changes. This code hasn't really supported any kind of concurrent access in a long time, so a lot of things (nhash.lock, nhash.locked-additions) have been adding some combination of clutter/overhead here. Inline a lot more, use more aggressive OPTIMIZE declarations. Try to avoid UNWIND-PROTECT overhead associated with locking and unlocking, disabling interrupts, inhibiting GC: wait for/grab the lock early, disable interrupts while doing so, assume that unlocking happens (without need to use UNWIND-PROTECT) because code is uninterruptible. Try to enforce restrictions on (SETF GETHASH)/REMHASH behavior while mapping is in effect. (Needs work.) Represent keys in deleted key/value pairs as SLOT-UNBOUND-MARKERs, so that we can tell whether an entry is empty or deleted with a single memory reference. Try to avoid excessive generality by using specialized lookup functions (at least for EQ/EQL tables). Might want to define these via MACROLET'ed code, since there's a lot of redundancy here; see %HASH-PROBE (which also tries to specialize the redundancy. 2005-09-05 06:23 gb * [r2490] : Don't invalidate cached hash_table key/value pairs here; let lisp code do it if/when it notices that rehashing is necessary. When deleting a (weak) pair, use slot_unbound (rather than unbound) to mark the key as having been deleted. 2005-09-05 06:21 gb * [r2489] : Define slot_unbound, slot_unbound_marker. 2005-09-05 06:20 gb * [r2488] : WITH-HASH-TABLE-ITERATOR: assume that locking, GC inhibition happen inside START-HASH-TABLE-ITERATION and STOP-HASH-TABLE-ITERATION; don't expand into code which does these things explicitly. 2005-09-05 06:20 gb * [r2487] : *** empty log message *** 2005-09-05 06:19 gb * [r2486] : SET-LISP-HEAP-GC-THRESHOLD does nothing (and returns NIL) if its argument isn't a non-negative fixnum. 2005-09-05 06:18 gb * [r2485] : Don't do any SYNCing or ISYNCing before or after %LOCK-GC-LOCK and %UNLOCK-GC-LOCK. (This needs to be tested on MP systems, which is one reason for checking this code in now.) 2005-09-05 06:15 gb * [r2484] : replace unused NHASH.MAP-COUNT with NHASH.ITERATOR. Add NHASH.FIND and NHASH.FIND-NEW function slots to NHASH accessors. 2005-09-05 06:13 gb * [r2483] : CHEADERS includes constants32.h. 2005-09-05 06:12 gb * [r2482] : secondary-keys-*-2 holds double values of the secondary-keys primes. Hash-table-iterators don't track "locked-additions", but do keep track of the previous iterator (for recursive calls.) 2005-09-05 05:56 gb * [r2481] : Reformat some comments. 2005-09-05 05:53 gb * [r2480] : Define *INTERRUPT-LEVEL* and try to guarantee that it's the first svar. (This is intended to allow WITHOUT-INTERRUPTS to be defined as (mostly) a special binding of *INTERRUPT-LEVEL*, since special binding is cheaper than UNWIND-PROTECT, but doing that sanely may take a while.) 2005-09-05 05:50 gb * [r2479] : WRITE-LOCK-RWLOCK-DISABLE-INTERRUPTS: does WRITE-LOCK-RWLOCK and disables interrupts in one operation. (Interrupts are enabled if it has to spin waiting for the lock.) 2005-09-05 05:48 gb * [r2478] : Use ACODE-S16-CONSTANT-P to recognize unbound/slot-unbound markers as 16-bit constants. 2005-09-05 05:47 gb * [r2477] : Be more careful about asserting that things that can have side-effects have the SIDE-EFFECT-FREE attribute. 2005-09-05 05:45 gb * [r2476] : ACODE-16-BIT-CONSTANT-P recognizes unbound/slot-unbound markers, as well as fixnum constants whose boxed representation is (SIGNED-BYTE 16). 2005-09-05 05:39 gb * [r2474] : Don't try to reference STACK-PTR when tracing in DEFCALLBACK-BODY; it may never have been bound. (Should probably try to separate the traced case from the non-traced case better in all of this.) 2005-09-05 00:20 gb * [r2473] : *** empty log message *** 2005-09-04 23:44 bryan * [r2472] : (def-type-translator complex ...) ratio and bignums aren't properly handled because they are represented by intersection and union ctypes. teach it just enough to deal with them and other simple variants. more work (and understanding on my part) will be needed to handle more complex (har har) types. ported from sbcl 0.9.4.23. 2005-09-02 10:20 gb * [r2471] : *** empty log message *** 2005-09-02 10:06 gb * [r2470] : GET-INTERNAL-REAL-TIME basically tracks elapsed time since the first time it's called in a session, is more likely to return fixnums after that first call. 2005-09-02 10:05 gb * [r2469] : pc_luser_xp(): allow catch_frame initialization to be interrupted, do the right thing if it is. 2005-09-02 10:04 gb * [r2468] : Recognize instructions (str(rX,nnn(nargs))) used to initialize catch frames. 2005-09-02 10:03 gb * [r2467] : Try to make catch/unwind-protect frames faster: allocate on the tstack in an unboxed frame, use NARGS as a base register, set each field individually and in order, then change the tstack frame to 'boxed'. pc_luser_xp() recognizes stores relative to nargs and zeroes out any partially initialized fields in the catch frame. 2005-09-02 07:20 gb * [r2466] : *** empty log message *** 2005-09-02 06:49 bryan * [r2465] : bump. 2005-09-02 05:20 gb * [r2464] : *** empty log message *** 2005-09-02 04:33 bryan * [r2463] : the reader gets first dibs on arithmetic errors. signal a reader-error when an "impossible number" is read. also display the (more informative) original error. 2005-09-01 14:20 gb * [r2462] : *** empty log message *** 2005-09-01 13:28 gb * [r2461] : be more careful about restoring saveregs after ff-call, syscall: set to zero, change tcr valence back to lisp, -then- vpop_saveregs. eabi_syscall wasn't saving/restoring NVRs at all; even though they're preserved across syscalls, we need to have their values somewhere that the GC can see them in case GC occurs while current thread is blocked in syscall. (Doing GC in Slime is one way to trigger this, since a few threads are blocked in syscall.) This doesn't seem to have been newly introduced, but NVRs may be more likely to contain relocatable lisp values when syscalls are entered because of register allocation changes. There was also a Linux-specific change that did a "stw" into tcr.valence; that wants to use the "str" macro for PPC64. These changes haven't yet been tested under Darwin; they -look- right, but beware. 2005-09-01 13:20 gb * [r2460] : Indentation changes only. 2005-08-31 22:20 gb * [r2459] : *** empty log message *** 2005-08-31 21:36 gb * [r2458] : NX1-1D-VSET, NX1-1D-VREF: only transform into a typed uvref if the array's declared/known to be simple. 2005-08-31 03:20 gb * [r2457] : *** empty log message *** 2005-08-31 03:00 gb * [r2456] : It shouldn't matter too much, but the newly documented WAIT-ON-SEMAPHORE uses an internal timeout of 1 second (vice 0 seconds.) 2005-08-31 02:59 gb * [r2455] : On older versions of LINUX, sem_timedwait() may not return error status correctly (see ); code here was recently changed to assume that errors were always returned in errno. Assume that errno is only set correctly if the return value is negative; this seems to fix at least some problems with SLIME, OpenMCL, and old Linux C libraries. 2005-08-30 21:20 gb * [r2454] : *** empty log message *** 2005-08-30 20:24 bryan * [r2453] : add doc strings 2005-08-30 20:20 gb * [r2452] : *** empty log message *** 2005-08-30 19:59 bryan * [r2451] : add more doc strings. 2005-08-30 17:20 gb * [r2450] : *** empty log message *** 2005-08-30 16:26 bryan * [r2449] : add doc strings to development-mode dictionary. 2005-08-30 16:20 gb * [r2448] : *** empty log message *** 2005-08-30 15:52 bryan * [r2447] : add doc strings to garbage collection dictionary 2005-08-30 08:20 gb * [r2446] : *** empty log message *** 2005-08-30 07:31 bryan * [r2445] : bump. 2005-08-30 07:26 bryan * [r2444] : add doc strings for operating-system dictionary 2005-08-30 07:20 gb * [r2443] : *** empty log message *** 2005-08-30 07:12 bryan * [r2442] : add doc strings for the foreign-function-interface dictionary. 2005-08-30 06:20 bryan * [r2441] : add doc strings for external-program dictionary. 2005-08-30 06:20 gb * [r2440] : *** empty log message *** 2005-08-30 06:06 bryan * [r2439] : add doc strings for sockets dictionary. 2005-08-30 05:24 bryan * [r2438] : add doc strings for the threads dictionary. 2005-08-29 04:20 gb * [r2437] : *** empty log message *** 2005-08-29 04:19 gb * [r2436] : Optionally suppress messages about cold-load functions and their addresses. 2005-08-29 04:18 gb * [r2435] : Base the choice of -lgcc vs. -lSystemStubs on the GCC version, not the availability of -lSystemStubs. 2005-08-29 00:20 gb * [r2434] : *** empty log message *** 2005-08-28 23:22 gb * [r2433] : TARGET-XLOAD-LEVEL-0 prints a message reminding us of the bootstrapping image's name and location. 2005-08-28 23:20 gb * [r2432] : *** empty log message *** 2005-08-28 23:13 gb * [r2431] : Bump. 2005-08-28 22:55 gb * [r2430] : WITH-HASH-TABLE-ITERATOR: get a clue, don't evaluate argform twice. 2005-08-28 22:45 gb * [r2429] : ppc64::tag-misc -> ppc64::fulltag-misc in ISTRUCT-TYPEP vinsn. 2005-08-28 22:44 gb * [r2428] : Hash-table accessors here. 2005-08-28 22:44 gb * [r2427] : Hash-table locking macros here. 2005-08-28 22:43 gb * [r2426] : Move hash-table accessors (nhash.*) to the more-public lispequ.lisp. 2005-08-28 22:42 gb * [r2425] : Move WITH-EXCLUSIVE-HASH-LOCK, WITH-HASH-WRITE-LOCK to macros.lisp, since we want them to be used in more-public code (WITH-HASH-TABLE-ITERATOR). 2005-08-28 08:20 gb * [r2424] : *** empty log message *** 2005-08-28 07:33 gb * [r2423] : Fix typo (extra #else). 2005-08-28 07:32 gb * [r2422] : Ignore errors (CHUD state) when cleaning up on exit. 2005-08-28 07:29 gb * [r2421] : Bigger red zone, larger stack alignment on ppc64. 2005-08-28 07:28 gb * [r2420] : Bump. 2005-08-28 07:27 gb * [r2419] : WITH-HASH-TABLE-ITERATOR locks table & disables GC. 2005-08-28 07:26 gb * [r2418] : If we have to obtain an exclusive lock on the hash table and disable GC (as is currently the case), wait for the lock first. 2005-08-28 07:25 gb * [r2417] : Inline ISTRUCT-TYPEP. 2005-08-28 07:20 gb * [r2416] : *** empty log message *** 2005-08-28 06:58 gb * [r2415] : Remove a few stale comments. 2005-08-28 04:20 gb * [r2414] : *** empty log message *** 2005-08-28 03:26 gb * [r2413] : Conditionalize SEM_WAIT_FOREVER for Darwin (especially Darwin64). 2005-08-26 11:20 gb * [r2412] : *** empty log message *** 2005-08-26 10:41 gb * [r2411] : Something's got to reset tcr->interrupt_level after an interrupt trap. 2005-08-25 07:20 gb * [r2410] : *** empty log message *** 2005-08-25 06:27 gb * [r2409] : Don't play with interrupt level before callbacks; callbacks can do so themselves. 2005-08-25 06:26 gb * [r2408] : Don't loop if SEM_WAIT_WITH_TIMEOUT gets interrupted; return to lisp code (so that the interrupt can be serviced) and let lisp code decide whether or not to retry. 2005-08-25 06:25 gb * [r2407] : Define SEM_WAIT_FOREVER, for kernel routines that should just restart a SEM_WAIT() call. 2005-08-25 06:24 gb * [r2406] : Format a comment better. 2005-08-25 06:23 gb * [r2405] : Semaphore waiting functions assume that %wait-on-semaphore-pointer returns a second value (a POSIX-style error number). Handle retries because of #$EINTR in lisp code (don't loop uninterruptibly in the kernel.) 2005-08-25 06:22 gb * [r2404] : Ensure that interrups are enabled around a break loop. 2005-08-25 06:21 gb * [r2403] : Arrgh. INITIALIZE-RANDOM-STATE uses %ILSL to shift, since ASH produces an arithmetically correct bignum, and we don't want that. 2005-08-25 06:20 gb * [r2402] : *** empty log message *** 2005-08-25 06:20 gb * [r2401] : Wait for lock's semaphore in one-second intervals. 2005-08-24 21:20 gb * [r2400] : *** empty log message *** 2005-08-24 20:21 gb * [r2399] : Restore some of the "locked additions" handling in GETHASH, since I probably haven't thought about the implications of removing it carefully enough. 2005-08-24 20:20 gb * [r2398] : *** empty log message *** 2005-08-24 20:19 gb * [r2397] : Fix handling of populations in recursive case of rmark. 2005-08-24 19:58 gb * [r2396] : Bump. 2005-08-24 19:58 gb * [r2395] : Tweak ROOM output. 2005-08-24 12:20 gb * [r2394] : *** empty log message *** 2005-08-24 11:47 gb * [r2393] : We don't actually use WITH-REHASH-LOCK, and don't want to allocate an nhash.rehash-lock. Remove a little (currently dead) code from the start of GETHASH. 2005-08-24 11:46 gb * [r2392] : We don't actually use WITH-REHASH-LOCK, and don't want to allocate an nhash.rehash-lock. 2005-08-24 11:45 gb * [r2391] : Remove a newline, add a semicolon at top of file. 2005-08-24 08:20 gb * [r2390] : *** empty log message *** 2005-08-24 07:41 gb * [r2389] : current_stack_pointer() already prototyped, inlined; need to cast to BytePtr. 2005-08-24 07:40 gb * [r2388] : current_stack_pointer() already prototyped, inlined. 2005-08-24 07:39 gb * [r2387] : Make rmark() recursive, if there's enough stack space. (Falls back to Deutch-Waite-Schorr pointer inversion if there isn't enough stack space.) 2005-08-24 07:38 gb * [r2386] : Prototype for inlined current_stack_pointer(). (Prototyped to return a "natural"; some previous uses prototyped to return a "BytePtr".) 2005-08-24 07:37 gb * [r2385] : Export *REPORT-TIME-FUNCTION*. 2005-08-24 07:36 gb * [r2384] : Add *REPORT-TIME-FUNCTION*, which allows customization of the presentation of TIME's output. 2005-08-24 07:35 gb * [r2383] : New file. 2005-08-24 02:20 gb * [r2382] : *** empty log message *** 2005-08-24 02:03 gb * [r2381] : Still more indentation problems. 2005-08-24 01:52 gb * [r2380] : Some of the work that needed to be done on 64-bit %FF-CALL involved indenting properly. Basic cases now seem to work. 2005-08-24 01:20 gb * [r2379] : *** empty log message *** 2005-08-24 01:11 gb * [r2378] : First cut of ppc64 %FF-CALL. Needs more work. 2005-08-24 01:10 gb * [r2377] : ALLOC-VARIABLE-EABI-C-FRAME. 2005-08-23 13:20 gb * [r2376] : *** empty log message *** 2005-08-23 12:40 gb * [r2375] : WITH-VARIABLE-C-FRAME. Lose %VREFLET. 2005-08-23 12:39 gb * [r2374] : WITH-VARIABLE-C-FRAME. PPC2-NLEXIT discards C frames. 2005-08-23 12:37 gb * [r2373] : ALLOC-VARIABLE-C-FRAME vinsns. 2005-08-23 12:37 gb * [r2372] : INITIALIZE-RANDOM-STATE uses ASH instead of CCL::%ILSL; note that ASH takes args in the opposite order from %ILSL. 2005-08-23 12:35 gb * [r2371] : Update the list of implementation-dependent special operators: drop %VREFLET and SPARC-LAP-FUNCTION (!), add WITH-C-FRAME and WITH-VARIABLE-C-FRAME. 2005-08-23 12:34 gb * [r2370] : Change major version to 1, minor version to 0. Change format of *OPENMCL-VERSION* format string (move version number before release level, ~A for platform.) "Upgrade" ppc32 to release level, ppc64 to beta. 2005-08-23 12:32 gb * [r2369] : Ensure that (COMPILE-CCL) updates NXENV before compiling the rest of the compiler. 2005-08-23 12:31 gb * [r2368] : Remove some unused constants. 2005-08-23 01:20 gb * [r2367] : *** empty log message *** 2005-08-23 01:07 gb * [r2366] : The seed values in a RANDOM-STATE can have their sign bits set on PPC32, so the PRINT-OBJECT method uses LDB to access the signigicant parts of those values. (Next bug: rewrite this stuff completely; it's not a very good random number generator.) 2005-08-23 00:20 gb * [r2365] : *** empty log message *** 2005-08-23 00:12 gb * [r2364] : Tweak Linux CPU-COUNT. 2005-08-22 23:59 gb * [r2363] : Provide a minimal definition of ED; use *RESIDENT-EDITOR-HOOK* to provide a real definition if/when a resident editor is available. Define CPU-COUNT, which returns ... the number of CPUs (1 or more) available. 2005-08-22 23:56 gb * [r2362] : Export *RESIDENT-EDITOR-HOOK* and CPU-COUNT. 2005-08-22 22:20 gb * [r2361] : *** empty log message *** 2005-08-22 22:19 gb * [r2360] : FASL-FILE-P thinks that files that start with a byte of #xFF are fasl files. 2005-08-22 20:20 gb * [r2359] : *** empty log message *** 2005-08-22 19:40 bryan * [r2358] : %compile-file: set *compile-file-truename* to (truename src). 2005-08-22 10:20 gb * [r2357] : *** empty log message *** 2005-08-22 10:13 gb * [r2356] : Wrap a HANDLER-BIND around compile-time evaluation, in case either compile-time evaluation or the compilation which implements part of it signals a warning that might be considered to have been "detected by the compiler". Another reasonable interpretation would seem to be that warnings that occur in that situation and are not detected by the compiler don't affect COMPILE-FILE's return value. 2005-08-22 10:08 gb * [r2355] : Disable #, which was removed from ANSI CL a long time ago. 2005-08-22 10:07 gb * [r2354] : #+ppc64 %NEXT-RANDOM-SEED accounts for el-bizarro shifting. 2005-08-22 10:05 gb * [r2353] : Use #& instead of #?, which is deprecated (the implementation shouldn't define #?). 2005-08-22 10:05 gb * [r2352] : The #? reader macro is supposed to be reserved for the user. Use #& for foreign variable lookup, and make #? an alias to it (for backward compatibility, for a while.) 2005-08-22 10:03 gb * [r2351] : RANDOM-STATE's PRINT-OBJECT method uses ASH instead of CCL::%ILSR. 2005-08-21 23:20 gb * [r2350] : *** empty log message *** 2005-08-21 22:57 gb * [r2349] : Try to parameterize more for ppc32/ppc64. 2005-08-21 22:57 gb * [r2348] : Add WITH-PROCESS-WHOSTATE, to make saving/restoring a process's WHOSTATE a little easier. %WITH-RECURSIVE-LOCK-PTR uses the new %LOCK-RECURSIVE-LOCK to better ensure that the lock is unlocked if and only if it was successfully locked. 2005-08-21 22:54 gb * [r2347] : Bump. 2005-08-21 22:54 gb * [r2346] : %TIMED-WAIT-ON-SEMAPHORE-POINTER takes an optional WHOSTATE argument. 2005-08-21 22:53 gb * [r2345] : Try harder to keep an (%UNBOUND-MARKER) as the primary return value to TOPLEVEL-PRINT from making *, **, *** unbound. 2005-08-21 22:50 gb * [r2344] : PROCESS-WAIT uses WITH-PROCESS-WHOSTATE. 2005-08-21 22:49 gb * [r2343] : THREAD-HANDLE-INTERRUPTS resets the process-whostate to "Active", using the new WITH-PROCESS-WHOSTATE macro. 2005-08-21 22:48 gb * [r2342] : Try to ensure that things that change the state of a recursive lock do so atomically, make %LOCK-RECURSIVE-LOCK take an extra CONS (optional for bootstrapping), which can be used to reliably indicate success. Add a %%LOCK-OWNER function for debugging. Note that there are some related changes in higher-level code; while bootstrapping, build ppc-boot and a new image before recompiling the higher-level parts of this. 2005-08-21 09:20 gb * [r2341] : *** empty log message *** 2005-08-21 08:28 gb * [r2340] : no SPARC 2005-08-21 03:20 gb * [r2339] : *** empty log message *** 2005-08-21 03:07 gb * [r2338] : 050820b. New binaries soon. 2005-08-21 02:47 gb * [r2337] : Several changes to enable larger functions/code-vectors and to allow more labels in LAP and the compiler. Some of this is tricky to bootstrap; new binaries soon. 2005-08-21 02:20 gb * [r2336] : *** empty log message *** 2005-08-21 01:54 bryan * [r2335] : fill-pointer: when throwing a type-error, set expected-type to '(and array (satisfies array-has-fill-pointer-p)). 2005-08-20 22:20 gb * [r2334] : *** empty log message *** 2005-08-20 21:30 bryan * [r2333] : use "CCL" instead of :ccl in in-package form. 2005-08-20 21:28 bryan * [r2332] : add (in-package "CCL") forms. 2005-08-20 21:20 gb * [r2331] : *** empty log message *** 2005-08-20 21:08 bryan * [r2330] : bump. 2005-08-20 21:05 bryan * [r2329] : remove all reader-conditionalized code for non-openmcl platforms. 2005-08-20 20:20 gb * [r2328] : *** empty log message *** 2005-08-20 20:06 bryan * [r2327] : remove all reader-conditionalized code for non-openmcl platforms. 2005-08-20 19:38 bryan * [r2326] : add (in-package "CCL") forms. 2005-08-20 19:26 bryan * [r2325] : add (in-package "CCL") forms. 2005-08-20 19:20 gb * [r2324] : *** empty log message *** 2005-08-20 19:08 bryan * [r2323] : add (in-package "CCL") forms. 2005-08-20 18:51 bryan * [r2322] : no longer required. all loop code is in loop.lisp. 2005-08-20 07:20 gb * [r2321] : *** empty log message *** 2005-08-20 07:08 bryan * [r2320] : bump. 2005-08-20 07:06 bryan * [r2319] : read-list: make sure the first form read isn't a dot. 2005-08-20 01:20 gb * [r2318] : *** empty log message *** 2005-08-20 01:17 bryan * [r2317] : bump. we now pass all PRINT*, PPRINT*, and FORMAT* ansi tests. 2005-08-20 01:14 bryan * [r2316] : array-readably-printable-p: new function. T if array can be readably printed. write-an-array: if *print-readably*, make sure that the array is readably-printable. 2005-08-19 23:20 gb * [r2315] : *** empty log message *** 2005-08-19 22:24 bryan * [r2314] : print-object (random-state t): cl:random-state was renamed to ccl::initialize-random-state. 2005-08-19 22:20 gb * [r2313] : *** empty log message *** 2005-08-19 21:44 bryan * [r2312] : write-pname: if *print-readably*, use the readtable-case of the "standard readtable". write-pname: if symbol could be read as a number, don't bother slashify-ing. just write it escaped in |..|. 2005-08-18 07:20 gb * [r2311] : *** empty log message *** 2005-08-18 07:14 bryan * [r2310] : bump. 2005-08-18 07:13 bryan * [r2309] : specifier-fn: convert to cond. convert-body: don't call copy-tree. copy-tree breaks pprint dispatch table entries with eql type specifiers. for example this form: (let ((obj '(foo bar))) (set-pprint-dispatch (list* 'eql (list obj)) ...)) 2005-08-16 20:20 gb * [r2308] : *** empty log message *** 2005-08-16 19:45 bryan * [r2307] : remf: according to 5.1.3, the indicator needs to be evaluated before the reading the "old value" of place. 2005-08-16 17:20 gb * [r2306] : *** empty log message *** 2005-08-16 17:16 gb * [r2305] : bump. 2005-08-16 17:13 gb * [r2304] : Use (expt 4 n) instead of (ash 4 n), which isn't the same thing ... 2005-08-16 17:11 gb * [r2303] : ALLOC-C-FRAME: ensure 32-byte alignment on PPC64. 2005-08-16 14:20 gb * [r2302] : *** empty log message *** 2005-08-16 13:35 gb * [r2301] : Remove gratuitous "ldr r2,tcr.native_thread_info(rcontext)" from _SPeabi_callback epilog: on linuxppc32 (the only platform that uses _SPeabi_callback), we should be keeping our hands off of r2 at all times, anyway. 2005-08-16 13:20 gb * [r2300] : *** empty log message *** 2005-08-16 13:18 gb * [r2299] : In xcmain callback (both of them), get rid of code which gratuitously smacks a 0 into the top word of the value stack in the current catch frame in response to an interrupt trap (trgti nargs, 0). (I -think- that this had something to do with reenabling interrupts in a long-forgotten cooperative scheduler.) It's possible that until recent changes to register allocation, the top word on the vstack at this point was either (a) always non-negative and therefore immune to getting smacked around or (b) something not very important. On Linux32, this was currently a pointer to a saved special binding, (happened to look negative because of where that stack was allocated) and setting it to 0 caused a crash as the binding stack was unwound. 2005-08-16 13:10 gb * [r2298] : In STARTUP-CCL, use LOAD-FILE (establishing a restart) instead of just calling LOAD directly when processing a --load command-line option. 2005-08-16 02:20 gb * [r2297] : *** empty log message *** 2005-08-16 02:01 bryan * [r2296] : bump. 2005-08-16 01:59 bryan * [r2295] : defclass: prefer MOP over ANSI behaivor. class options are canonicalized and passed to ensure-class. 2005-08-16 01:20 gb * [r2294] : *** empty log message *** 2005-08-16 01:15 bryan * [r2293] : bump. 2005-08-16 01:09 bryan * [r2292] : defclass: add new case for locally implemented class options (currently only 'primary-p). throw program-error if not one of the ansi or locally implemented class options. this means that all future local class options will need to be added to the defclass macro. 2005-08-16 00:29 bryan * [r2291] : %destructure-lambda-lisp: don't quote the default-initial-value when &key is in list form. the supplied default needs to be evaluated. 2005-08-16 00:20 gb * [r2290] : *** empty log message *** 2005-08-16 00:14 bryan * [r2289] : %destructure-lambda-lisp: make sure default-initial-value is properly quoted in getf form. 2005-08-15 23:40 bryan * [r2288] : defclass: revert previous commit. 2005-08-15 20:20 gb * [r2287] : *** empty log message *** 2005-08-15 20:00 bryan * [r2286] : defclass: signal a program-error if invalid class option seen. 2005-08-15 19:20 gb * [r2285] : *** empty log message *** 2005-08-15 19:00 gb * [r2284] : Reformat a comment; one down, a few thousand more to go ... 2005-08-15 18:59 gb * [r2283] : Enable -O2, at least until the next time something needs to be debugged. 2005-08-15 18:58 gb * [r2282] : Write barrier things check to see if bit is set before entering the (expensive) interlocked load/store loop. (Old objects that're destructively modified are often modified multiple times.) 2005-08-15 18:55 gb * [r2281] : Bum %PL-SEARCH a little. 2005-08-15 18:54 gb * [r2280] : (very) approximate loop-weighting for VAR refrences. Bad, but seems to win more often than it loses. 2005-08-15 13:20 gb * [r2279] : *** empty log message *** 2005-08-15 12:38 gb * [r2278] : Fix typo in version check. 2005-08-14 00:20 gb * [r2277] : *** empty log message *** 2005-08-13 23:25 gb * [r2276] : CCL::EDIT-DEFINITION-P didn't know that all of a method's specializers were specializers, so it sometimes called CLASS-NAME on an EQL-SPECIALIZER. SLIME calls CCL::EDIT-DEFINITION-P, but I'm not really sure that that's a good idea. 2005-08-13 20:20 gb * [r2275] : *** empty log message *** 2005-08-13 20:09 bryan * [r2274] : %load: allow loading from string-input-streams. load-from-stream already supported it, but the path to get there was littered with file-stream traps. 2005-08-13 18:20 gb * [r2273] : *** empty log message *** 2005-08-13 18:05 bryan * [r2272] : file-error: set initform of "" for pathname. 2005-08-13 17:20 gb * [r2271] : *** empty log message *** 2005-08-13 16:38 bryan * [r2270] : pprint-indent: truncate the arg here instead of later in set-indentation-prefix. 2005-08-13 16:20 gb * [r2269] : *** empty log message *** 2005-08-13 15:53 gb * [r2268] : Check that we've got at least CHUD 4.1. A version of CHUD 4.2.x gets installed via XCode 2.1; that version doesn't seem to be usable. Versions of CHUD 4.2.x that're distributed on Apple's Performance Tools site seem to work better. 2005-08-13 15:51 gb * [r2267] : Define and use DCODE-TOO-FEW-ARGS. %%NTH-ARG-DCODE was checking the argument count incorrectly; this could have something random to have been used (in the &lexpr case) in dispatching. Note that this detects -some- cases of missing args to a GF, but generic functions that use %%NTH-ARG-DCODE (at least) don't do rigorous number-of- arg checking on entry; cases like: (add-method 6) still generate a NO-APPLICABLE-METHOD without noticing that an argument's missing. (-maybe- partly fixable in the short term). 2005-08-13 15:44 gb * [r2266] : Make the REPORT method for wrong-number-of-arguments errors mention the function (which isn't always the same as what's reported as "while executing ...".) 2005-08-13 15:43 gb * [r2265] : Defvar *SIMPLE-ARRAY-CLASS* (not a toplevel form, so the compiler may warn the fist time this is compiled.) CLASS-OF returns that class for simple multidimensional arrays (but see CLHS.) 2005-08-13 07:20 gb * [r2264] : *** empty log message *** 2005-08-13 07:10 bryan * [r2263] : write-a-structure: change *print-length* handling. 2005-08-13 05:20 gb * [r2262] : *** empty log message *** 2005-08-13 04:49 bryan * [r2261] : print-object, write-a-structure: structures with no slots are considered to have no components, so ignore *print-level*. 2005-08-12 11:20 gb * [r2260] : *** empty log message *** 2005-08-12 11:16 gb * [r2259] : Be a little more careful in XLOAD-SET-DOCUMENTATION; only call SET-DOCUMENTATION if the argument might be a string. 2005-08-12 11:14 gb * [r2258] : Skip the second (more global) check for duplicate component names, per a suggestion from Christophe Rhodes on #lisp. This may help to ensure that McCLIM builds. 2005-08-12 11:13 gb * [r2257] : SD accessors use SVREF, not %SVREF. (Not sure what the bug is, but this change may help to catch it earlier.) 2005-08-12 11:12 gb * [r2256] : bump. 2005-08-12 11:12 gb * [r2255] : %PUT-DOCUMENTATION: if "doc" is non-null, ensure that it's a string. 2005-08-12 04:20 gb * [r2254] : *** empty log message *** 2005-08-12 03:20 gb * [r2253] : Needed to say #. before 64-bit vector type names in %UVECTOR-REPLACE. 2005-08-11 15:20 gb * [r2252] : *** empty log message *** 2005-08-11 15:12 gb * [r2251] : handle single-float case in %UVECTOR-REPLACE. 2005-08-11 13:20 gb * [r2250] : *** empty log message *** 2005-08-11 12:54 gb * [r2249] : Update for CHUD 4.1.x. 2005-08-11 12:54 gb * [r2248] : Do 3-arg char, numeric comparisons at least partly inline. 2005-08-11 12:53 gb * [r2247] : Don't do SOME/EVERY when evaluating VINSN predicates. 2005-08-11 12:52 gb * [r2246] : New file. Needs some work, but somewhat useful. 2005-08-09 18:20 gb * [r2245] : *** empty log message *** 2005-08-09 18:18 bryan * [r2244] : bump after fix for ~F bug. 2005-08-09 18:15 bryan * [r2243] : revert part of changes made in v1.14. format-fixed-aux: don't forget to write #\0 after decimal point and significant digits. 2005-08-09 12:20 gb * [r2242] : *** empty log message *** 2005-08-09 11:59 gb * [r2241] : Use (record-length :linger) instead of 8 when setting the #$SO_LINGER socket option. 2005-08-09 11:27 gb * [r2240] : On (32-and-64-bit) Linux, link_map entries (used to find shared libraries) may have their :link_map.l_addr fields set to NULL; this caused INIT-SHARED-LIBRARIES to think that all libraries were mapped to the same address (0). In %SHLIB-FROM-MAP-ENTRY, use #_dladdr to find the library base address that contains the library's dynamic section if the link_map.l_addr field is NULL. 2005-08-09 03:20 gb * [r2239] : *** empty log message *** 2005-08-09 02:44 bryan * [r2238] : write-a-symbol: add #: to "apparently uninterned" symbols which are those having a home package of nil. 2005-08-09 00:09 gb * [r2237] : *** empty log message *** 2005-08-09 00:01 bryan * [r2236] : new-numtoken: if token is a float, ignore *read-base* and read it in base 10. 2005-08-08 23:20 gb * [r2235] : *** empty log message *** 2005-08-08 23:04 gb * [r2234] : Don't assume that the CTR gets saved in a jmp_buf (it appears not to be under Tiger, which causes the whole %associate-jmp-buf-with-catch-frame hack to fail.) Instead, set r14 in the jmp_buf to THROW-TO-CATCH-FRAME's address, and add an extra instruction to *SETJMP-CATCH-LR-CODE* to copy r14 to the CTR. 2005-08-08 12:20 gb * [r2233] : *** empty log message *** 2005-08-08 11:45 gb * [r2232] : Ignore SIG40. 2005-08-08 11:45 gb * [r2231] : Linux thread stack size stuff wants parameters to be of natural word size. 2005-08-08 11:44 gb * [r2230] : Conditionalize on EABI/POWEROPEN ABI, not Linux/Darwin in DEFINE-CALLBACK, too. 2005-08-08 11:44 gb * [r2229] : Conditionalize on EABI/POWEROPEN ABI, not Linux/Darwin. 2005-08-08 11:43 gb * [r2228] : Fix a few LinuxPPC64 constants. Can do callbacks now, but may still be bugs. 2005-08-08 11:42 gb * [r2227] : Fix type: #+LINUX-PPC-TARGET, not #+LINUX. 2005-08-08 11:42 gb * [r2226] : %GET-FP-ARG-REGS is #+POWEROPEN-TARGET, not #+DARWINPPC-TARGET. Add a %LOAD-FP-ARG-REGS, for runtim %FF-CALL implementation. 2005-08-08 11:40 gb * [r2225] : Handle part of the hard case in 32-bit %EXTEND-VECTOR. I think that this was causing a 32-bit VECTOR-PUSH-EXTEND test to fail, but VECTOR-PUSH-EXTEND can't always use %EXTEND-VECTOR, anyway. 2005-08-08 11:39 gb * [r2224] : Add :POWEROPEN-TARGET to *FEATURES* if it was there at compile-time (e.g., if it was on BACKEND-TARGET-SPECIFIC-FEATURES of the target backend.) 2005-08-08 11:37 gb * [r2223] : Define %UVECTOR-REPLACE, intended to handle REPLACE when both args are vectors of the same element type. Fix a few fenceposts/bugs in VECTOR-PUSH-EXTEND: if old vector size is 0, ensure that new vector size is non-zero (4, arbitrarily.) When explicitly displaced, we may not be displaced to the entire vector, so using %EXTEND-VECTOR to do the copying can't work (use %UVECTOR-REPLACE instead.) After extending, the vector is no longer explicitly displaced, so clear the bit that claims that it is. 2005-08-08 07:20 gb * [r2222] : *** empty log message *** 2005-08-08 06:43 bryan * [r2221] : set-indentation-prefix: truncate new-position since fractional ems aren't currently supported. 2005-08-08 06:20 gb * [r2220] : *** empty log message *** 2005-08-08 05:23 bryan * [r2219] : (print-object standard-object ..): add *print-level* check. 2005-08-08 00:20 gb * [r2218] : *** empty log message *** 2005-08-07 23:29 bryan * [r2217] : write-perverted-string: the :capitalize case (har har) requires slightly different behavior if the readtable case is :downcase. 2005-08-07 18:20 gb * [r2216] : *** empty log message *** 2005-08-07 17:29 gb * [r2215] : Say "r11" instead of just "11", since doing otherwise would lead to harder stuff. 2005-08-07 17:27 gb * [r2214] : Define rTOC when defining other platform dependencies. 2005-08-07 17:27 gb * [r2213] : "conditional" definition of rTOC didn't work; do it another way, elsewhere. 2005-08-07 17:26 gb * [r2212] : Honestly intend to read the #+/#- section one of these days. 2005-08-07 17:25 gb * [r2211] : Fix reference to old subprim names. 2005-08-07 17:20 gb * [r2210] : *** empty log message *** 2005-08-07 16:58 gb * [r2209] : Start to provide 64/32-bit versions of things; more work needs to be done here. 2005-08-07 16:57 gb * [r2208] : Subprim renaming, to better reflect EABI/PowerOpen-ness. Deal with the TOC if it exists. 2005-08-07 16:56 gb * [r2207] : Subprim renaming, to better reflect EABI/PowerOpen-ness. 2005-08-07 16:56 gb * [r2206] : Reload the TOC into rTOC when exiting from start_lisp on platforms that use a TOC (LinuxPPC64). 2005-08-07 16:54 gb * [r2205] : Need to specify PROT_EXEC; PPC64 MMUs can apparently enforce execute protection, and Linux pays attention to the PROT_EXEC bit in mmap/ mprotect calls. 2005-08-07 16:53 gb * [r2204] : Try to get LinuxPPC64 exception frame accessors right. (Floating-point may still be bogus.) 2005-08-07 16:52 gb * [r2203] : Y-OR-N-P, YES-OR-NO-P: try to ensure that output side of *DEBUG-IO* realizes that a #\NewLine has been read. Don't beep in irritation; do beep in YES-OR-NO-P, 'cause we're supposed to. (Hey, don't blame me if Emacs doesn't beep when it sees a ^G.) 2005-08-07 16:49 gb * [r2202] : Bump. 2005-08-07 16:49 gb * [r2201] : Callback trampolines for LinuxPPC64 are two (64-bit) word "transfer vectors". 2005-08-07 16:48 gb * [r2200] : Impement a STREAM-READ-LINE for TWO-WAY-ECHOING-STREAM: after a line has been read (along with the terminating #\NewLine), output side should be at column 0. 2005-08-07 16:47 gb * [r2199] : Use 64-bit memory accessors for some TCR slots ("os" thread ID, "native" thread ID, etc.) 2005-08-07 16:45 gb * [r2198] : WRITE-A-CONS: use WRITE-INTERNAL for both CAR and CDR, to ensure bogusity checks. 2005-08-07 16:44 gb * [r2197] : Don't claim to implement %FF-CALL for LinuxPPC64 (yet.). 2005-08-07 16:43 gb * [r2196] : Reference 64-bit ELF structure fields if on a 64-bit platform. 2005-08-07 16:42 gb * [r2195] : LinuxPPC64 has :poweropen-target (not :eabi-target) on its backend-target-specific-features. 2005-08-07 16:41 gb * [r2194] : Gratuitously rename subprimitives and vinns dealing with foreign-function calls, system calls, and callback to reflect whether they're oriented towards the (Linux 32 bit) EABI or some flavor of the PowerOpen ABI (derived from AIX). 2005-08-07 07:20 gb * [r2193] : *** empty log message *** 2005-08-07 06:39 bryan * [r2192] : get-properties: use '(satisfies proper-list-p) instead of '(satisfies plistp). plistp will throw a type-error if not a proper list. all occurrences of plistp (there aren't many) should probably use proper-list-p instead. 2005-08-07 05:20 gb * [r2191] : *** empty log message *** 2005-08-07 05:06 bryan * [r2190] : with-output-to-string: element-type needs to be evaluated regardless of use. pull it into the let and defer which function to call to make the output stream until runtime. 2005-08-07 04:20 gb * [r2189] : *** empty log message *** 2005-08-07 04:02 bryan * [r2188] : expand-package-iteration-macro: pull pkg-spec into the surrounding let* in case its evaluation has an interesting side effect. 2005-08-07 02:20 gb * [r2187] : *** empty log message *** 2005-08-07 02:14 bryan * [r2186] : princ shouldn't rebind *print-circle*. ~A format directive was calling princ which in turn was rebinding *print-circle*. 2005-08-06 19:20 gb * [r2185] : *** empty log message *** 2005-08-06 19:10 bryan * [r2184] : if *pretty-print*, (funcall (formatter ...) ...) wouldn't return the unused args to the caller. maybe-initiate-xp-printing and xp-print were the culprits. 2005-08-06 17:20 gb * [r2183] : *** empty log message *** 2005-08-06 17:01 bryan * [r2182] : #\Backspace and #\Rubout should have invalid constituent traits. 2005-08-06 07:20 gb * [r2181] : *** empty log message *** 2005-08-06 06:54 bryan * [r2180] : if coerce fails when dealing with vectors, signal a type-error with an expected-type of '(vector ,element-type ,length) instead of '(vector ,length). 2005-08-06 05:20 gb * [r2179] : *** empty log message *** 2005-08-06 04:23 bryan * [r2178] : correct a comment that said %collect-xtoken returns 3 values.. it actually returns 4. 2005-08-06 00:20 gb * [r2177] : *** empty log message *** 2005-08-05 23:24 gb * [r2176] : Use the same IMAGE_BASE_ADDRESS on LinuxPPC64 as on DarwinPPC64. 2005-08-05 23:24 gb * [r2175] : Replace a few tests for DARWIN with tests for POWEROPENABI. 2005-08-05 23:23 gb * [r2174] : LinuxPPC64 uses (yet another variant of) the PowerOpen ABI, so define POWEROPENABI (and not EABI) when LINUX and PPC64 are both defined. (Lots of assembler source uses e.g. __ifdef([DARWIN]) when it's really just concerned about the type of foreign stack frame to create; this needs to be carefully reviewed.) 2005-08-05 23:21 gb * [r2173] : r2 is rTOC when in C code on LinuxPPC64. (it'd always be the same value for all threads, so we can keep it in a global.) (Could assume that LinuxPPC64 implies TLS-based libc, and make the TCR be a __thread variable.) 2005-08-05 23:20 gb * [r2172] : *** empty log message *** 2005-08-05 23:18 gb * [r2171] : Dont' pass in lisp version stuff anymore. Disable -O2, to make debugging easier. 2005-08-05 23:17 gb * [r2170] : Change the subrpims jump table origin to #x5000 by moving the image base to #x4000. 2005-08-05 23:16 gb * [r2169] : Don't error out, but the LinuxPPC64 execption frame stuff is likely to be very different from its 32-bit counterpart. 2005-08-05 23:15 gb * [r2168] : Conditionalize socketcalls for #+(and linuxppc-target ppc64-target). May need to also conditionalizer their callers in some cases. 2005-08-05 23:14 gb * [r2167] : Process ifaddrs.h 2005-08-05 23:14 gb * [r2166] : Ignore *.p64fsl 2005-08-05 23:14 gb * [r2165] : Ignore ppc-boot* 2005-08-05 23:00 bryan * [r2164] : fix a *print-length* corner case -- printing (A B . C) at *print-length* 2 should print as (A B . C) and not (A B . ...) 2005-08-05 22:20 gb * [r2163] : *** empty log message *** 2005-08-05 22:01 bryan * [r2162] : write-an-array: no need for "..." when printing bit-vectors at any length. 2005-08-05 20:20 gb * [r2161] : *** empty log message *** 2005-08-05 19:23 bryan * [r2160] : copy-pprint-dispatch: check arg, if specified, is either a pprint-dispatch-table or nil. 2005-08-05 19:20 gb * [r2159] : *** empty log message *** 2005-08-05 18:49 bryan * [r2158] : pprint-logical-block: - error if both :prefix and :per-line-prefix specified. - typecheck :prefix, :per-line-prefix, and :suffix as 'string. 2005-08-05 07:20 gb * [r2157] : *** empty log message *** 2005-08-05 07:11 bryan * [r2156] : pprint-newline: check type of arg all the time, not just when pretty printing. if not '(member :linear :miser :fill :mandatory), signal a type-error. 2005-08-05 06:53 gb * [r2155] : %GET-SINGLE-FLOAT-FROM-DOUBLE-PTR has to cons up a SINGLE-FLOAT object on PPC32. 2005-08-05 06:52 gb * [r2154] : (%CALLBACK-FUNCTION pointer) looks up and returns the function on the lisp side of the callback whose foreign side is "pointer". 2005-08-05 06:31 bryan * [r2153] : format-logical-block: if colon modifier, prefix and suffix default to open and close parentheses instead of null string. 2005-08-05 05:20 gb * [r2152] : *** empty log message *** 2005-08-05 04:56 bryan * [r2151] : last one for 22.3.5.2 compliance? logical-block directives not allowed inside justification directives. 2005-08-05 04:20 gb * [r2150] : *** empty log message *** 2005-08-05 03:57 bryan * [r2149] : more 22.3.5.2 cleanup.. logical-block directives can not co-exist with ~<..~:;..~>. 2005-08-05 03:20 gb * [r2148] : *** empty log message *** 2005-08-05 02:39 bryan * [r2147] : move *logical-block-p* defvar from format.lisp to pprint.lisp since it gets built first. 2005-08-05 01:20 gb * [r2146] : *** empty log message *** 2005-08-05 00:53 bryan * [r2145] : format-justification: ignore minpad when there is only one segment. format-justification: algorithm for determining width is different than the one used for ~A, ~S. format-justification-round-columns: new function. 2005-08-04 21:20 gb * [r2144] : *** empty log message *** 2005-08-04 21:12 bryan * [r2143] : ansi section 22.3.5.2 cleanup - format-find-command: ~:T is not allowed within justification format directives. pprint directives (~I, ~W, ~_, ~:T) are not allowed to coexist in the same format control string as a justification directive (~<...~>) that uses ~:;. *format-pprint* will be T whenever a pprint directive has been seen. *format-justification-semi* will be T whenever a ~<..~:;..~> has been seen. 2005-08-04 12:20 gb * [r2142] : *** empty log message *** 2005-08-04 11:24 gb * [r2141] : Bump again (mostly so that the date matches binaries.) 2005-08-04 11:23 gb * [r2140] : Make sure that (SETF GENERIC-FUNCTION-NAME) is implemented, and that it calls REINITIALIZE-INSTANCE to change the slot's value. (This was caught by Pascal Costanza's mop-features tests.) 2005-08-04 09:20 gb * [r2139] : *** empty log message *** 2005-08-04 08:35 bryan * [r2138] : bump. 2005-08-04 08:31 bryan * [r2137] : changes to recognize a "pretty printing stream" and act accordingly. a pretty printing stream is created with pprint-logical-block or the ~<...~:> format directive. *logical-block-p* will be T if we are inside a logical block. pprint-logical-block, format-logical-block: set *logical-block-p* to T. ~_ (conditional newline), pprint-newline, pprint-indent+, pprint-tab: do nothing if not *print-pretty* or not a pretty printing stream. pprint-tab+; do nothing only if ~:T and not *print-pretty* or not a pretty printing stream. ~T format parameter only behaves like pprint-tab when ~:T. change ~T handler to skip pprint-tab and call straight to pprint-tab+. notes -- to test for a pretty printing stream, it's not enough to just check whether we have an xp-stream or xp-structure. the other existing symbols like *current-level* or *logical-block-xp* were not consistently non-nil with a pretty printing stream. changes to pprint-xxx+ functions affect both the cl symbol pprint-xxx and its corresponding format handler. pprint-newline+ had to be left alone since it is used by other code that handles justification. instead i pulled changes out a level to pprint-newline and the ~_ handler. 2005-08-03 10:20 gb * [r2136] : *** empty log message *** 2005-08-03 10:10 gb * [r2135] : SIMPLE-STREAM-ERROR's :report method prints the result of applying FORMAT to the control string and arguments, rather than just printing the control string. 2005-08-03 09:20 gb * [r2134] : *** empty log message *** 2005-08-03 09:05 gb * [r2133] : Try to get "revert" working ... 2005-08-03 09:04 gb * [r2132] : Force all things that need to have the class in an OBJC-METHOD-INFO structure to go through the function GET-OBJC-METHOD-INFO-CLASS to do so; make sure that that function handles metaclasses correctly (the failure to do so was causing (SEND x 'name) to fail.) Remove redundant subclasses when generating CASE/COND forms for SEND; prefer COND to use of union types. 2005-08-03 08:20 gb * [r2131] : *** empty log message *** 2005-08-03 07:39 bryan * [r2130] : bump. 2005-08-03 07:36 bryan * [r2129] : format-fix-aux: skip leading or trailing 0 when number (and sign if @F) would fit the specified width perfectly. 2005-08-03 03:20 gb * [r2128] : *** empty log message *** 2005-08-03 02:40 bryan * [r2127] : pprint-tab: only do something if *print-pretty* and we have a pretty printing stream. 2005-08-03 02:20 gb * [r2126] : *** empty log message *** 2005-08-03 01:21 bryan * [r2125] : set-syntax-from-char: certain characters become invalid if from-char is a constituent. see 2.1.4.2 and figure 2.8. if one of those illegal characters tries to act as a constituent character, signal a reader-error. 2005-08-02 21:20 gb * [r2124] : *** empty log message *** 2005-08-02 20:40 bryan * [r2123] : read-dispatch: signal a reader-error instead of just calling it a reader-error in the error message. 2005-08-02 20:20 gb * [r2122] : *** empty log message *** 2005-08-02 19:59 bryan * [r2121] : find-method: canonicalize specializers before comparison. 2005-08-02 18:20 gb * [r2120] : *** empty log message *** 2005-08-02 18:14 bryan * [r2119] : (setf macro-function): since the spec explicitly notes that a symbol with a previous function definition will be overwritten, change the cerror to a warn. like *ignore-extra-close-parenthesis*, there should probably be some variables that let users change clobbering behavior. 2005-08-02 14:20 gb * [r2118] : *** empty log message *** 2005-08-02 13:47 gb * [r2117] : Fencepost in CFSTR binary search. 2005-08-02 13:47 gb * [r2116] : Don't do MODIFYING-BUFFER around indentation (was an attempt to defer redisplay.) 2005-08-02 13:46 gb * [r2115] : The "Next Line" command doesn't insert newlines at the end of the buffer by default; there's a global Hemlock variable that controls this behavior. 2005-08-02 08:20 gb * [r2114] : *** empty log message *** 2005-08-02 07:53 bryan * [r2113] : file-length: switch to typecase and explicitly throw the type-error so we can specify the expected-type as file-stream. 2005-08-02 05:20 gb * [r2112] : *** empty log message *** 2005-08-02 04:33 bryan * [r2111] : readtable-case, setf (readtable-case): arg must be a readtable. 2005-08-02 04:20 gb * [r2110] : *** empty log message *** 2005-08-02 03:33 bryan * [r2109] : check-sequence-bounds: report-bad-arg if (< end 0). the (> start end) clause would catch this, but we want to report end not start in the error message. 2005-08-02 03:20 gb * [r2108] : *** empty log message *** 2005-08-02 02:53 bryan * [r2107] : make-condition: arg is a type-specifier, so allow both things like 'arithmetic-error and (find-class 'arithmetic-error). 2005-08-02 02:35 gb * [r2106] : bump. 2005-08-01 23:20 gb * [r2105] : *** empty log message *** 2005-08-01 23:14 gb * [r2104] : When setting up standard fd-streams, recognize that FD-OPEN returns a negative (non-NULL) value on failure. 2005-08-01 20:20 gb * [r2103] : *** empty log message *** 2005-08-01 20:01 bryan * [r2102] : ctype-of-number: consider the type of imagpart of complex numbers when determining type. ? (type-of #c(1 1/2)) (COMPLEX (RATIONAL 1/2 1)) ; not (COMPLEX (INTEGER 1/2 1)) 2005-08-01 11:20 gb * [r2101] : *** empty log message *** 2005-08-01 10:55 gb * [r2100] : Allow redisplay when getting key events during search commands. 2005-08-01 10:54 gb * [r2099] : Don't export CHECK-REGION-QUERY-SIZE. 2005-08-01 10:52 gb * [r2098] : Try to ensure that Meta-Space runs "Just One Space". 2005-08-01 10:52 gb * [r2097] : First stab at balancing BEGIN/END-EDITING operations around recursive edits. May still be wrong. 2005-08-01 10:50 gb * [r2096] : Don't call CHECK-REGION-QUERY-SIZE. 2005-08-01 10:50 gb * [r2095] : Wrap BEGIN-EDITING/END-EDITING around command processing. The good news is that this suppresses display after every buffer modification. The bad news is that some extended/recursive commands have to do hairy things to allow redisplay to happen. (The hair mostly has to do with closing the outer BEGIN-EDITING and re-opening it to keep things balanced; sometimes this needs to happen when reading characters from the event queue, since some types of redisplay happen then.) Summary: less hysterical redisplay code, but it may still be suppressed too much; some things that don't really modify the buffer may wrap a MODIFYING-BUFFER around a sequence of lower- level operations, and this may cause the buffer to be marked as modified even when nothing changes. 2005-08-01 10:44 gb * [r2094] : Dont do CHECK-REGION-QUERY-SIZE (I think that that's actually the real name ...). 2005-08-01 10:43 gb * [r2093] : Dont do CHECK-REGION-SIZE-QUERY in the "Filter Region" command. 2005-08-01 10:43 gb * [r2092] : Dont do CHECK-REGION-SIZE-QUERY before indenting a region. 2005-08-01 10:40 gb * [r2091] : Has been edited with the Cocoa/Hemlock editor a lot; indentation differences. 2005-08-01 10:40 gb * [r2090] : In BUILD-CALL (which might be currently unused), recognize : directly. 2005-08-01 10:38 gb * [r2089] : PPC64 %SIMPLE-BIT-BOOLE - advance by 8 bytes (not 4) in loop. 2005-08-01 05:20 gb * [r2088] : *** empty log message *** 2005-08-01 05:09 bryan * [r2087] : sync latest version (1.87) from cclan. 2005-07-31 23:20 gb * [r2086] : *** empty log message *** 2005-07-31 22:55 bryan * [r2085] : check-type: (if string string) === string 2005-07-31 22:50 bryan * [r2084] : check-type: if string not specified, don't replace it with typespec in call to ensure-value-of-type. ensure-value-of-type: remove init-form from &optional typename. 2005-07-31 22:35 bryan * [r2083] : ensure-value-of-type: if typename is specified, create a new format string that uses it instead of binding it to the type-error condition's :expected-type. the spec says that :expected-type must be a typespec. 2005-07-30 06:20 gb * [r2082] : *** empty log message *** 2005-07-30 05:21 gb * [r2081] : New definition of WITH-PACKAGE-ITERATOR. 2005-07-30 05:20 gb * [r2080] : New runtime support for WITH-PACKAGE-ITERATOR. 2005-07-30 05:20 gb * [r2079] : New stuff for WITH-PACKAGE-ITERATOR. 2005-07-30 03:20 gb * [r2078] : *** empty log message *** 2005-07-30 03:16 gb * [r2077] : PROCESS-INPUT-WAIT: don't compute tick count if it's not needed. (Is it ever needed ?) 2005-07-30 02:20 gb * [r2076] : *** empty log message *** 2005-07-30 01:55 gb * [r2075] : In DARWIN version of default_image_name, use strcpy() to copy string to malloc'ed memory. (strcat() only worked - had that effect - if the first byte of the destination was #\Nul.) 2005-07-30 01:52 gb * [r2074] : Start claiming to be an 0.14.4 prerelease. 2005-07-30 01:52 gb * [r2073] : Try to ensure that REPL functions use *STANDARD-INPUT* and *STANDARD-OUPUT* and/or pay attention to their input-stream & output-stream args (rather than doing all I/O to/from *TERMINAL-IO*). Don't print the listener prompt if *QUIET-FLAG* is set. (Should only be set if *BATCH-FLAG* is also set, so this affects REPL and break loop prompts, not inspector, etc.) 2005-07-30 01:48 gb * [r2072] : Don't print the Welcome herald if *quiet-flag* is set. 2005-07-30 01:47 gb * [r2071] : Declare *QUIET-FLAG*. Ensure that *STDOUT* (not necessarily *TERMINAL-OUTPUT*) gets its output flushed periodically. 2005-07-30 01:45 gb * [r2070] : Add support for a -Q/--quiet command-line argument. 2005-07-29 09:20 gb * [r2069] : *** empty log message *** 2005-07-29 09:03 gb * [r2068] : Define %SET-CSTRING; make it the SETF inverse of %GET-CSTRING. Restore %PUT-CSTRING (used by UFFI); implement it in terms of (SETF (%GET-CSTRING ...) ...) 2005-07-29 09:02 gb * [r2067] : Don't do (DEFSETF %GET-STRING %SET-STRING); %GET-STRING is only defined in "ccl:lib;pascal-strings.lisp", and a SETF function for it is defined in that file. 2005-07-28 14:20 gb * [r2066] : *** empty log message *** 2005-07-28 13:48 gb * [r2065] : OBJC-MESSAGE-INFO structure has "ambiguous methods" slot, used by bridge. 2005-07-28 13:47 gb * [r2064] : Don't modify buffer if doing so would change a protected region. 2005-07-28 13:46 gb * [r2063] : Works again (doesn't do much, but WebKit is pretty impressive.) Need new interfaces. 2005-07-28 13:46 gb * [r2062] : Handle ambiguiuty in SEND; convert %SEND to use new (decl-based) scheme. No need to parse ObjC types or introspect to find method/message info. 2005-07-28 13:45 gb * [r2061] : Remove the commented-out call at the end of file; no need to update type signatures via introspection 2005-07-28 13:44 gb * [r2060] : No need to canonicalize class references in type signature table. 2005-07-28 13:43 gb * [r2059] : No need to declare some internal classes (was needed to keep %SEND happy.) No need to update old method type-signature table. 2005-07-28 13:42 gb * [r2058] : Conditionalize out the ObjC GF integration stuff (until it can be adapted to the New Order.) Note that we may need a way to obtain all (declared) ObjC messages - as well as tweak this to use the new method/message decl structures - to revive this. 2005-07-28 13:40 gb * [r2057] : Set bold weight to 9.0 (not a factor of the point size). ObjC "count" message is unambiguous, so no need for type decl. 2005-07-28 13:38 gb * [r2056] : Make %FF-CALL stuff explicitly 32-bit only. (Didn't this already happen last week ?) 2005-07-28 13:35 gb * [r2055] : New file. 2005-07-28 13:34 gb * [r2054] : Ensure that more .cvsignore files exist and that files matching *~.* (e.g., Cocoa NSDocument backup files) are ignored. 2005-07-28 13:30 gb * [r2053] : Ensure that more .cvsignore files exist and that files matching *~.* (e.g., Cocoa NSDocument backup files) are ignored. 2005-07-28 13:23 gb * [r2052] : CCL::TRUE and CCL::FALSE. (There are things that wind up calling TRUE at runtime; there's no reason to actually do that.) 2005-07-28 13:20 gb * [r2051] : *** empty log message *** 2005-07-23 10:20 gb * [r2050] : *** empty log message *** 2005-07-23 09:38 gb * [r2049] : First cut at handling (lots of) FP args in Darwin callbacks. (Probably needs some further testing: there are lots of boundary cases.) 2005-07-23 09:36 gb * [r2048] : %GET-FP-ARG-REGS, for Darwin callbacks. 2005-07-22 08:20 gb * [r2047] : *** empty log message *** 2005-07-22 08:01 gb * [r2046] : Look for pointers in data sections (cfstrings) via binary search (vice linear search.) 2005-07-22 08:00 gb * [r2045] : Don't bind HI::NOW-TICK per-thread. 2005-07-22 07:59 gb * [r2044] : In %SET-BUFFER-MODIFIED: do QUEUE-BUFFER-CHANGE (e.g., update modified status in modeline) after status change. 2005-07-22 07:33 gb * [r2043] : Remove a FORMAT (debugging) call. 2005-07-22 07:20 gb * [r2042] : *** empty log message *** 2005-07-22 06:58 gb * [r2041] : Haven't bumped in a while. 2005-07-22 06:57 gb * [r2040] : Ensure that ASCII info in trailer block is space-delimited. 2005-07-22 06:57 gb * [r2039] : %GET-ERRNO: errno value is always 32-bit int; pointer to errno may be 32/64 bits ... 2005-07-22 06:56 gb * [r2038] : Add SETF-FUNCTION-NAME-P, use it in VALID-FUNCTION-NAME-P. 2005-07-22 06:52 gb * [r2037] : Include /usr/include/ifaddrs.h, for #_getifaddrs/#_freeifaddrs. 2005-07-22 06:51 gb * [r2036] : Kinder, gentler %GET-IP-INTERFACES, based on #_getifaddrs. #_getifaddrs seems to be supported on Darwin at least as far back as Jaguar, and on Linux at least as far back as ... libc versions from 2002 or so. (Note that /usr/include/ifaddrs.h hasn't been included in the Linux interfaces up until this point; note also that the interface names aren't returned correctly on PPC64 Darwin as of OS release 10.4.1). 2005-07-16 01:20 gb * [r2035] : *** empty log message *** 2005-07-16 00:31 gb * [r2034] : Don't assume that all function names that're lists of 2 symbols are SETF function names in VALID-FUNCTION-NAME-P. (Unless they -are- SETF function names, of course.) 2005-07-16 00:28 gb * [r2033] : %SVAR-BINDING-ADDRESS fixes. 2005-07-16 00:27 gb * [r2032] : PPC64::PPC64-ARRAY-TYPE-NAME-FROM-CTYPE should return a keyword or NIL, not ppc32::subtag-simple-vector (sic). 2005-07-16 00:27 gb * [r2031] : PPC32::PPC32-ARRAY-TYPE-NAME-FROM-CTYPE should return a keyword or NIL, not ppc32::subtag-simple-vector. 2005-07-15 19:20 gb * [r2030] : *** empty log message *** 2005-07-15 18:29 gb * [r2029] : Clear the "private" class tree on application exit; ensure that it gets re-populated on application launch. 2005-07-15 13:20 gb * [r2028] : *** empty log message *** 2005-07-15 12:47 gb * [r2027] : Reference *CUBE*, not CUBE. (Need to ask Hamilton about this, and incorporate his "ctrl-left-moust = right-mouse" patches.) 2005-07-15 12:45 gb * [r2026] : Clarify that foreign direct slotds's "offset" is a bit offset; change the name of the slot and its accessor/initarg to reflect this. Foreign effective slot definitions generally treat the location as a byte offset (except the slot's type is a bit or bitfield.) Fix some bugs in calculating instance size of lisp-defined classes, which had led to general chaos/memory corruption problems. 2005-07-15 12:42 gb * [r2025] : Declare the class NSConcreteAttributedString, until SEND is converted to always use decl info. Recognize instances of private (i.e., undeclared) classes. When creating a foreign direct slot, ensure that the type's bits have been determined; use the new :BIT-OFFSET initarg. 2005-07-15 12:38 gb * [r2024] : NEW-BACKTRACE-INFO records the (new) break level. 2005-07-15 12:38 gb * [r2023] : Add a bt.break-level field to backtrace info. 2005-07-15 12:34 gb * [r2022] : Comments. 2005-07-15 12:33 gb * [r2021] : Change some indentation; make UI-OBJECT-DO-OPERATION use EQL specializers on the operation name. 2005-07-15 12:32 gb * [r2020] : Set backtrace window title to indicate process and its break level. 2005-07-15 12:31 gb * [r2019] : SEND still punts to old SEND code sometimes, but formats the message differently. 2005-07-15 11:20 gb * [r2018] : *** empty log message *** 2005-07-15 10:25 gb * [r2017] : Look for an RLDICL that clears the left 61 (not 62) bits, preceding a (TDNEI PPC64::TAG-FIXNUM) trap, in XCMAIN. 2005-07-15 07:20 gb * [r2016] : *** empty log message *** 2005-07-15 06:30 gb * [r2015] : Hamilton's patch for the right-mouse-button challenged. 2005-07-14 21:20 gb * [r2014] : *** empty log message *** 2005-07-14 21:11 gb * [r2013] : Bump. 2005-07-14 21:01 gb * [r2012] : PPC64 VOID-ALLOCPTR is "most-positive-dnode" (e.g, non-negative.) 2005-07-14 21:00 gb * [r2011] : In #+ppc64 walk-dynamic-area, use a signed alloc trap; ensure that ALLOCBASE contains the new (non-negative) VOID_ALLOCPTR. 2005-07-13 16:20 gb * [r2010] : *** empty log message *** 2005-07-13 15:58 gb * [r2009] : Only recognize unsigned alloc_trap on PPC32; only recognize signed alloc_trap on PPC64. 2005-07-13 15:57 gb * [r2008] : Don't pass major/minor release version to CPP. 2005-07-13 15:28 gb * [r2007] : Write newfangled ABI-VERSION to image header. 2005-07-13 15:27 gb * [r2006] : Check abi version in image header; SAVE-APPLICATION writes current version to header. 2005-07-13 15:26 gb * [r2005] : Define min,max, current ABI version (arbitrarily, 1001. I've lost count.) 2005-07-13 15:24 gb * [r2004] : Dont define version numbers in Makefile. 2005-07-13 15:20 gb * [r2003] : *** empty log message *** 2005-07-13 14:59 gb * [r2002] : Bump. Will need new binaries to bootstrap. 2005-07-13 14:58 gb * [r2001] : Increment FASL version. 2005-07-13 14:41 gb * [r2000] : Increase heap-image-base for DarwinPPC32. 2005-07-13 14:41 gb * [r1999] : Define platform-specific alloc_trap, use it. 2005-07-13 14:40 gb * [r1998] : Redefine RATIO, COMPLEX subtags. 2005-07-13 14:40 gb * [r1997] : new (higher) HEAP_IMAGE_BASE for DarwinPPC32. 2005-07-13 14:38 gb * [r1996] : New NUMBERP/REALP for PPC64 (depends on new complex/ratio typecodes.) Redefine PPC64 *NODEHEADER-TYPES*. 2005-07-13 14:37 gb * [r1995] : Redefine PPC64::SUBTAG-RATIO and PPC64::SUBTAG-COMPLEX, so that a mask of all numeric typecodes fits (just barely) in a fixnum. (This enables things like NUMBERP and REALP to check several typecodes via LOGBITP, rather than via a 6-or-7 arm CASE statement.) This requires some bootstrapping and will require new binaries. 2005-07-13 14:33 gb * [r1994] : Define some missing vinsns (some of which were found/defined by jamesjb) for 64-bit vector set, SINGLE-FLOAT-BITS. 2005-07-13 14:32 gb * [r1993] : Fix some cases of PPC2-VSET involving 64-bit arrays. 2005-07-13 14:31 gb * [r1992] : Define MISSING-VINSNS, which found a handful of them. 2005-07-13 14:30 gb * [r1991] : Revert to using "trap logical" alloc-traps on PPC32. Rationale: the largest object we can allocate is about 2^27 bytes in size (a DOUBLE-FLOAT vector with almost 2^24 elements). If HEAP-IMAGE-BASE is effectively above 2^27, we can never cause allocptr to change sign from positive to negative (to wrap around 0), but we -might- cause it to change sign from negative to positive (to wrap around #x80000000). On PPC64, we'd have to be able to guarantee that HEAP-IMAGE-BASE was above 2^59 to ensure this. That's hard to do, so we have to basically limit ourselves to the low 63 bits of memory ("positive" addresses) and do signed comparisons. (Wrapping around 0 should trap; wrapping around #x8000000000000000 shouldn't be possible.) 2005-07-13 14:20 gb * [r1990] : *** empty log message *** 2005-07-13 14:19 gb * [r1989] : Comment formatting. 2005-07-13 07:20 gb * [r1988] : *** empty log message *** 2005-07-13 07:17 gb * [r1987] : destbind1 is now a local label (there was some PPC32 code that thought it was still global.) 2005-07-13 05:20 gb * [r1986] : *** empty log message *** 2005-07-13 04:40 gb * [r1985] : Inline-lambda+keyword args support for PPC64. (Needs more testing, but at least the PPC64 changes are nominally there.) 2005-07-13 04:38 gb * [r1984] : Don't inhibit %SET-SINGLE-FLOAT on PPC64; -do- inhibit %SETF-SINGLE-FLOAT there. 2005-07-12 00:20 gb * [r1983] : *** empty log message *** 2005-07-11 23:50 gb * [r1982] : Bump. 2005-07-11 23:50 gb * [r1981] : Chritian Nybo's change to %%YIELD-TERMINAL-TO (handle the case where *TERMINAL-IO* is a SYNONYM-STREAM.) 2005-07-11 23:48 gb * [r1980] : Chritian Nybo's fix to :Y command. 2005-07-11 23:48 gb * [r1979] : Gotta start somewhere. 2005-07-11 15:20 gb * [r1978] : *** empty log message *** 2005-07-11 14:45 gb * [r1977] : ALLOC-LAP-OPERAND-VECTOR takes an optional SIZE argument, only uses freelisting if the requested size matches the constant LAP-OPERAND-VECTOR-SIZE. 2005-07-11 09:20 gb * [r1976] : *** empty log message *** 2005-07-11 09:07 gb * [r1975] : MEM-REF-C-S16 was generating "lha dest,src(index)" instead of "lha dest,index(src)". 2005-07-10 09:20 gb * [r1974] : *** empty log message *** 2005-07-10 08:40 gb * [r1973] : Try to be saner about DOCUMENTATION methods on SETF function names (as reported by alanr); define methods on the class CONS (not LIST), and only treat the CONS as a SETF function name if it is one. 2005-07-10 08:38 gb * [r1972] : Use ;;; for file-level comments. (A losing battle ...) Use target:: package prefix for stack-overflow fixup kernel call. 2005-07-10 08:37 gb * [r1971] : Use ;;; for file-level comments. (A losing battle ...) 2005-07-09 14:20 gb * [r1970] : *** empty log message *** 2005-07-09 14:13 gb * [r1969] : ref-type shouldn't contain ppc32 subtag. 2005-07-09 14:12 gb * [r1968] : target:: package prefix. 2005-07-09 14:11 gb * [r1967] : Bump. 2005-07-09 14:11 gb * [r1966] : Check for 8-bit element type on ppc64, too. 2005-07-09 14:10 gb * [r1965] : target:: package prefix. Don't print block/go tags, 'cause we don't use them anymore. 2005-07-09 14:09 gb * [r1964] : Change package in buffer attribute line. Always require PPC-ARCH. 2005-07-08 11:20 gb * [r1963] : *** empty log message *** 2005-07-08 10:58 gb * [r1962] : exit if initial mmap fails 2005-07-06 18:20 gb * [r1961] : *** empty log message *** 2005-07-06 18:09 gb * [r1960] : Bump again, since I botched the checking comment the last time ... 2005-07-06 18:09 gb * [r1959] : Move the #+pppc64-target 64-bit arrays later in the typecase (some other things - like FUNCTION - are probably more common.) 2005-07-06 18:00 gb * [r1958] : Fix a few word-size assumptions in CALL-WITH-STRING-VECTOR; do pipe via libc on Darwin (since the syscall interface is pretty bizarre.) 2005-07-06 17:58 gb * [r1957] : Do STRIP-TAG-TO-FIXNUM in LAP (since I'm nervous about playing with (LOGAND -1)). 2005-07-06 17:57 gb * [r1956] : Define %%SCALE-SFLOAT for PPC64. 2005-07-06 17:56 gb * [r1955] : Early PROVIDE pushes (STRING MODULE) on *MODULES*. 2005-07-06 17:56 gb * [r1954] : Don't define key-mangling functions at compile-time (so that it's easier to experiment with them.) Mangle keys on PPC64, which seems to greatly improve GETHASH performance. 2005-07-06 17:54 gb * [r1953] : SCALE-FLOAT calls %%SCALE-SFLOAT on PPC64. 2005-07-06 02:20 gb * [r1952] : *** empty log message *** 2005-07-06 02:04 gb * [r1951] : natural ALU ops were ... missing; problems with inlined boxing/unboxing of 64-bit integers. 64-bit aref. 2005-07-06 01:56 gb * [r1950] : natural ALU ops take 32-bit args, return 32-bit results on PPC32. 2005-07-06 01:48 gb * [r1949] : natural-vref, natural-vset for ppc64. Handle 64-bit integer vectors. 2005-07-06 01:47 gb * [r1948] : Two-arg addition, subtraction on unsigned natural-sized integers. 2005-07-06 01:46 gb * [r1947] : NX-BINARY-NATURAL-OP-P may want to look at result type. 2005-07-06 01:44 gb * [r1946] : Error codes for not-u64, not-s64 were backwards. (continue from constant redefinition CERRORs, recompile everything). 2005-07-05 14:20 gb * [r1945] : *** empty log message *** 2005-07-05 13:25 gb * [r1944] : Need NUMBER-CASE-MACRO at compile-time. 2005-07-05 13:20 gb * [r1943] : *** empty log message *** 2005-07-05 13:12 gb * [r1942] : LOGBITP: fencepost on fixnum-only bound (can only access most significant bit of fixnum, not beyond ...). (Was looking at ppc32 constants, so this didn't lead to erroneous code, but may have been very inefficient in some cases.) 2005-07-05 13:11 gb * [r1941] : LOGBITP: fencepost on fixnum-only bound (can only access most significant bit of fixnum, not beyond ...). 2005-07-05 11:20 gb * [r1940] : *** empty log message *** 2005-07-05 10:25 gb * [r1939] : Bump. 2005-07-05 10:20 gb * [r1938] : *** empty log message *** 2005-07-05 10:19 gb * [r1937] : Add (IN-PACKAGE "CCL"). 2005-07-05 10:18 gb * [r1936] : Add (IN-PACKAGE "CCL"). Do SET-PACKAGE and FIND-PACKAGE here, so that IN-PACKAGE can be used in level-0 and early level-1 files. (All of these files are in the CCL package; the IN-PACKAGE forms are supposed to help text editors set the buffer package reasonably when editing these files.) 2005-07-05 10:12 gb * [r1935] : Add (IN-PACKAGE "CCL"). 2005-07-05 09:20 gb * [r1934] : *** empty log message *** 2005-07-05 08:52 gb * [r1933] : Add (IN-PACKAGE "CCL"). %LDB-FIXNUM-FROM-BIGNUM: byte may span as many as 3 digits, not just 2. 2005-07-05 08:51 gb * [r1932] : Add (IN-PACKAGE "CCL"). 2005-07-05 08:50 gb * [r1931] : REMOVE . before removing :UP in REMOVE-UP. 2005-07-05 08:50 gb * [r1930] : Move SET-PACKAGE stuff into level-0. 2005-07-05 08:49 gb * [r1929] : Integrate 64-bit vectors into type system better. 2005-07-05 08:49 gb * [r1928] : #+sparc-target ? no. 2005-07-05 08:48 gb * [r1927] : Recognize 64-bit traps, 64-bit BOUNDP traps. 2005-07-05 08:47 gb * [r1926] : Recognize LD instructions, for debugger. 2005-07-05 08:47 gb * [r1925] : Add an IN-PACKAGE; format/layout changes. 2005-07-05 08:46 gb * [r1924] : Nuke CLtL1 backward-compatibility stuff in IN-PACKAGE. 2005-07-05 08:45 gb * [r1923] : %ILSL : zero result if shift count > 63 (not 31). 2005-07-05 08:44 gb * [r1922] : Fix vcell locations in inlined lambda &optional/&key processing. 2005-07-05 05:20 gb * [r1921] : *** empty log message *** 2005-07-05 04:22 gb * [r1920] : no it wasn't 2005-07-05 04:20 gb * [r1919] : file was empty 2005-07-04 08:20 gb * [r1918] : *** empty log message *** 2005-07-04 07:55 gb * [r1917] : _SPbuiltin_length does list-length/circularity check on PPC64. _SPbuiltin_seqtype was checking (random) imm1 value, rather than typecode in imm0; this means that sequence functions sometimes tried to treat non-list arguments as vectors (rather than complaining about non-sequences.) (This seems to have been present on both platforms, and may have been long-standing.) 2005-07-04 07:52 gb * [r1916] : Use reentrant #_localtime_r. GCTIME uses (host) :timeval size, not hardwired constant. 2005-07-04 07:51 gb * [r1915] : Don't override field offsets assigned by interface translator! Assume that DarwinPPC64 uses natural PPC struct alignment. (AFAIK, Apple hasn't bothered to document the ABI yet.) 2005-07-04 07:49 gb * [r1914] : Only use character names for non-graphic-standard-chars (for #\Newline and for non-standard-chars) when writing chars with #\ syntax. *NAME-CHAR-ALIST*: reorder so that standard/semi-standard char names are preferred. 2005-07-04 07:47 gb * [r1913] : Make *TOTAL-GC-MICROSECONDS* big enough. 2005-07-04 07:46 gb * [r1912] : fix-big-logandc2: start complementing at second digit, since fixnum op handled low 2 digits. 2005-07-04 07:45 gb * [r1911] : EQL hashing treats +/- 0.0 the same. 2005-07-04 07:44 gb * [r1910] : %EXTEND-VECTOR: handle bit-vector case with non-zero start offset. 2005-07-04 07:43 gb * [r1909] : Functions can't (easily) overflow on PPC64. 2005-07-04 07:42 gb * [r1908] : NX1-ASH: (ash X 0) should typecheck X. 2005-07-04 07:41 gb * [r1907] : TRANSFORM-LOGOP: some logical operations aren't commutative. 2005-07-03 09:20 gb * [r1906] : *** empty log message *** 2005-07-03 08:38 gb * [r1905] : Try to schedule special-variable reference subprims a little better. (Hard, since we can't do speculative execution, but at least move a test further from a dependent branch.) 2005-07-03 08:36 gb * [r1904] : Single-digit LOGANDC1, LOGANDC2: store value in DEST if non-nil (was clobbering source bignum arg.) 2005-07-03 08:35 gb * [r1903] : Bignum LOGXOR does LOGXOR (not LOGIOR ...) on each digit. Bignun-X-fixnum LOGANDC1, LOGANDC2 assume that the fixnum op has set low 2 (32-bit) digits. 2005-07-03 08:32 gb * [r1902] : %DOUBLE-FLOAT-SIGN, %SHORT-FLOAT-SIGN back in LAP, return T iff sign bit is set. 2005-07-03 08:31 gb * [r1901] : Some paranoia in SXHASH-AUX; ensure that the "similar" -0.0 and 0.0 SXHASH to the same value on PPC64. 2005-07-03 08:30 gb * [r1900] : Move %DOUBLE-FLOAT-SIGN, %SINGLE-FLOAT-SIGN back to LAP (and make them return boolean valures, as their callers expect.) PPC64: LOG-E: coerce X to SINGLE-FLOAT before calling %SHORT-FLOAT-ABS on it. (Note: bugs like this may clobber the tcr's single-float convert slot, making it hard to box/unbox single-floats in that thread. If user code can do this (or if system code keeps being buggy ...), should make it easier to fix the tcr slot.) 2005-07-03 08:22 gb * [r1899] : (ASH x 0) - return X, but typecheck it 2005-07-03 08:21 gb * [r1898] : In default compiler policy, don't trust declarations or do unsafe things if SAFETY optimize quantity is 3. 2005-07-03 02:20 gb * [r1897] : *** empty log message *** 2005-07-03 02:01 gb * [r1896] : TARGET-ARCH-CASE for some hash-table macros. (l0-hash.lisp functions have been a little confused about types/tags.) 2005-07-03 02:00 gb * [r1895] : Store low 32 bits, not high 32 when storing into a (SIMPLE-ARRAY (SIGNED-BYTE 32) (*)). 2005-07-03 01:59 gb * [r1894] : Avoid "%08x" in printf format strings, when size of arg isn't known to be 32 bits. 2005-07-03 01:57 gb * [r1893] : Lisp image du jour. 2005-07-03 01:56 gb * [r1892] : Make a real (SETF MACRO-FUNCTION) function; make the old SET-MACRO-FUNCTION call it. (Note that this may be a little tricky to bootstrap: it may be necessary to do: ? (remhash 'macro-function ccl::%setf-methods%) to remove the old SETF inverse established by DEFSETF.) 2005-07-03 01:54 gb * [r1891] : So many mis-indented comments, so little time. 2005-07-03 01:52 gb * [r1890] : Make TYPE-ERROR objects with implicit EXPECTED-TYPE slots if no type info available. 2005-07-03 01:51 gb * [r1889] : target package in %SLOT-UNBOUND-TRAP. 2005-07-03 01:50 gb * [r1888] : APPEND has something to do with the CL definition ... 2005-07-03 01:49 gb * [r1887] : LOGTEST of 2 bignums looks at both args (rather than LOGTESTing the first with itself.) LOGTEST of fixnum and bignum looks at second 32-bit digit if it exists and no intersection on first digit. 2005-07-03 01:47 gb * [r1886] : %SET-OBJECT stores a natural-sized value, not a 32-bit word ... 2005-07-03 01:46 gb * [r1885] : PROPER-LIST-P returns false on dotted lists (as someone may have suggested a few months ago.) 2005-07-03 01:46 gb * [r1884] : Comment formatting changes only. 2005-07-02 23:20 gb * [r1883] : *** empty log message *** 2005-07-02 22:52 gb * [r1882] : PPC64-specific changes: Get upper bound of (UNSIGNED-BYTE 64) right. 2005-07-02 22:32 gb * [r1881] : PPC64-specific changes: makeu64: make a 3-digit bignum if we need 3 digits misc_set_common: get type-checking right on (SIGNED-BYTE 16), (UNSIGNED-BYTE 64). 2005-07-02 22:31 gb * [r1880] : ; -> ;;; on section comments. 2005-07-02 21:20 gb * [r1879] : *** empty log message *** 2005-07-02 20:42 gb * [r1878] : FIND-LOCAL-NAME: encoded stack-frame addresses are a function of target::word-size. 2005-07-02 12:20 gb * [r1877] : *** empty log message *** 2005-07-02 11:51 gb * [r1876] : %MULTIPLY returns (values "high 32 bits of product" low 32 bits of product") in that order. (%MULTIPLY is currently only used to validate the "guess" digit in BIGNUM-TRUNCATE's aux functions, so getting things completely backwards here only caused some bad guess digits to be accepted.) 2005-07-02 11:47 gb * [r1875] : Some comments had been line-wrapped (presumably causing some words to become lap labels). 2005-07-02 11:45 gb * [r1874] : 64-bit %ROUND-NEAREST-SHORT-FLOAT->FIXNUM stores FP0 in the right place ... 2005-07-01 22:20 gb * [r1873] : *** empty log message *** 2005-07-01 22:00 gb * [r1872] : %ADD-WITH-CARRY and %SUBTRACT-WITH-BORROW operate on the low 32 bits of their arguments (e.g, -1 is #x00000000ffffffff)). 2005-07-01 10:20 gb * [r1871] : *** empty log message *** 2005-07-01 09:55 gb * [r1870] : ppc64 bug fixes. 2005-07-01 09:53 gb * [r1869] : use T vice (+ ppc64::nil-value ppc64::t-offset) in EQL. 2005-07-01 09:52 gb * [r1868] : Bug fixes. 2005-07-01 09:50 gb * [r1867] : Indentation. 2005-07-01 09:49 gb * [r1866] : yet another way to change default directory (CCL::CD). 2005-07-01 09:48 gb * [r1865] : SET-MACRO-FUNCTION accepts/ignores env arg. 2005-07-01 09:48 gb * [r1864] : Remove some comments. 2005-07-01 09:47 gb * [r1863] : Check for sign-extension via arithmetic shift (not logical shift) in makes128. 2005-07-01 09:46 gb * [r1862] : Comment indentation. 2005-07-01 09:46 gb * [r1861] : Several bug fixes ... 2005-07-01 09:43 gb * [r1860] : ADJUST-SP was missing. 2005-07-01 09:43 gb * [r1859] : Handle LABELS declarations differently; this may need more thought/work. 2005-07-01 01:20 gb * [r1858] : *** empty log message *** 2005-07-01 00:36 gb * [r1857] : bump. 2005-07-01 00:20 gb * [r1856] : *** empty log message *** 2005-06-30 23:47 gb * [r1855] : Zero the CTR when entering/returning from lisp code. Fix valence when reentering lisp with exception-handler swap. 2005-06-30 23:46 gb * [r1854] : commit_pages() may fail, so it returns Boolean. This needs work (overcommit, etc.) 2005-06-30 23:45 gb * [r1853] : Use signed alloc traps. 2005-06-30 23:45 gb * [r1852] : resize_dyamic_heap() may fail, so it returns a Boolean. Define new alloc trap instructions; contunue to define old ones, for bootstrapping. 2005-06-30 23:43 gb * [r1851] : Recognize old (unsigned) alloc traps as well as new (signed) ones. Check for failure in memory allocation requests; fixup allocptr before signalling lisp errors in case of failure. 2005-06-30 23:41 gb * [r1850] : What exactly is a "locatitive" ? 2005-06-30 23:40 gb * [r1849] : Compute an effective slot-definition's type predicate based on the value of its type slot, not the (possibly unsupplied) :TYPE initarg. 2005-06-30 23:39 gb * [r1848] : Recognize (UNSIGNED-BYTE 56) as an expected type in some UUO type errors. 2005-06-30 23:38 gb * [r1847] : Slot-definition's TYPE slots default to T, damn it. Compute the intersection of all direct slots' types when determining an effective slot's type. (We should probably complain if the intersection is empty, and there may be cases where it can't be computed.) 2005-06-30 23:34 gb * [r1846] : Allow BIGNUM-REF/BIGNUM-SET to inline. 2005-06-30 23:34 gb * [r1845] : Use signed alloc traps. (This means that valid heap addresses are limited to the low half of the address space; this has been true because of OS limitations.) 2005-06-30 23:32 gb * [r1844] : (partially) revive %TYPED-MISCREF and %TYPED-MISCSET. 2005-06-30 23:31 gb * [r1843] : Recognize SRDI. 2005-06-29 23:20 gb * [r1842] : *** empty log message *** 2005-06-29 22:52 gb * [r1841] : bump. 2005-06-29 22:30 gb * [r1840] : %ilogand2: use clear-left, clear-right if appropriate. NOTE: clear-left doesn't ensure that tag bits are cleared; code which says (logand (the fixnum x) mask) - and lies about the type of X - is unsafe, but any such code is pretty bogus to begin with. 2005-06-29 22:20 gb * [r1839] : *** empty log message *** 2005-06-29 22:12 gb * [r1838] : CLEAR-LEFT and CLEAR-RIGHT. 2005-06-29 21:20 gb * [r1837] : *** empty log message *** 2005-06-29 21:06 gb * [r1836] : Reorganize TRUNCATE and related functions. 2005-06-29 21:04 gb * [r1835] : %SUB-TIMEVALS in lisp. 2005-06-29 21:03 gb * [r1834] : Kinder, gentler %usedbytes. 2005-06-29 21:02 gb * [r1833] : Comment indentation. 2005-06-29 21:01 gb * [r1832] : Fix typos in conditionalization. 2005-06-29 21:00 gb * [r1831] : Missing close-paren in comment. 2005-06-29 20:59 gb * [r1830] : lisp_heap_threshold, egc thresholds are platform-specific (twice as large on 64-bit.) Enable EGC on PPC64. 2005-06-29 20:58 gb * [r1829] : PPC64 codevector header stuff. When handling an allocation trap, ensure that displacement is a "signed natural" (signed int of natural word size), not an "int" (signed-byte 32). Don't explicitly zero pages that should already be zeroed in new_heap_segment; the PPC64-specific bug was elsewhere. 2005-06-29 20:54 gb * [r1828] : If EGC memoization is missing, say "doublenode" instead of "doubleword". When masking out bits at the end of the "refbits" bit vector, ensure that C does 64-bit arithmetic. 2005-06-29 20:52 gb * [r1827] : heap_segment_size (and log2 of it) are platform-specific. Make many constants explicitly "long", to discourage 32-bit arithmetic. 2005-06-29 20:50 gb * [r1826] : heap_segment_size (and log2 of it) are platform-specific. 2005-06-29 20:49 gb * [r1825] : Do SUB-TIMEVALS in lisp. (Did this cons or something ? Why was it in LAP ?) 2005-06-29 20:48 gb * [r1824] : Remove stray DBG call. 2005-06-17 11:20 gb * [r1823] : *** empty log message *** 2005-06-17 10:48 gb * [r1822] : s16 and u16 cases were backwards in misc_ref. 2005-06-17 10:48 gb * [r1821] : EGC not ready on PPC64 yet. 2005-06-17 10:47 gb * [r1820] : misc_dloat_offset isn't misc_header_offset ... double_float_header reflects the view of a double-float as two 32-bit elements. 2005-06-17 10:45 gb * [r1819] : Sigh. May not be ppc32, use target package. 2005-06-17 10:44 gb * [r1818] : Define and use XP-GPR-SIGNED-DOUBLEWORD. 2005-06-17 10:43 gb * [r1817] : Don't do APPEND two lists at a time ... 2005-06-17 10:29 gb * [r1816] : Don't reuse NVRs in hard case of %EXTEND-VECTOR on either platform. 2005-06-16 23:20 gb * [r1815] : *** empty log message *** 2005-06-16 23:13 gb * [r1814] : Fix ppc64 bignum case of integer-sign. Some stores and loads (FP constants, fpscr, etc.) need to be 32-bit, not ldr/str. 2005-06-16 23:11 gb * [r1813] : Just say "%lx", not %08lx". 2005-06-16 23:11 gb * [r1812] : Forgot to say "%lx" in some printf calls. 2005-06-16 23:10 gb * [r1811] : c_reg_save has room for N nodes, not N 4-byte things. 2005-06-16 23:08 gb * [r1810] : Define and use %LOGXOR, and set carry on entry to negation loop. 2005-06-16 23:07 gb * [r1809] : ARRAY-RANK-LIMIT, CALL-ARGUMENTS-LIMIT, LAMBDA-PARAMETERS-LIMIT are functions of target::node-size. 2005-06-16 12:20 gb * [r1808] : *** empty log message *** 2005-06-16 11:44 gb * [r1807] : %truncate-double-float for PPC64. 2005-06-16 11:43 gb * [r1806] : INIT-RANDOM-STATE-SEEDS on both platforms. 2005-06-16 11:41 gb * [r1805] : Use old (safe-but-slow) bignum GCD routine on PPC64, for now. Define ONE-BIGNUM-FACTOR-OF-TWO. 2005-06-16 11:39 gb * [r1804] : 050616. 2005-06-16 11:38 gb * [r1803] : Don't try to determine GCC version from "reading specs from ..." verbose output if specs aren't read from a file (assume modern GCC in that case.) 2005-06-16 11:36 gb * [r1802] : Enable EGC on PPC64. 2005-06-16 11:36 gb * [r1801] : Zero enough bits (shift by "bitmap_shift", not "5") when tenuring to an area other than the oldest. 2005-06-16 06:20 gb * [r1800] : *** empty log message *** 2005-06-16 05:41 gb * [r1799] : Bump fasl version. 2005-06-16 05:41 gb * [r1798] : Remove embedded control char. 2005-06-16 05:20 gb * [r1797] : *** empty log message *** 2005-06-16 05:16 gb * [r1796] : Bump; claim "Alpha" status for PPC64. 2005-06-16 05:15 gb * [r1795] : Fixes in %INIT-MISC. 2005-06-16 05:15 gb * [r1794] : %MAKE-CODE-EXECUTABLE fixes for PPC64. 2005-06-16 05:14 gb * [r1793] : Word-size fixes. 2005-06-16 05:13 gb * [r1792] : don't do unsigned divide in %FIXNUM-TRUNCATE. Bad %NEXT-RANDOM-SEED is arguably better than none. 2005-06-16 05:12 gb * [r1791] : Fix in #+ppc64-target walk-static-area. 2005-06-16 05:11 gb * [r1790] : Lots-o-bug-fixes. 2005-06-16 05:10 gb * [r1789] : specialized-array-element-types include ppc64 64-bit array types. 2005-06-16 05:09 gb * [r1788] : RANDOM-STATE changes for ppc64; %single-float-atanh for ppc64. 2005-06-16 05:08 gb * [r1787] : Set up ppc64 signed/unsigned 64-bit vectors in class table. 2005-06-16 05:07 gb * [r1786] : Define image-base-address per-backend. 2005-06-16 05:06 gb * [r1785] : New fasl ops for 32-bit, 64-bit signed integers. Image-base-address determined by backend. 2005-06-16 05:06 gb * [r1784] : New fasl ops for 32-bit, 64-bit signed integers. 2005-06-16 05:04 gb * [r1783] : Remove (obsolete) reference defs. 2005-06-16 05:03 gb * [r1782] : Ignore 64-bit FASL files. 2005-06-16 05:02 gb * [r1781] : New file. 2005-06-16 05:01 gb * [r1780] : Need to check gcc version in some other way (actually, we can assume >= 4.0 for DarwinPPC64.) 2005-06-16 05:00 gb * [r1779] : So, Apple ? What's with all the SIGTTIN's ? 2005-06-16 04:59 gb * [r1778] : Fix bit vectore ref/set, small-bignum initialization after shifts/multiplies for ppc64. 2005-06-16 04:58 gb * [r1777] : Print some more uvector types in kernel debugger. 2005-06-16 04:57 gb * [r1776] : Create much larger reserved area on ppc64. Fix handling of HeapHighWaterMark, especially in cases when EGC is off. 2005-06-16 04:56 gb * [r1775] : Ensure that _align_to_power_of_2 DTRT on ppc64. 2005-06-16 04:54 gb * [r1774] : Word-size changes. 2005-06-16 04:49 gb * [r1773] : (S) command to find/describe symbol with matching pname. 2005-06-16 04:49 gb * [r1772] : PPC64 code-vector-prefix. 2005-06-16 04:48 gb * [r1771] : Lots of word-size fixes. 2005-06-16 04:47 gb * [r1770] : 64-bit IMAGE_BASE_ADDRESS for ppc64. 2005-06-16 04:46 gb * [r1769] : APPLY+ changes for PPC64. 2005-06-16 04:45 gb * [r1768] : Use target::fixnumshift in PRINT-OBJECT method on RANDOM-STATE. 2005-06-16 04:44 gb * [r1767] : Dump small integers (signed-byte 16, 32, 64) via specialized ops. 2005-06-16 04:20 gb * [r1766] : *** empty log message *** 2005-06-16 04:14 gb * [r1765] : Parameterize %ilogbitp, other things for ppc64. 2005-06-16 04:13 gb * [r1764] : :cdb-datum.size is (:unsigned 32), not :unsigned-long. 2005-06-16 04:12 gb * [r1763] : Portability changes. 2005-06-16 04:11 gb * [r1762] : Conditionalize REGISTERS-USED-BY. 2005-06-16 04:11 gb * [r1761] : target: package changes. 2005-06-16 04:10 gb * [r1760] : Remove commented-out code. 2005-06-16 04:09 gb * [r1759] : ONE-BIGNUM-FACTOR-OF-TWO, LOGBITP here for PPC32. 2005-06-16 04:06 gb * [r1758] : primsizes array isn't of (non-portable) element-type 'fixnum. New fasl ops for s32, s64. Count package htab elements based on SYMBOLP only when resizing. 2005-06-16 04:02 gb * [r1757] : Fix NUMBERP typo. 2005-06-16 04:02 gb * [r1756] : Move ONE-BIGNUM-FACTOR-OF-TWO, LOGBITP elsewhere. 2005-06-16 04:01 gb * [r1755] : Look at the right area slots in %RESERVEDBYTES. 2005-06-16 03:58 gb * [r1754] : Indentation and comment changes. 2005-06-16 03:57 gb * [r1753] : ppc64 fixes. 2005-06-16 03:55 gb * [r1752] : Several fixes, mostly in bit addressing. 2005-06-16 03:54 gb * [r1751] : UNBOX-BASE-CHAR fixes for PPC64; 16-bit digit-part accessors are ppc32-only. 2005-06-16 03:53 gb * [r1750] : :RAW option to PPC-XDISASSEMBLE, ppc-gpr handles target rcontext. 2005-06-16 03:52 gb * [r1749] : Check platform-specific "naturalness" in comparisons. 2005-06-13 21:20 gb * [r1748] : *** empty log message *** 2005-06-13 21:17 gb * [r1747] : minor_opcode_TR cut-and-paste mishap. 2005-06-12 22:20 gb * [r1746] : *** empty log message *** 2005-06-12 21:29 gb * [r1745] : Missed another thing (in %USE-PACKAGE-CONFLICT-CHECK) that accesses package hash tables directly (and was doing so incorreclty, per the New Order.) 2005-06-10 22:20 gb * [r1744] : *** empty log message *** 2005-06-10 22:13 gb * [r1743] : sigaction flags, node_size changes in thread startup. 2005-06-10 22:12 gb * [r1742] : start_lisp: save/restore C registers, build C stack frame in node_size multiples (not 4-byte multiples.) 2005-06-10 22:11 gb * [r1741] : Print (immedate) SINGLE-FLOATs on ppc64. 2005-06-10 22:09 gb * [r1740] : sigaction flags. 2005-06-10 22:08 gb * [r1739] : Define MIN_KERNEL_GLOBAL. 2005-06-10 22:08 gb * [r1738] : We seem to be committed to the "new" DarwinSigReturn. Add a little bit of abstraction for trap opcodes. 2005-06-10 22:04 gb * [r1737] : Decode some traps more portably; still some work to be done here. 2005-06-10 22:02 gb * [r1736] : Word-size issues; use MIN_KERNEL_GLOBAL rather than platform-dependent constant. 2005-06-10 21:58 gb * [r1735] : Lots of word-size issues. 2005-06-10 21:57 gb * [r1734] : Split tcr.flags into two 32-bit words, because the .h file does so. 2005-06-10 21:56 gb * [r1733] : Always do N-ary +/- as a sequence of 2-arg operations. 2005-06-10 21:55 gb * [r1732] : Comment indentation. 2005-06-10 21:20 gb * [r1731] : *** empty log message *** 2005-06-10 21:18 gb * [r1730] : Version de jour. 2005-06-10 21:16 gb * [r1729] : target:: package prefix changes. 2005-06-10 21:16 gb * [r1728] : HOST-PLATFORM returns second (64-bit-p) value; PLATFORM-DESCRIPTION decodes that, LISP-IMPLEMENTATION-VERSION uses that info. 2005-06-10 21:13 gb * [r1727] : Bug fixes ... 2005-06-10 21:12 gb * [r1726] : Use trlgei in (IIRC) %%SAVE-APPLICATION trap. 2005-06-10 21:11 gb * [r1725] : Move some #+ppc64 code to lisp. 2005-06-10 21:09 gb * [r1724] : Finish package changes. 2005-06-10 21:08 gb * [r1723] : Conditionalize ARRAY-TOTAL-SIZE-LIMIT; add *features* for word size, endianness. 2005-06-10 21:07 gb * [r1722] : Some 64-bit conditionalization. 2005-06-10 21:04 gb * [r1721] : Need to define ONE-DIGIT-BIGNUM-HEADER after all. 2005-06-10 21:04 gb * [r1720] : SET-GLOBAL sets all 64 bits on PPC64. 2005-06-09 22:20 gb * [r1719] : *** empty log message *** 2005-06-09 21:29 gb * [r1718] : Compile all of target-xdev-modules, not just the cdr of the list (why was it ever skipping the car of that list ???) 2005-06-08 06:20 gb * [r1717] : *** empty log message *** 2005-06-08 05:35 gb * [r1716] : Initial element in cloned package htabs is 0, too. New binaries. 2005-06-08 05:20 gb * [r1715] : *** empty log message *** 2005-06-08 04:49 gb * [r1714] : 050607, once in a lifetime. 2005-06-08 04:48 gb * [r1713] : Finish package bootstrapping (unless I missed something): package hash tables contain symbols (NILSYM on PPC32), 0 (unused/free entries), or (PACKAGE-DELETED-MARKER)s. 2005-06-08 03:20 gb * [r1712] : *** empty log message *** 2005-06-08 03:06 gb * [r1711] : Don't stop in GDB on SIGKILL. 2005-06-08 03:06 gb * [r1710] : ASH can return 64-bit results, as can fixnum multiplication. 2005-06-08 03:05 gb * [r1709] : 64-bit changes. 2005-06-08 03:04 gb * [r1708] : define some PPC64 opcodes/masks. 2005-06-08 03:03 gb * [r1707] : Propagate exceptions on Darwin. 2005-06-08 03:02 gb * [r1706] : 64-bit changes/fixes. Ensure that area_dnode returns natural. 2005-06-08 03:01 gb * [r1705] : 64-bit changes/fixes. 2005-06-08 03:00 gb * [r1704] : 4&5-bit bignum headers. 2005-06-08 02:59 gb * [r1703] : TCR_FLAG_BIT_PROPAGATE_EXCEPTION. 2005-06-08 02:59 gb * [r1702] : 64-bit changes. 2005-06-08 02:57 gb * [r1701] : Use target::target-most-positive-fixnum when applicable. 2005-06-08 02:57 gb * [r1700] : Decode and handle 64-bit traps. 2005-06-08 02:56 gb * [r1699] : Use PACKAGE-DELETED-MARKER. 2005-06-08 02:55 gb * [r1698] : COLUMN moved here. 2005-06-08 02:54 gb * [r1697] : Recognize 64-bit bogus-objects. 2005-06-08 02:53 gb * [r1696] : lower-case 1 char. 2005-06-08 02:53 gb * [r1695] : Remove some duplication. 2005-06-08 02:52 gb * [r1694] : Remove some commented-out code. 2005-06-08 02:51 gb * [r1693] : 64-bit (DarwinPPC) callbacks. 2005-06-08 02:50 gb * [r1692] : Move COLUMN definition earlier. 2005-06-08 02:47 gb * [r1691] : PACKAGE-DELETED-MARKER, some comment/formatting changes. 2005-06-08 02:46 gb * [r1690] : Bug fixes. 2005-06-08 01:20 gb * [r1689] : *** empty log message *** 2005-06-08 00:36 gb * [r1688] : Comment formatting. 2005-06-08 00:35 gb * [r1687] : GC returns target::nil-value. 2005-06-08 00:34 gb * [r1686] : Do an unsigned divide; there may be another problem here (this is almost certainly wrong.) 2005-06-08 00:31 gb * [r1685] : Start to recognize 0 as an unused package htab entry. 2005-06-08 00:30 gb * [r1684] : target-package stuff. 2005-06-08 00:29 gb * [r1683] : Some more target conditionalization. 2005-06-08 00:29 gb * [r1682] : INTEGER-DECODE-DOUBLE-FLOAT doesn't have to make bignums on ppc64. 2005-06-08 00:27 gb * [r1681] : trap-unless-list traps if tag is unsignged< or unsigned> ppc64::fulltag-cons, to distinguish itself from trap-unless-cons cases. 2005-06-08 00:25 gb * [r1680] : tag-imm-2. 2005-06-08 00:25 gb * [r1679] : Target-specific shift counts. 2005-06-08 00:24 gb * [r1678] : LOCKP, INTEGERP use target types. 2005-06-08 00:23 gb * [r1677] : Parameterize ASH for target fixnum size. 2005-06-03 15:20 gb * [r1676] : *** empty log message *** 2005-06-03 15:10 gb * [r1675] : Bump: 050603. 2005-06-03 15:09 gb * [r1674] : 32/64-bit versions of some utilities; need to do (gdb) set $ppc64=1 manually to enable 64-bit versions. 2005-06-03 15:08 gb * [r1673] : No more *LDS*. 2005-06-03 15:08 gb * [r1672] : Target:: package changes, code-vector prefix. 2005-06-03 15:07 gb * [r1671] : Target:: package changes. 2005-06-03 15:05 gb * [r1670] : No more LDS/*LDS*. 2005-06-03 15:02 gb * [r1669] : Remove some obsolete/archaic stuff. 2005-06-03 15:01 gb * [r1668] : Some target-package/cross-compilation stuff. 2005-06-03 15:01 gb * [r1667] : Heap-ivector, other target-package fixes. 2005-06-03 15:00 gb * [r1666] : Don't use (archaic) LDS macro. 2005-06-03 14:55 gb * [r1665] : Node size when getting command-line arguments. Hide some pathname creation forms from COMPILE-FILE toplevel processing (may not be necessary now that cross-compilation binds *target-backend*.) 2005-06-03 14:53 gb * [r1664] : LDB: look at target-package constants. 2005-06-03 14:52 gb * [r1663] : Lots of target-package changes. 2005-06-03 14:51 gb * [r1662] : Use the right target-package symbols to initialize MOST-POSITIVE-FIXNUM and MOST-NEGATIVE-FIXNUM. 2005-06-03 14:50 gb * [r1661] : Hide toplevel MAKE-ARRAY call (in *FORMAT-CHAR-TABLE* initial value); may not be necessary now that cross-compilation binds *target-backend*. 2005-06-03 14:49 gb * [r1660] : Initialize *.fasl-pathname* correctly. Hide toplevel macro call (in *default-pathname-defaults* initial value); may not be necessary now that cross-compilation binds *target-backend*. 2005-06-03 14:47 gb * [r1659] : Remove a (meaningless) RESIDENT declaration. 2005-06-03 14:46 gb * [r1658] : %STANDARD-INSTANCE-P uses TARGET-ARCH-CASE. 2005-06-03 14:45 gb * [r1657] : Fix PPC64 ivector class names. Lots of formatting changes. 2005-06-03 14:44 gb * [r1656] : More conditionalized pathnames. 2005-06-03 14:43 gb * [r1655] : Some ppc64 changes + fixes. 2005-06-03 14:41 gb * [r1654] : CLASS-OF, heap-walking fixes. 2005-06-03 14:40 gb * [r1653] : Don't bother conditionalizing pname hashes (32 bits is plenty, and we need to rehash after cross-compilation anyhow.) 2005-06-03 14:39 gb * [r1652] : PPC64 EQUAL: recover NFN from FN after recursing. 2005-06-03 14:38 gb * [r1651] : Fake %NEXT-RANDOM-SEED for ppc64. 2005-06-03 14:37 gb * [r1650] : Lots of fixes in 64-bit code. 2005-06-03 14:35 gb * [r1649] : Move some stuff used to do integer truncation elsewhere. 2005-06-03 14:32 gb * [r1648] : All PPC targets seem to have the same set of level-1 modules. Compile target-xdev modules when cross-compiling. Bind *TARGET-BACKEND* when cross-compiling, so that macros that use TARGET-ARCH-CASE or similar dtrt. 2005-06-03 14:30 gb * [r1647] : Comment/formatting changes only. 2005-06-03 14:29 gb * [r1646] : Small bignums have two (32-bit) words on PPC64. 2005-06-03 14:28 gb * [r1645] : No more *FRED-SPECIAL-INDENT-ALIST*. (Might someday want something similar.) 2005-06-03 14:27 gb * [r1644] : Word size issues, target package. 2005-06-03 14:26 gb * [r1643] : Multiple subdirs. 2005-06-03 14:24 gb * [r1642] : Lots of bug fixes, 64-bit changes. 2005-06-03 14:22 gb * [r1641] : Fix character printing (use charcode_shift), NIL printing on PPC64. 2005-06-03 14:21 gb * [r1640] : Disable EGC on PPC64, until it works. 2005-06-03 14:20 gb * [r1639] : Print addresses with "l" qualifier. 2005-06-03 14:20 gb * [r1638] : *** empty log message *** 2005-06-03 14:19 gb * [r1637] : Quiet breakpoint trap, start to abstract 64-bit traps. 2005-06-03 14:18 gb * [r1636] : Suppress warnings when mapping between TCRs and exception ports. Recognize "quiet" lisp-breakpoints (no possibility of C-string messages, no changes to LR/PC, etc.) 2005-06-03 14:16 gb * [r1635] : Start to learn about 64-bit traps, use %lx (etc) when printing 64-bit values. 2005-06-03 14:14 gb * [r1634] : Comments, formatting changes. 2005-06-03 14:12 gb * [r1633] : New file. 2005-06-03 14:11 gb * [r1632] : Lots of word-size and other ppc64 changes. 2005-06-03 14:10 gb * [r1631] : Implement %extend-vector for ppc64. 2005-06-03 14:09 gb * [r1630] : Lots-o-changes. 2005-06-03 14:08 gb * [r1629] : Use target package prefix. 2005-06-03 14:04 gb * [r1628] : Very minor change in %ENSURE-SVAR. 2005-06-03 14:03 gb * [r1627] : %TYPE-OF for PPC64. 2005-06-03 14:02 gb * [r1626] : target-package, some minor changes. 2005-06-03 14:01 gb * [r1625] : Some cleanup/conditionalization of *FEATURES*. 2005-06-03 14:00 gb * [r1624] : Don't do unsafe fixnum arithmetic to "mixup" or "rotate" hash codes on ppc64. 2005-06-03 13:59 gb * [r1623] : Fix a few typos. 2005-06-03 13:58 gb * [r1622] : Move functionality into target-specific subdirectory. 2005-06-03 13:52 gb * [r1621] : Lots of fixes, more consistent use of 64-bit rotates/shifts. 2005-06-03 13:50 gb * [r1620] : Only require PPC64-VINSNS if we're the native target. Change the Darwin architecture identifier. 2005-06-03 13:49 gb * [r1619] : Only require PPC32-VINSNS if we're the native target. 2005-06-03 13:48 gb * [r1618] : Some 64-bit fixes. 2005-06-03 13:47 gb * [r1617] : BKPT trap; a few fixes (including 64-bit INT-TO-FREG.) 2005-06-03 13:45 gb * [r1616] : CLRLSLDI[.] had arguments inverted. 2005-06-03 13:45 gb * [r1615] : SYMBOLP. 2005-06-03 13:43 gb * [r1614] : NX-BINARY-FIXNUM-OP-P is sensitive to target fixnum size. 2005-05-27 21:20 gb * [r1613] : *** empty log message *** 2005-05-27 20:44 gb * [r1612] : 050527a. 2005-05-27 20:43 gb * [r1611] : rcontext defined per-architecture. 2005-05-27 20:43 gb * [r1610] : rcontext = r13. 2005-05-27 20:41 gb * [r1609] : rcontext = r2. Fix several typos/inconsistencies. 2005-05-27 20:40 gb * [r1608] : Print wide registers (needs some tweaking.) 2005-05-27 20:38 gb * [r1607] : Word-size fixes; preserve r13 when setting up pseudo-signal trampoline on ppc64. 2005-05-27 20:37 gb * [r1606] : Define ALLOCPTR_TRAP for PPC64. 2005-05-27 20:36 gb * [r1605] : Use .machine directive in _beginfile. Define stack-frame-type equates. 2005-05-27 20:34 gb * [r1604] : header_length looks at all of the bits, as does extract_unsigned_byte_bits_. 2005-05-27 20:33 gb * [r1603] : Use dlsym() on Darwinppc64; print wide addresses on ppc64. 2005-05-27 20:32 gb * [r1602] : Remap the subprims jump table on Darwinppc64 2005-05-27 20:31 gb * [r1601] : Print "natural-sized" integers. 2005-05-27 20:23 gb * [r1600] : Lots of word-size fixes, flesh out misc-set, deal with Tiger ld64/as64 alignment crap. (Filed a bug.) 2005-05-27 20:22 gb * [r1599] : Add, export labels, so that we can work around Tiger ld64 lossage. 2005-05-27 20:21 gb * [r1598] : Use lwz to load each half of fp conversion constant. (Hmmm. May need another constant for 64-bit conversions.) 2005-05-27 20:20 gb * [r1597] : *** empty log message *** 2005-05-27 20:19 gb * [r1596] : Lots of word-size/tagging conditionalization. 2005-05-27 20:18 gb * [r1595] : PPC-LAP-WORD (this version, at least ...) accounts for code-vector prefix. 2005-05-27 20:17 gb * [r1594] : Load 64-bit fasls. 2005-05-27 20:16 gb * [r1593] : Fix constant handling, slightly. 2005-05-27 20:16 gb * [r1592] : Handle code-vectors more portably. 2005-05-27 20:15 gb * [r1591] : A few word-size fixes; show addresses of cold-load functions, for debugging. 2005-05-27 20:14 gb * [r1590] : Give up (for the time being) on the notion of getting Apple's ld64 to do anything useful. 2005-05-27 20:12 gb * [r1589] : Bugs. 2005-05-27 20:11 gb * [r1588] : rcontext = r2. Not TCR bias; no low words in TCR def. Fix kernel-global, nrs-offset. 2005-05-27 20:10 gb * [r1587] : SCALE-NODE-MISC-OFFSET. 2005-05-27 20:09 gb * [r1586] : rcontext = r13 for ppc32. 2005-05-27 20:08 gb * [r1585] : fix lots of word-size assumptions in stack-unwinding, lambda-list processing. 2005-05-27 20:07 gb * [r1584] : rcontext is target-dependent fix int-to-freg for ppc64. 2005-05-27 20:05 gb * [r1583] : r2, r16: usage is target-dependent. (Need to tell disassembler about this.) 2005-05-24 06:20 gb * [r1582] : *** empty log message *** 2005-05-24 06:04 gb * [r1581] : Move image up a page. 2005-05-24 06:03 gb * [r1580] : Fix botched #+ conditionalization. 2005-05-24 06:01 gb * [r1579] : Increment FASL version. 2005-05-24 06:00 gb * [r1578] : Don't pad jump table by extra page; revert to pc-relative branches. 2005-05-24 06:00 gb * [r1577] : Don't copy spjump table anymore. 2005-05-24 05:59 gb * [r1576] : Adjust image start address. 2005-05-24 05:36 gb * [r1575] : New binaries soon. 2005-05-24 05:29 gb * [r1574] : 64-bit xloader, cross-xloader. New image header stuff for PPC64. 2005-05-24 05:28 gb * [r1573] : Mach Header too far away from old spjump address, so we'll have to move the spjump table. 2005-05-24 05:28 gb * [r1572] : Duplicate the subprims jump table (bootstrapping kludge.) Observe word size when vpushing initial function. 2005-05-24 05:26 gb * [r1571] : Use STRU when allocating TSP frames. 2005-05-24 05:25 gb * [r1570] : Fix bignum boxing for __ifdef([PPC64]) case. 2005-05-24 05:24 gb * [r1569] : Do absolute jumps, to make relocation of the jump table itself simpler. 2005-05-24 05:24 gb * [r1568] : Remove some very old stuff. 2005-05-24 05:23 gb * [r1567] : Fix at least one Mach exception-handling case. 2005-05-24 05:22 gb * [r1566] : Allow loading 64-bit images. 2005-05-24 05:21 gb * [r1565] : Load PPC64 images. 2005-05-24 05:21 gb * [r1564] : Fix t_offset def. Align TCR fields correctly. 2005-05-24 05:20 gb * [r1563] : area_codes are native fixnums. 2005-05-24 05:20 gb * [r1562] : Call non-destructive single-float stuff on ppc64. 2005-05-24 05:20 gb * [r1561] : *** empty log message *** 2005-05-24 05:19 gb * [r1560] : Fix compile-time compilation (env/function name screwup). Fasdump bignums using new fasl-op. 2005-05-24 05:17 gb * [r1559] : FIND-LOCAL-NAME: fix some incorrect stack-frame layout assumptions. 2005-05-24 05:15 gb * [r1558] : Lots of #+ppc64 bug fixes and conditionalization. 2005-05-24 05:14 gb * [r1557] : More non-destuctive short/single-float stuff. 2005-05-24 05:14 gb * [r1556] : LOAD-DOUBLE-FLOAT-CONSTANT in 2 32-bit halves. 2005-05-24 05:13 gb * [r1555] : Double-float -cell indices. 2005-05-24 05:12 gb * [r1554] : Renamed syscall stuff. 2005-05-24 05:12 gb * [r1553] : SINGLE-FLOAT-CONVERT: use TCR. prefix. 2005-05-24 05:11 gb * [r1552] : Subprims jump table moved from 0x4000 to 0x5000 (at the moment, it's both places.) 2005-05-24 05:10 gb * [r1551] : Rename syscall operators. 2005-05-24 05:10 gb * [r1550] : Use (new, renamed) POWEROPEN-SYSCALL for both PPC64 platforms. 2005-05-24 05:08 gb * [r1549] : Define more error/UUO codes. 2005-05-20 14:20 gb * [r1548] : *** empty log message *** 2005-05-20 13:38 gb * [r1547] : 050520b. 2005-05-20 13:38 gb * [r1546] : System calls defined in OS-specific hash tables, not in the FTD. System call indices are in the "SYSCALLS" package, not "OS". (These changes make it a little easier to cross-compile between different architectures on the same OS, and perhaps a little more awkward to cross-compile between OSes.) 2005-05-20 10:20 gb * [r1545] : *** empty log message *** 2005-05-20 10:18 gb * [r1544] : TARGET-OS-CASE. 2005-05-20 09:26 gb * [r1543] : OS is now a nickname for the platform-specific package. Remove %COERCE-TO-UVECTOR. Whatever it was ... 2005-05-20 09:23 gb * [r1542] : Fix typos in code which initializes ppc64 class table. 2005-05-20 09:22 gb * [r1541] : Avoid "consing" single-floats/destructive operations on single-floats on PPC64. 2005-05-20 09:21 gb * [r1540] : setup CCL pathname translations for 64-bit fasls. 2005-05-20 09:20 gb * [r1539] : htonl, ntohs, etc. may not be defined in Tiger header files (and are no-ops on big-endian machines, anyway.) Use lisp macros to remind ourselves that there may be byte-order issues here, someday. 2005-05-20 09:20 gb * [r1538] : *** empty log message *** 2005-05-20 09:19 gb * [r1537] : Conditionalize MAKE-HEAP-IVECTOR. 2005-05-20 09:18 gb * [r1536] : 050520a. 2005-05-20 09:17 gb * [r1535] : CROSS-COMPILE-CCl. A lot of the stuff in this file is way too complicated (left over from 68k->PPC MCL bootstrapping a long time ago); there were and may still be some duplicate function definitions and other nonsense here. 2005-05-20 09:15 gb * [r1534] : Allow "the wrong target" to open .cdb files (as needs to be done during cross-compilation.) Need a better solution/scheme here, to prevent this from happening by accident. 2005-05-20 09:13 gb * [r1533] : Don't use the :PRINT-OBJECT DEFSTRUCT option (just define a couple of methods.) 2005-05-20 09:12 gb * [r1532] : Some cross-compilation infrastructure. Careful about references to 64-bit array types when cross-compiling. 2005-05-20 09:10 gb * [r1531] : Avoid using %MAKE-SFLOAT/destructive single-float operations on ppc64. 2005-05-20 09:09 gb * [r1530] : Fix #+ppc64-target typos in SIMPLE-1D-ARRAY-SUBSEQ. 2005-05-20 09:08 gb * [r1529] : Make the target argument &optional, not &key. Install the target backend while we're parsing, so that .cdb files will have the right signature even if they're cross-compiled. 2005-05-20 09:03 gb * [r1528] : Don't map arg/result types to representation types here (now happens in compiler frontend). 2005-05-20 08:59 gb * [r1527] : FSQRT: do the single-float case non-destructively on PPC64. 2005-05-20 08:58 gb * [r1526] : u32-shift-* -> natural-shift-*, and some other fixes... 2005-05-20 08:57 gb * [r1525] : u32-shift-* -> natural-shift-*. 2005-05-20 08:56 gb * [r1524] : listp: don't expand into references to the gensym on ppc64. 2005-05-20 08:55 gb * [r1523] : Ignore more kinds of fasl files. 2005-05-20 08:47 gb * [r1522] : Map syscall arg/result types to representation types here. 2005-05-19 08:20 gb * [r1521] : *** empty log message *** 2005-05-19 08:02 gb * [r1520] : Compiler backend stuff. 2005-05-19 08:02 gb * [r1519] : Record types passed by address on Linux (and any other sane ABI.) Integer encoding is in natural-sized words (32/64) when ABI is not sane. 2005-05-19 08:00 gb * [r1518] : A few things (at least a few of those that we know about) take/return :LONG or :UNSIGNED-LONG. 2005-05-19 07:59 gb * [r1517] : Pass args/results through FOREIGN-TYPE-TO-REPRESENTATION-TYPE (to handle things like :LONG, etc.) 2005-05-19 07:58 gb * [r1516] : Fix UNBOX-S32. 2005-05-19 07:58 gb * [r1515] : :signed/unsigned-HALFword. 2005-05-19 06:20 gb * [r1514] : *** empty log message *** 2005-05-19 05:35 gb * [r1513] : Remove FORMAT call. 2005-05-19 05:20 gb * [r1512] : *** empty log message *** 2005-05-19 05:17 gb * [r1511] : Fix parenthesization in PPC2-POWEROPEN-FOREIGN-ARGS. 2005-05-19 05:10 gb * [r1510] : Don't warn about tokens after #endif. (Don't know how to suppress warnings about tokens after #include .) Don't pre-include float.h in limits.h 2005-05-19 05:07 gb * [r1509] : ppc64 changes in poweropen-ffi-arglist stuff. 2005-05-19 05:01 gb * [r1508] : Pretend that :darwinppc64/:linuxppc64 targets follow PowerOpen ABI, which is kind of half-right, at least. 2005-05-19 05:00 gb * [r1507] : XCOMPILE-LAMBDA binds target-ftd. 2005-05-19 04:59 gb * [r1506] : Fix eep.address. 2005-05-19 04:56 gb * [r1505] : No more PPC64::ONE-DIGIT-BIGNUM-HEADER; fix a few things that thought that there was. 2005-05-19 04:55 gb * [r1504] : New PPC2-UNBOXED-INTEGER-ARG-TO-REG. 2005-05-19 02:20 gb * [r1503] : *** empty log message *** 2005-05-19 02:07 gb * [r1502] : CDB-OPEN creates a CDB structure (which may have a null FID); %CDB-GET does nothing if the FID is null. (This is intended to avoid repeated attempts to open a missing .cdb file, as when populating an interface directory for the first time.) 2005-05-19 02:04 gb * [r1501] : Handle 128-bit integers here. (Well, they aren't really handled anywhere, but we should at least recognize the data types.) 2005-05-19 01:20 gb * [r1500] : *** empty log message *** 2005-05-19 01:05 gb * [r1499] : Fix missing close-paren. 2005-05-19 01:01 gb * [r1498] : a 64-bit opcode or two. 2005-05-19 01:00 gb * [r1497] : Bignum header constants. 2005-05-19 01:00 gb * [r1496] : %init-misc for PPC64. 2005-05-19 00:59 gb * [r1495] : fix copy/past bug in box-u8. 2005-05-19 00:59 gb * [r1494] : 2, 3 digit bignum headers are interesting; 1 (32-bit) digit headers don't exist. 2005-05-19 00:57 gb * [r1493] : a bit more ppc64 conditionaliztion. (This is probably pretty close to complete - modulo 64-bit ivectors - but probably pretty buggy.) 2005-05-19 00:55 gb * [r1492] : extract-unsigned-byte-bits. for PPC64. 2005-05-18 07:20 gb * [r1491] : *** empty log message *** 2005-05-18 06:58 gb * [r1490] : Remove some more old/unreferenced stuff. 2005-05-18 06:58 gb * [r1489] : 050518a, since it's reasonable to assume that there will be more than 1-per-day. 2005-05-18 06:48 gb * [r1488] : no longer 1994 2005-05-18 06:44 gb * [r1487] : Lose some more old stuff. 2005-05-18 06:38 gb * [r1486] : Kinder, gentler .SPgetxlong (u32/s32 only.) 2005-05-18 06:38 gb * [r1485] : Remove some unused foreign struct definitions; make PTASKSTATE an istruct. 2005-05-18 06:37 gb * [r1484] : All (lisp-level) foreign structs defined per-ftd. 2005-05-18 06:37 gb * [r1483] : Don't try to describe exception frames ... 2005-05-18 06:37 gb * [r1482] : Export STREAM-DEVICE; don't export some old MCL stuff, %PUT-* functions. 2005-05-18 06:35 gb * [r1481] : Move foreign struct definitions elsewhere. 2005-05-18 06:34 gb * [r1480] : Don't eval-redef %put-* functions. 2005-05-18 06:34 gb * [r1479] : Don't mess with stack overflow limits on new stacks. 2005-05-18 06:33 gb * [r1478] : PTASKSTATE is now an istruct. 2005-05-18 06:33 gb * [r1477] : New memory access stuff. 2005-05-18 06:32 gb * [r1476] : Recognize LA to TSP. 2005-05-18 06:32 gb * [r1475] : Drop old (undocumented, problematic) %PUT-* memory update functions. Try to handle some 64-bit memory reference/assignments, make encoding a little more sane. 2005-05-18 06:30 gb * [r1474] : Enable vinsn debugging in XCOMPILE-LAMBDA. 2005-05-18 06:30 gb * [r1473] : Recognize :SIGNED-NATURAL alias, as well. 2005-05-17 16:20 gb * [r1472] : *** empty log message *** 2005-05-17 15:20 gb * [r1471] : old old file 2005-05-17 15:20 gb * [r1470] : *** empty log message *** 2005-05-17 14:43 gb * [r1469] : change target ftd arg handling 2005-05-17 14:20 gb * [r1468] : *** empty log message *** 2005-05-17 13:22 gb * [r1467] : Arrgh. 2005-05-17 13:20 gb * [r1466] : *** empty log message *** 2005-05-17 13:18 gb * [r1465] : new files. 2005-05-17 13:16 gb * [r1464] : new file; generates some warnings. 2005-05-17 13:12 gb * [r1463] : new files. 2005-05-17 13:11 gb * [r1462] : New file; use Fedora Core 4 test 3 headers. 2005-05-17 13:10 gb * [r1461] : make & install foreign type data for non-native backends. 2005-05-17 13:09 gb * [r1460] : Straighten out :void type identity. 2005-05-17 11:20 gb * [r1459] : *** empty log message *** 2005-05-17 11:05 gb * [r1458] : Pass the ftd around, mostly so that we can determine the (ftd-specific/ target-specific) size of a pointer. 2005-05-17 10:44 gb * [r1457] : fix linux typo. 2005-05-17 10:44 gb * [r1456] : um, 050517b. 2005-05-17 10:32 gb * [r1455] : Get GETF arg order right. Set long/pointer type sizes based on :bits-per-word. 2005-05-17 10:32 gb * [r1454] : Get GETF arg order right ... 2005-05-17 10:20 gb * [r1453] : *** empty log message *** 2005-05-17 09:48 gb * [r1452] : ftd-attributes is now a plist; use its :bits-per-word key to determine bit size of primitive :long, :unsigned-long types. 2005-05-17 09:45 gb * [r1451] : Pass the target ftd around a bit more, but that's probably hopeless. ftd-attributes is now a plist. 2005-05-17 09:44 gb * [r1450] : Conditionalize more; pass the target ftd around a bit more, but that's probably hopeless. Type translators are ftd-specific after all; basic types get installed per-ftd. ftd-attributes is now a plist. 2005-05-17 08:20 gb * [r1449] : *** empty log message *** 2005-05-17 07:32 gb * [r1448] : Bump version. 2005-05-17 07:32 gb * [r1447] : Bump fasl version. 2005-05-17 07:20 gb * [r1446] : *** empty log message *** 2005-05-17 07:04 gb * [r1445] : Lose ftype-info/ordinal stuff. (Causes bootstrapping problems, as does its removal.) Primitive types (translators) are global, not per-ftd. 2005-05-17 07:02 gb * [r1444] : WRITE-A-MACPTR: no ftype info. 2005-05-17 05:20 gb * [r1443] : *** empty log message *** 2005-05-17 04:29 gb * [r1442] : fix typos. 2005-05-17 04:20 gb * [r1441] : *** empty log message *** 2005-05-17 04:10 gb * [r1440] : Define a lot of PPC64 stuff. 2005-05-17 04:10 gb * [r1439] : Migrate builtin_function indices, fpscr bits to constants.s. Add lowtag defs. Fix double-float struct. 2005-05-17 04:09 gb * [r1438] : Migrate builtin_function indices, fpscr bits to constants.s. 2005-05-17 04:09 gb * [r1437] : Migrate builtin_function indices, fpscr bits here from arch-specific files. 2005-05-17 04:08 gb * [r1436] : Work around lack of 2005-05-17 04:07 gb * [r1435] : extract_lowtag for PPC64. 2005-05-17 04:07 gb * [r1434] : Add error_object_not_unsigned_byte_56. 2005-05-17 04:03 gb * [r1433] : Target packages, interface dirs. 2005-05-17 04:03 gb * [r1432] : Don't garble error messages when checking trailer. 2005-05-17 04:02 gb * [r1431] : Rename OS package. 2005-05-17 04:02 gb * [r1430] : Try to distinguish between referencing fullwords and "natural" words in memory. 2005-05-17 04:02 gb * [r1429] : Don't dereference pointers more than once (old %HGET stuff.) Try to distinguish between referencing fullwords and "natural" words in memory. 2005-05-17 04:00 gb * [r1428] : Recognize a few more operations. 2005-05-17 03:59 gb * [r1427] : %SET-SINGLE-FLOAT is ppc32-specific. Don't dereference pointers more than once (old %HGET stuff.) 2005-05-15 12:20 gb * [r1426] : *** empty log message *** 2005-05-15 11:29 gb * [r1425] : Bump; new binaries soon. 2005-05-15 11:29 gb * [r1424] : Lose old, obsolet LFUN-MAKER argument. May need to use XCOMPILE-CCL to bootstrap. 2005-05-15 07:20 gb * [r1423] : *** empty log message *** 2005-05-15 07:18 gb * [r1422] : Shift in the right direction when converting bit- to byte-offsets. 2005-05-15 07:18 gb * [r1421] : Default to *target-backend* (need to re-think some of this), use new (more specialized) fasl ops. 2005-05-15 07:16 gb * [r1420] : XLOAD-DNODE-ALIGN: always account for header. 2005-05-15 07:15 gb * [r1419] : Change an STW to an STR. 2005-05-15 07:15 gb * [r1418] : A few target-package changes, comments, indentation. 2005-05-15 07:14 gb * [r1417] : Add vector/array headers to target-uvector alist. 2005-05-15 06:39 gb * [r1416] : Recognize new fasl operators (specialized uvector/array types.) 2005-05-15 06:39 gb * [r1415] : More fasl operators, for specialized array types. Remove some (very) old code/concepts. 2005-05-15 06:38 gb * [r1414] : More fasl operators, for specialized array types. 2005-05-14 23:20 gb * [r1413] : *** empty log message *** 2005-05-14 23:18 gb * [r1412] : very old, irrelevant 2005-05-13 21:20 gb * [r1411] : *** empty log message *** 2005-05-13 20:50 gb * [r1410] : Fix typo in makeu64. 2005-05-13 11:20 gb * [r1409] : *** empty log message *** 2005-05-13 10:21 gb * [r1407] : U32/natural changes, start handling 64-bit register copies. 2005-05-13 10:20 gb * [r1406] : misc byte-count stuff. 2005-05-13 10:20 gb * [r1405] : *** empty log message *** 2005-05-13 10:19 gb * [r1404] : rename most u32 operations to "natural" ones. Handle some of them, as well as some u64/s64 boxing/unboxing operations. 2005-05-13 10:18 gb * [r1403] : rename most u32 operations to "natural" ones. 2005-05-13 10:18 gb * [r1402] : Implement byte-count stuff. 2005-05-13 10:17 gb * [r1401] : u32/natural changes. 2005-05-13 10:16 gb * [r1400] : Recognize "natural" constants and operands, as well as U32 ones. 2005-05-13 10:15 gb * [r1399] : Change most *u32* operator names to *natural*. 2005-05-13 10:14 gb * [r1398] : FIXUP-VINSN-TEMPLATE: account for the possibility that some (target-specific) vinsn templates may be missing. 2005-05-13 10:13 gb * [r1397] : Note when vinsn parameters/results have modes :s64/:u64, to prevent the corresponding registers from being used as temps. 2005-05-13 10:12 gb * [r1396] : Accept :NATURAL as an alias for an unsigned "natural width" (32/64-bit) integer. AVAILABLE-IMM-TEMP defaults mode to :NATURAL, as does the WITH-IMM-TARGET macro. 2005-05-13 10:10 gb * [r1395] : Architecture-specific ARRAY-DATA-SIZE-FUNCTION. 2005-05-13 10:09 gb * [r1394] : Skip the library version checks (at least in theory, the only classes/foreign methods we believe in are those described by the interface files.) Make %MLIST-CONTAINING force canonicaliazation of class pointers when updating message signature tables. 2005-05-13 10:07 gb * [r1393] : UPDATE-TYPE-SIGNATURES-FOR-METHOD optionally canonicalizes the class, so that references to user-defined classes in method descriptor tables get revived when a saved image is restarted. 2005-05-13 10:05 gb * [r1392] : %FIXNUM-REF-U32 -> %FIXNUM-REF-NATURAL, and other 32/64-bit changes. 2005-05-13 10:04 gb * [r1391] : %FIXNUM-REF-U32 -> %FIXNUM-REF-NATURAL. 2005-05-13 10:04 gb * [r1390] : fix some 32/64-bit boxing stuff. 2005-05-13 10:02 gb * [r1389] : 050513. 2005-05-12 05:20 gb * [r1388] : *** empty log message *** 2005-05-12 05:10 gb * [r1387] : save/load NVRs without using stmw/lmw (since there aren't 64-bit equivalents.) use new __ifdef mechanism (old mechanism interfered with source line number scheme.) 2005-05-12 05:10 gb * [r1386] : save/load NVRs without using stmw/lmw (since there aren't 64-bit equivalents.) use new __ifdef mechanism (old mechanism interfered with source line number scheme.) A little bit of __ifdef([PPC64]) stuff works. 2005-05-12 05:06 gb * [r1385] : save/load NVRs without using stmw/lmw (since there aren't 64-bit equivalents.) 2005-05-12 05:05 gb * [r1384] : New __ifdef mechanism. 2005-05-12 05:05 gb * [r1383] : Use new __ifdef mechanism. 2005-05-12 05:04 gb * [r1382] : min_array_subtag, min_vector_subtag. 2005-05-12 05:04 gb * [r1381] : Enumerate all NVR names. Apple's assembler sucks. 2005-05-12 05:03 gb * [r1380] : Provide a more ... descriptive description for lisp exceptions. 2005-05-12 05:02 gb * [r1379] : Declare a few more private classes, for document load/save. Need a better long-term solution, obviously. 2005-05-12 05:01 gb * [r1378] : Don't call CCL::DBG on duplicate Hemlock buffer names. Disable "smart insertion and deletion"; apparently, this has something to do with frameworks. Turn on backup files, which would be named "foo~.lisp". Not sure how easy this is to change. 2005-05-12 04:59 gb * [r1377] : Failures in OBJC-CLASS-OF are often spurious, but we really shouldn't be asking this question in a lot of cases. When #+debug, try to print the object's class name, so that we can special-case it. 2005-05-12 04:20 gb * [r1376] : *** empty log message *** 2005-05-12 04:14 gb * [r1375] : Remove a DBG trap; don't TIME the call to import method type signatures. 2005-05-12 04:04 gb * [r1374] : Be more careful to distinguish between (ARRAY *) and (ARRAY T) cases. Treat DOUBLE-FLOAT objects as if they consisted of 2 32-bit elements. 2005-05-12 04:04 gb * [r1373] : Use TARGET-ARCH-CASE; try to simplify simple MAKE-ARRAY calls 2005-05-12 04:03 gb * [r1372] : Be more careful to distinguish between (ARRAY *) and (ARRAY T) cases. 2005-05-12 04:02 gb * [r1371] : Don't use WITH-STACK-SHORT-FLOATS/destuctive sfloat operations on PPC64. 2005-05-12 04:01 gb * [r1370] : Changes for immediate single/short floats on ppc64. 2005-05-12 04:00 gb * [r1369] : Remove a comment. 2005-05-12 04:00 gb * [r1368] : Add #\Newline at EOF. 2005-05-12 03:59 gb * [r1367] : %PASCAL-FUNCTIONS% starts out a little bigger. 2005-05-12 03:58 gb * [r1366] : DELETE-PASCAL-FUNCTION, be a little more careful about assumptions that PFEs are named/named globally. 2005-05-12 03:57 gb * [r1365] : %MAKE-SIMPLE-ARRAY. 2005-05-12 03:57 gb * [r1364] : Make shorter trampolines (use absolute branches.) 2005-05-12 03:56 gb * [r1363] : 050511. 2005-05-12 03:56 gb * [r1362] : Use TARGET-ARCH-CASE in most macros. 2005-05-12 03:55 gb * [r1361] : Some non-destructive SINGLE/SHORT-FLOAT stuff for PPC64. 2005-05-12 03:54 gb * [r1360] : Comment indentation. 2005-05-12 03:54 gb * [r1359] : Use architecture-specific NUMERIC-TYPE-NAME-TO-TYPECODE-FUNCTION. 2005-05-12 03:54 gb * [r1358] : NUMERIC-TYPE-NAME-TO-TYPECODE-FUNCTION. 2005-05-11 03:20 gb * [r1357] : *** empty log message *** 2005-05-11 02:47 gb * [r1356] : Don't revert to standard float printer if MARKER argument provided, in FORMAT-EXP-AUX. 2005-05-10 19:20 gb * [r1355] : *** empty log message *** 2005-05-10 19:16 gb * [r1354] : Bump. 2005-05-10 19:16 gb * [r1353] : Use ! to indicate destructive operation in LOG-E. 2005-05-10 19:16 gb * [r1352] : Bring back (and comment-out) the old %ILSR-C vinsn, just in case (I had some problems reading float contants around the time that the previous change was made; I was messing with how shift operations were compiled at the time, and the problem may have been elsewhere. Just in case.) 2005-05-10 18:45 gb * [r1351] : conditionalize count_leading_zeros, conditional loads/stores. 2005-05-10 18:44 gb * [r1350] : New scheme for assembler kernel globals. 2005-05-10 18:43 gb * [r1349] : Move SUBTAG_CHARACTER. 2005-05-10 18:42 gb * [r1348] : Move SUBTAG_CHARACTER. New scheme for assembler kernel globals. 2005-05-10 18:39 gb * [r1347] : CLEAR_ALLOC_TAG: parenthesize operands. 2005-05-10 18:39 gb * [r1346] : Typo. 2005-05-10 18:39 gb * [r1345] : Colons. 2005-05-10 18:38 gb * [r1344] : #+PPC64-target: special-case (SYMBOL-NAME NIL). 2005-05-10 18:37 gb * [r1343] : LOGIOR X 0 is X, LOGIOR 0 Y is Y. Sheesh. 2005-05-10 18:37 gb * [r1342] : There were 2 (different) definitions of %ILSR-C; the second (which took effect) generated slightly worse code (for an operation that's rarely used.) 2005-05-10 18:35 gb * [r1341] : Change source file indicator. 2005-05-10 18:34 gb * [r1340] : Parameterize LOGBITP a bit. 2005-05-10 18:34 gb * [r1339] : Lots of changes, mostly for the better ... 2005-05-10 18:33 gb * [r1338] : Target package-name, t-offset. Lots of other changes, including reassigning SUBTAG-CHARACTER to make SUBTAG-SINGLE-FLOAT easier to identify. 2005-05-10 18:32 gb * [r1337] : Target package-name, t-offset. 2005-05-10 18:31 gb * [r1336] : Support T as a (constant) operand. 2005-05-10 18:31 gb * [r1335] : Conditionalize for ppc64. 2005-05-10 18:30 gb * [r1334] : Conditionalize for ppc64, partly (but see MAKE-ARRAY). 2005-05-10 18:29 gb * [r1333] : XCOMPILE-LAMBDA, to test cross-compiler. 2005-05-10 18:29 gb * [r1332] : Package name, t-offset in target-arch struct. 2005-05-10 07:20 gb * [r1331] : *** empty log message *** 2005-05-10 06:39 gb * [r1330] : Parameterize, try to force a few more things to inline. 2005-05-10 06:39 gb * [r1329] : Link with -lmx. 2005-05-10 06:39 gb * [r1328] : Add a non-destructive %SINGLE-FLOAT-LOG for PPC64. (Assumes -lmx on Darwin64). 2005-05-10 06:37 gb * [r1327] : Use target::most-positive/negative-fixnum. Remove a few MCL-isms. 2005-05-10 06:36 gb * [r1326] : some more #+ppc64-target conditionalization; still more to be done (especially bignum stuff.) 2005-05-10 06:33 gb * [r1325] : remove empty files. (Remember to remove old fasls before (XLOAD-LEVEL-0). 2005-05-10 06:32 gb * [r1324] : NODE-SIZE. 2005-05-10 06:31 gb * [r1323] : interlocked loads/stores, some more shifts, interrupt-checking fixes. 2005-05-09 14:20 gb * [r1322] : *** empty log message *** 2005-05-09 13:42 gb * [r1321] : conditionalize for ppc64. 2005-05-09 13:41 gb * [r1320] : EXTRACT-LOWBYTE for ppc64. 2005-05-09 12:20 gb * [r1319] : *** empty log message *** 2005-05-09 11:43 gb * [r1318] : Conditionalize some for ppc64; warn at compile time about missing ppc64 functions. 2005-05-09 11:30 gb * [r1317] : Use CL arithmetic to create 32-bit integers ... 2005-05-09 11:30 gb * [r1316] : Remove some unused functions. 2005-05-09 11:29 gb * [r1315] : TARGET package. 2005-05-09 11:28 gb * [r1314] : lowtag constants. 2005-05-09 11:28 gb * [r1313] : TRLGEI, EXTRACT-LOWTAG. 2005-05-09 11:27 gb * [r1312] : 64-bit vectors, *CLASS-TABLE* changes for #+ppc64-target. 2005-05-09 05:20 gb * [r1311] : *** empty log message *** 2005-05-09 05:01 gb * [r1310] : Start to conditionalize for #+ppc64-target; -mostly- involves changing PPC32 package prefix to TARGET (but some -real- differences.) 2 down, about 100 to go. 2005-05-09 05:01 gb * [r1309] : Start to conditionalize for #+ppc64-target; -mostly- involves changing PPC32 package prefix to TARGET. 2005-05-09 04:59 gb * [r1308] : Define lowtag constants, min-{vector,array}-subtag values. 2005-05-09 03:20 gb * [r1307] : *** empty log message *** 2005-05-09 03:14 gb * [r1306] : Fix some tag definitions, add missing ratio/complex. 2005-05-09 03:14 gb * [r1305] : Fix some tag definitions, add missing ratio/complex, define tag_of_nil. 2005-05-09 03:12 gb * [r1304] : Define tag_of_nil, for lisp_globals parameterization. 2005-05-09 03:11 gb * [r1303] : Parameterize lisp_globals struct origin stuff. 2005-05-08 13:20 gb * [r1302] : *** empty log message *** 2005-05-08 12:25 gb * [r1301] : Bump; time for new binaries. 2005-05-08 12:25 gb * [r1300] : Remove old code-vector/svar loading cases. 2005-05-08 12:20 gb * [r1299] : *** empty log message *** 2005-05-08 12:17 gb * [r1298] : Remove old code-vector/svar dumping cases. 2005-05-08 11:59 gb * [r1297] : AREA- enumeration constants are PPC-specific, not platform-specific. (AREA. accessors -are- platform-specific). 2005-05-08 11:57 gb * [r1296] : Use $fasl-svar, $fasl-codevector. 2005-05-08 11:20 gb * [r1295] : *** empty log message *** 2005-05-08 11:11 gb * [r1294] : Define (and use) $FASL-CODE-VECTOR, $FASL-SVAR. 2005-05-08 10:30 gb * [r1293] : Remove obsolete function. 2005-05-08 10:20 gb * [r1292] : FASL-DUMP-FUNCTION. 2005-05-08 10:20 gb * [r1291] : *** empty log message *** 2005-05-08 10:19 gb * [r1290] : Report file position of bad fasl opcode. 2005-05-08 09:56 gb * [r1289] : Fix typo. 2005-05-08 09:51 gb * [r1288] : Drop old operators; add $FASL-FUNCTION. 2005-05-08 09:46 gb * [r1287] : Drop old operators; add $FASL-FUNCTION. 2005-05-08 09:30 gb * [r1286] : Use new $fasl-vetab-alloc/count, not $fasl-etab-alloc/long. 2005-05-08 09:20 gb * [r1285] : *** empty log message *** 2005-05-08 09:07 gb * [r1284] : Increment fasl version. 2005-05-08 08:52 gb * [r1283] : FASL-DUMP-DISPATCH: don't forget to dump character constants, and use ETYPECASE to keep from quietly neglecting to dump other types in the future. (All of this is getting tricky to bootstrap; there will be new binaries soon.) 2005-05-08 07:20 gb * [r1282] : *** empty log message *** 2005-05-08 07:17 gb * [r1281] : Know (a bit) less about platform-dependent typecodes when dumping; still more work to be done here. 2005-05-08 07:16 gb * [r1280] : Define (and implement) $FASL-SFLOAT. 2005-05-08 07:11 gb * [r1279] : DOUBLE-FLOAT-BITS, DOUBLE-FLOAT-FROM-BITS. 2005-05-08 07:09 gb * [r1278] : DEFINE-CONSTANT: report new, old value in CERROR message. 2005-05-08 07:08 gb * [r1277] : $FASL-SFLOAT. 2005-05-08 07:07 gb * [r1276] : BIGNUM, DOUBLE-FLOAT have 32-bit elements on PPC64 (to ease bootstrapping.) 2005-05-08 06:20 gb * [r1275] : *** empty log message *** 2005-05-08 05:52 gb * [r1274] : (host) support to get/set single-float bits. 2005-05-08 05:51 gb * [r1273] : Use new (so far ...) operators. 2005-05-08 05:51 gb * [r1272] : Add support for new (so far) operators. 2005-05-08 05:50 gb * [r1271] : Rename $FASL-FLOAT to $FASL-DFLOAT. 2005-05-08 05:50 gb * [r1270] : Ignore (ignored) extended-string flag in obsolete FASL-READSTR and equally obsolete callers. 2005-05-08 04:20 gb * [r1269] : *** empty log message *** 2005-05-08 04:01 gb * [r1268] : Remove (old, unused) extended-string stuff. 2005-05-08 03:46 gb * [r1267] : FASL-OUT-COUNT, to encode arbitrary-sized UNSIGNED-BYTE values. 2005-05-08 03:45 gb * [r1266] : implement some new fasl operators. 2005-05-08 03:45 gb * [r1265] : Define some new fasl operators. 2005-05-08 00:20 gb * [r1264] : *** empty log message *** 2005-05-07 23:54 gb * [r1263] : use $DEFSTUCT-STRUCT, to simplify cross-compilation. 2005-05-07 23:36 gb * [r1262] : define $DEFSTUCT-STRUCT, to simplify cross-compilation. 2005-05-07 23:34 gb * [r1261] : Provide a couple of doc strings. 2005-05-07 23:34 gb * [r1260] : Function (macro, special-operator) documentation associated with function object. 2005-05-07 23:28 gb * [r1259] : Don't use %VECT-SUBTYPE. 2005-05-07 23:28 gb * [r1258] : Associate documentation with the definition. 2005-05-07 23:27 gb * [r1257] : Don't define %VECT-SUBTYPE, to remove temptation to use it. 2005-05-07 23:27 gb * [r1256] : TRAP-UNLESS-CHARACTER, TRAP-UNLESS-CONS. 2005-05-07 23:26 gb * [r1255] : Use :struct to name structure type. 2005-05-07 23:26 gb * [r1254] : Hide a few more ppc32 dependencies. Still issues when native typecodes are involved (%gvector, %alloc-misc, etc.) 2005-05-07 23:24 gb * [r1253] : Remove doc string from IF lap macro. 2005-05-07 05:20 gb * [r1252] : *** empty log message *** 2005-05-07 04:59 gb * [r1251] : Typed-uvref/uvset operators accept keywords. 2005-05-07 04:59 gb * [r1250] : Use the new (keyword-based) typed-uvref/uvset operators. 2005-05-07 04:21 gb * [r1249] : Remove some ppc32 package constants. 2005-05-07 04:20 gb * [r1248] : *** empty log message *** 2005-05-07 04:15 gb * [r1247] : ARRAY-TYPE-NAME-FROM-CTYPE-FUNCTION in arch struct, and PPC32 implementation. 2005-05-07 02:20 gb * [r1246] : *** empty log message *** 2005-05-07 01:54 gb * [r1245] : Don't use %TYPED-MISC-REF, %TYPED-MISC-SET anymore. 2005-05-07 01:54 gb * [r1244] : SINGLE-FLOAT-BITS. 2005-05-07 01:54 gb * [r1243] : Don't define %TYPED-MISC-REF, %TYPED-MISC-SET anymore. 2005-05-07 00:20 gb * [r1242] : *** empty log message *** 2005-05-07 00:17 gb * [r1241] : Fix typo. 2005-05-07 00:06 gb * [r1240] : Change typecode name in TYPE-KEYWORD-CODE call, do the call at readtime, not runtime. 2005-05-07 00:05 gb * [r1239] : TYPE-KEYWORD-CODE uses vector types from target. 2005-05-07 00:05 gb * [r1238] : Define vector type names. Flesh out uvector type codes. 2005-05-07 00:04 gb * [r1237] : Add vector type names (gvector, n-bit-ivector.) 2005-05-03 23:20 gb * [r1236] : *** empty log message *** 2005-05-03 22:31 gb * [r1235] : No need to zero_last_page() anymore. 2005-05-03 22:31 gb * [r1234] : use HeapHighWaterMark to mark highest mapped address known to be zero-filled. 2005-05-03 22:30 gb * [r1233] : No memory protection anymore, so no need to zero_last_page(). 2005-05-03 22:30 gb * [r1232] : Add a missing #\Newline. 2005-05-03 22:29 gb * [r1231] : New handling of heap resizing and segment allocation/zeroing. 2005-05-03 22:28 gb * [r1230] : Turn the EGC on again; track HeapHighWaterMark. 2005-05-03 22:28 gb * [r1229] : Use allocate_tcr() to ... allocate TCR pointers, to ensure that a Mach port can be assigned the TCR's address as a "name". 2005-05-03 22:27 gb * [r1228] : Support doc strings in level-0. 2005-05-03 22:26 gb * [r1227] : Return another value from %STAT-VALUES. 2005-05-03 22:25 gb * [r1226] : MEMBER/ASSOC support, MACRO-FUNCTION moved to level-0 (to handle early doc strings.) 2005-05-03 22:25 gb * [r1225] : Handle cold-load documentation. 2005-05-03 22:24 gb * [r1224] : Some MEMBER/ASSOC support here. 2005-05-03 22:24 gb * [r1223] : Change the wording in ROOM: it's not 1989 anymore. 2005-05-03 22:23 gb * [r1222] : Remove (this version of) SET-DOCUMENTATION, which might or might not have been overridden by another version ... Define a SET-FUNCTION-INFO, which handles the case where a function's information is just a doc string. Move the definition of MACRO-FUNCTION here. 2005-05-02 00:20 gb * [r1221] : *** empty log message *** 2005-05-01 23:53 gb * [r1220] : Hamilton's fix in RENDER-FOR-SELECTION. 2005-04-29 10:20 gb * [r1219] : *** empty log message *** 2005-04-29 09:35 gb * [r1218] : Increment FASL version. 2005-04-29 09:20 gb * [r1217] : *** empty log message *** 2005-04-29 09:07 gb * [r1216] : Writes that may introduce intergenerational references should all be handled here; ensuring atomicity and handling locatives requires some support in pc_luser_xp(). 2005-04-29 09:05 gb * [r1215] : New subprims replace some old ones. 2005-04-29 09:05 gb * [r1214] : No more hardware EGC write barrier. EGC off by default (for bootstrapping new write barrier.) 2005-04-29 09:03 gb * [r1213] : No more hardware EGC write barrier. 2005-04-29 09:02 gb * [r1212] : A little more 32/64 macrology; some support for bitvector operations. 2005-04-29 09:01 gb * [r1211] : No more pageentry type. 2005-04-29 09:01 gb * [r1210] : New globals (REFBITS, OLDSPACE_DNODE_COUNT). 2005-04-29 09:00 gb * [r1209] : No more protectin-based write barrier. Some new pc-luser support to make software write barrier appear to be atomic (needs testing.) 2005-04-29 08:59 gb * [r1208] : No more pagemap. 2005-04-29 08:59 gb * [r1207] : New (software) EGC write barrier; drop support for the old one ... 2005-04-29 08:58 gb * [r1206] : Move lisp-global definitions elsewhere; define some bitvector-related values. 2005-04-29 08:57 gb * [r1205] : Define t_value. 2005-04-29 08:56 gb * [r1204] : Move lisp-global definitions here. 2005-04-29 08:54 gb * [r1203] : Some word-size issues; still very 32-bit-centric. 2005-04-29 08:53 gb * [r1202] : Add atomic_ior. 2005-04-29 08:51 gb * [r1201] : Bump. 2005-04-29 08:51 gb * [r1200] : %STORE-NODE-CONDITIONAL goes out-of-line, for EGC write barrier. %STORE-IMMEDIATE-CONDITIONAL skips that, when fixnums/immediates are known to be involved. 2005-04-29 08:50 gb * [r1199] : %SET-HASH-TABLE-VECTOR-KEY goes out of line, to handle EGC write barrier. 2005-04-29 08:49 gb * [r1198] : Use %STORE-IMMEDIATE-CONDITIONAL when fixnums are involved. 2005-04-29 08:48 gb * [r1197] : Remove misc-set-node and vcell-set. 2005-04-29 08:48 gb * [r1196] : lwi -> lri. Do setq of closed-over vcells via .SPgvset, for EGC. 2005-04-29 08:46 gb * [r1195] : Replace some obsolete globals with refbits and oldspace-dnode-count. 2005-04-29 08:45 gb * [r1194] : Replace something obsolete with .SPset-hash-key. 2005-04-26 17:20 gb * [r1193] : *** empty log message *** 2005-04-26 17:02 gb * [r1192] : Make Darwin assembler happy. (Well, happier than it ordinarily is.) 2005-04-26 17:01 gb * [r1191] : Declare NSConcreteData. 2005-04-26 17:00 gb * [r1190] : SPstore_node_conditional. 2005-04-26 16:58 gb * [r1189] : Remove (very) obsolete %BIND-SPECIAL. 2005-04-26 16:57 gb * [r1188] : Do rplaca/rplacd (and variants), svset (and variants) out-of-line, for new write-barrier scheme. 2005-04-26 16:55 gb * [r1187] : A few indentation changes. 2005-04-26 16:55 gb * [r1186] : Add .SPstore-node-conditional. 2005-04-26 11:20 gb * [r1185] : *** empty log message *** 2005-04-26 10:57 gb * [r1184] : Updated. 2005-04-26 10:56 gb * [r1183] : Disable EGC, initially. 2005-04-26 10:54 gb * [r1182] : Start new write-barrier scheme. 2005-04-26 07:20 gb * [r1181] : *** empty log message *** 2005-04-26 06:55 gb * [r1180] : Partially implement SEND in terms of interface information. 2005-04-26 06:54 gb * [r1179] : Avoid message arg type ambiguity. 2005-04-26 06:53 gb * [r1178] : Declare SEND arg type. 2005-04-26 06:52 gb * [r1177] : Handle foreign slots differently, since offset/width info is available from interfaces. 2005-04-26 06:52 gb * [r1176] : Derive class info from interfaces, not via introspection. 2005-04-26 06:51 gb * [r1175] : Don't references private slots or use SEND to send undeclared messages. 2005-04-26 06:50 gb * [r1174] : Bump. 2005-04-26 06:50 gb * [r1173] : Start thinking about making some of this target-sensitive. Pre-define : as an integer type. 2005-04-26 06:49 gb * [r1172] : Handle protocol methods, protocols in ObjC class decls. 2005-04-26 06:47 gb * [r1171] : Add MAP-SPLAY-TREE-KEYS-AND-VALUES. 2005-04-24 06:20 gb * [r1170] : *** empty log message *** 2005-04-24 06:01 gb * [r1169] : Ignore fill-pointers when looking at array dimensions in RESULT-BIT-ARRAY. 2005-04-20 13:20 gb * [r1168] : *** empty log message *** 2005-04-20 12:25 gb * [r1167] : Write trailer on CDB files, check for it. 2005-04-20 12:25 gb * [r1166] : Bump, once again. 2005-04-20 12:24 gb * [r1165] : New file. 2005-04-20 09:20 gb * [r1164] : *** empty log message *** 2005-04-20 09:08 gb * [r1163] : Prepend 2K of NULs to .CDB files, for backward compatibility (sort of). Still more bootstrapping ... 2005-04-20 07:20 gb * [r1162] : *** empty log message *** 2005-04-20 06:53 gb * [r1161] : Changes to how some things are encoded in .cdb files; field offsets and widths are now provided explicitly. This will require new .cdb files and is a little hard to bootstrap ... 2005-04-20 06:50 gb * [r1160] : Bryan's change to allow potential numbers with package prefixes to be treated as symbols. 2005-04-20 06:49 gb * [r1159] : Updated, a bit. 2005-04-20 06:48 gb * [r1158] : Assume new-style location for gnome.h 2005-04-20 06:47 gb * [r1157] : Ignore backup .cdb files. 2005-04-20 06:46 gb * [r1156] : New file. 2005-04-18 07:20 gb * [r1155] : *** empty log message *** 2005-04-18 06:48 gb * [r1154] : Fix typo: "dictionaryWithObjectsAndKeys:". 2005-04-18 06:47 gb * [r1153] : Support for ObjC class info, method info in interface files. 2005-04-18 06:46 gb * [r1152] : Export CURRENT-DIRECTORY; drop COMPILE-LOAD. 2005-04-18 06:44 gb * [r1151] : Bump, yet again. 2005-04-18 06:44 gb * [r1150] : Implement CURRENT-DIRECTORY and (SETF CURRENT-DIRECTORY). 2005-04-18 06:44 gb * [r1149] : Some toplevel commands for getting/setting the current directory. 2005-04-18 06:43 gb * [r1148] : %DEFTYPE clears the type cache. 2005-04-18 06:43 gb * [r1147] : Remove COMPILE-LOAD, which doesn't offer much functionality beyond that offered by (compile ... :load t), does it ? 2005-04-18 06:41 gb * [r1146] : Say "FASL", not "PFSL" in error messages. 2005-04-18 06:41 gb * [r1145] : Use more recent headers (current Panther headers.) 2005-04-18 06:40 gb * [r1144] : Process agl.h as well. 2005-04-18 06:40 gb * [r1143] : New file. 2005-04-18 06:38 gb * [r1142] : Ignore *.cdb-BAK as well. 2005-04-18 06:37 gb * [r1141] : New files. 2005-04-18 06:26 gb * [r1140] : Handle the 2-argument case of case-sensitive character comparison functions. 2005-04-18 02:20 gb * [r1139] : *** empty log message *** 2005-04-18 01:29 alex * [r1138] : edited %objc-domain-slots-vector to allocate slot vectors on demand for object instantiated outside of lisp (AppKit). 2005-04-18 01:27 alex * [r1137] : added a trace flag to the callback table and extended %pascal-functions% to set a special variable *callback-trace-p* during execution. 2005-04-18 01:26 alex * [r1136] : extended the trace facility to support tracing of callbacks, a :backtrace argument and hook variables for the IDE 2005-04-18 01:25 alex * [r1135] : Extended DEFINE-CALLBACK and DEFCALLBACK-BODY to support tracing. 2005-04-17 15:20 gb * [r1134] : *** empty log message *** 2005-04-17 14:56 beer * [r1133] : evised %PARSE-MESSAGE and CHECK-MESSAGE-ARG-COUNT to fix a bug with interpreting some variable-arity message sends 2005-04-15 13:20 gb * [r1132] : *** empty log message *** 2005-04-15 12:23 gb * [r1131] : Bump. 2005-04-15 12:20 gb * [r1130] : *** empty log message *** 2005-04-15 12:13 gb * [r1129] : Fix EVAL-REDEF of %ALLOC-MISC. 2005-04-15 12:00 gb * [r1128] : Define charcodeshift, ncharchodebits consistently with what ppc64-arch says. 2005-04-15 12:00 gb * [r1127] : Apparently, it isn't necessary to use the well-known -weak_references_mismatch argument when that has nothing to do with anything. ld64 just keeps getting better and better! 2005-04-15 11:59 gb * [r1126] : TRAP-UNLESS-MACPTR. Do a few more 64-bit comparisons. 2005-04-15 11:57 gb * [r1125] : TRAP-UNLESS-MACPTR. 2005-04-15 11:57 gb * [r1124] : Need a 1-word code-vector prefix here. 2005-04-15 11:56 gb * [r1123] : No code-vector prefix here. 2005-04-15 11:56 gb * [r1122] : Maybe emit a "prefix" on the code vector. More PPC64 conditionalization. 2005-04-15 11:53 gb * [r1121] : Maybe emit a "prefix" on the code vector. 2005-04-15 11:52 gb * [r1120] : Skip over any code-vector prefix. Start to conditionalize to make PPC64 disassembly better. 2005-04-15 11:51 gb * [r1119] : Define, use, NX1-TARGET-FIXNUMP. 2005-04-15 11:50 gb * [r1118] : CODE-VECTOR-PREFIX, for PPC64 alignment. 2005-04-15 04:20 gb * [r1117] : *** empty log message *** 2005-04-15 03:27 gb * [r1116] : charcode-shift is 8; ncharcodebits is 8 (for now). 2005-04-15 03:26 gb * [r1115] : Actually bother to define LOAD-CHARACTER-CONSTANT. 2005-04-15 03:26 gb * [r1114] : :u64const, :s64const. 2005-04-15 03:20 gb * [r1113] : LWI -> LRI, PPC2-LWI -> PPC2-LRI. 2005-04-15 03:20 gb * [r1112] : LWI -> LRI. 2005-04-15 03:20 gb * [r1111] : *** empty log message *** 2005-04-15 03:19 gb * [r1110] : Use TARGET-ARCH-CASE, no need for PPC-LAP-TARGET-ARCH-CASE. 2005-04-15 03:18 gb * [r1109] : TARGET-ARCH-CASE here. 2005-04-14 08:20 gb * [r1108] : *** empty log message *** 2005-04-14 07:39 gb * [r1107] : Finish bootstrapping backend changes. 2005-04-14 07:20 gb * [r1106] : *** empty log message *** 2005-04-14 07:13 gb * [r1105] : Start to remove obsolete target slots. 2005-04-14 06:51 gb * [r1104] : Fix typos, exposed by Bryan's fix to #+/#- readers. 2005-04-14 06:21 gb * [r1103] : Bump version. 2005-04-14 06:20 gb * [r1102] : #\B in typestrings is :. C arrays are always passed by reference. 2005-04-14 06:20 gb * [r1101] : Factor out some typecode-specific stuff. 2005-04-14 06:19 gb * [r1100] : Backend/arch parameterization. 2005-04-14 06:18 gb * [r1099] : Use target-specific parameters from structure, not individually. 2d-aref/aset acode changes. 2005-04-14 06:18 gb * [r1098] : Use target-specific parameters from structure, not individually. 2005-04-14 06:17 gb * [r1097] : Require "ARCH" at load-time. 2005-04-14 06:16 gb * [r1096] : Try to keep target tags out of acode (for 2d array access only, so far.) 2005-04-14 06:16 gb * [r1095] : Add a "target-arch" slot to backend structure (this should allow several other slots to be removed.) 2005-04-14 06:15 gb * [r1094] : Start moving parameters from "backend" structure to new "target-arch" structure. 2005-04-13 18:23 bryan * [r1093] : "file not found" during compile-file should be a file-error. 2005-04-13 08:13 gb * [r1092] : *** empty log message *** 2005-04-13 04:57 bryan * [r1091] : per 2.4.8.17, "#+(or) #+(or) nil nil t" should probably read as T 2005-04-01 10:23 gb * [r1090] : No they weren't: parse PreferencePanes.h 2005-04-01 07:15 gb * [r1089] : Minor target package changes. 2005-03-30 05:54 gb * [r1088] : Start to add target-uvector-subtags to backend. 2005-03-29 14:06 gb * [r1087] : First cut at a Makefile. 2005-03-29 14:05 gb * [r1086] : Linker script for linux64; force kernel into low memory. 2005-03-29 14:05 gb * [r1085] : May be grief including # on linux64. Note that fact. 2005-03-29 14:04 gb * [r1084] : extract_lowbyte uses clrlri. 2005-03-29 14:04 gb * [r1083] : Don't save/restore PT_MQ; this had something to do with saving/restoring Altivec registers, a long time ago. 2005-03-29 14:03 gb * [r1082] : Don't import __trampoline_setup on linux64. 2005-03-29 14:02 gb * [r1081] : No eabi_c_frame on linux64. 2005-03-28 11:34 gb * [r1080] : The actual cocoa interfaces (from darwin-interfaces-040505.tar.gz) were built from this, not from what's been in CVS. 2005-03-28 11:33 gb * [r1079] : TRAP-UNLESS-{SINGLE,DOUBLE}-FLOAT, refer to the least significant bit as "lowbit", not as "bit31". A few other things implemented; a lot to go. 2005-03-28 11:31 gb * [r1078] : TRAP-UNLESS-{SINGLE,DOUBLE}-FLOAT, refer to the least significant bit as "lowbit", not as "bit31". 2005-03-28 11:28 gb * [r1077] : Some fixes & cleanups; TCR-BIAS. 2005-03-28 11:26 gb * [r1076] : More compilation-target parameters in the backend. 2005-03-28 11:25 gb * [r1075] : Start to remove some ppc32-isms. 2005-03-28 11:24 gb * [r1074] : MOST (not all) things run-time conditionalized for ppc32/ppc64. 2005-03-28 11:24 gb * [r1073] : Remove some PPC32-isms. 2005-03-28 11:23 gb * [r1072] : Put more target stuff in the backend structure. 2005-03-28 11:22 gb * [r1071] : GF protos, trampolines: LAP moved here. 2005-03-28 11:21 gb * [r1070] : Float sign, zerop functions in lisp. 2005-03-28 11:21 gb * [r1069] : A little bit of conditionalization (for a little bit of code.) 2005-03-28 11:20 gb * [r1068] : A little bit of conditionalization (needs more work.) 2005-03-28 11:19 gb * [r1067] : %DOUBLE-FLOAT-PLUSP, %SHORT-FLOAT-PLUSP in lisp. 2005-03-28 11:18 gb * [r1066] : %DOUBLE-FLOAT-ZEROP, %SHORT-FLOAT-ZEROP in lisp. 2005-03-28 11:18 gb * [r1065] : %DOUBLE-FLOAT-SIGN, %SHORT-FLOAT-SIGN in lisp. 2005-03-28 11:17 gb * [r1064] : Bump; again, a little hard to bootstrap. 2005-03-28 11:16 gb * [r1063] : Remove some commented-out code. 2005-03-28 11:16 gb * [r1062] : Define PPC64-VINSNS. 2005-03-28 11:15 gb * [r1061] : Include PPC64-ARCH in *PPC-COMPILER-MODULES*. 2005-03-28 11:14 gb * [r1060] : Define TCR_BIAS. 2005-03-28 11:13 gb * [r1059] : Don't use NATURAL_POINTER_FIELD. Define TCR_BIAS. 2005-03-28 11:13 gb * [r1058] : Define TCR_BIAS; ensure that TCR has single_float_convert field. 2005-03-28 11:12 gb * [r1057] : No TCR_BIAS on Darwin32; rcontext is now r13. 2005-03-28 11:11 gb * [r1056] : Don't (try to) support 32-bit foreign pointers in 64-bit code. 2005-03-28 11:10 gb * [r1055] : Don't define NATURAL_POINTER_FIELD. 2005-03-28 11:10 gb * [r1054] : Don't use NATURAL_POINTER_FIELD. 2005-03-28 11:09 gb * [r1053] : Don't use NATURAL_POINTER_FIELD. Define TCR_BIAS in constants{32,64}.h 2005-03-28 11:08 gb * [r1052] : Move LAP code to level-0. 2005-03-27 21:36 gb * [r1051] : Increment min fasl version. 2005-03-27 21:30 gb * [r1050] : Increment fasl version. 2005-03-27 21:22 gb * [r1049] : Bump. 2005-03-27 21:22 gb * [r1048] : Increment $fasl_vers. 2005-03-27 21:17 gb * [r1047] : Don't define old-rcontext. 2005-03-27 21:17 gb * [r1046] : Stop maintaining old_rcontext. Don't need special get_interrupt_tcr() on Linux; review other TCR-related cruft. 2005-03-27 20:38 gb * [r1045] : r13 is rcontext; maintain r2 (old_rcontext) in parallel, until we're bootstrapped. 2005-03-27 20:38 gb * [r1044] : r13 is rcontext; r2 (maintained by the kernel) is old-rcontext, until we stop referencing it. 2005-03-27 19:03 gb * [r1043] : Stop maintaining OLD-VSP; R13 will soon (one more round of bootstrapping) become RCONTEXT, and then the fasl version will change and there'll be new binaries. 2005-03-27 18:27 gb * [r1042] : Maintain OLD-VSP, not NEW-VSP. 2005-03-27 18:15 gb * [r1041] : VSP is now r15, OLD-VSP is r13. Keep them in sync. Don't use PPC2-FIX-STACKREG or vinsn ADJUST-STACK-REGISTER, for bootstrapping reasons. 2005-03-27 18:08 gb * [r1040] : VSP is now r15, OLD-VSP is r13. 2005-03-27 18:04 gb * [r1039] : VSP is now r15, OLD-VSP is r13. They should be equivalent until we're finished bootstrapping; use check_vsp to enforce that at discrete execution points. 2005-03-27 14:40 gb * [r1038] : Ensure vsp = new-vsp, harder. 2005-03-27 14:39 gb * [r1037] : do check_vsp on subprim entry. 2005-03-27 14:38 gb * [r1036] : define check_vsp (trap if vsp != new_vsp) 2005-03-27 13:55 gb * [r1035] : Really no more old-fn. R15 is new-vsp; try to keep it in synch with VSP (soon to become "old-vsp") at all times. 2005-03-27 13:01 gb * [r1034] : No more old-fn. 2005-03-27 11:53 gb * [r1033] : Copy FN to OLD-FN in %APPLY-LEXPR-TAIL-WISE. 2005-03-27 11:52 gb * [r1032] : Lisp believes that FN = R16, copies it to OLD-FN whenever it's set (for bootstrapping). 2005-03-27 10:04 gb * [r1031] : Remove remaining uses of temp4; define r16 as "new_fn". All assignments to fn copy to new_fn, as well (more bootstrapping). 2005-03-27 07:50 gb * [r1030] : Back out of previous change, since even current versions of Debian unstable don't provide a TLS-aware gcc/glibc. Set TCR_BIAS to 0 where relevant. Start to take another approach (moving rcontext to r13), by starting to phase out temp4 (= r16). Don't let the compiler allocate temp4, don't use temp4 in LAP code. 2005-03-27 03:35 gb * [r1029] : Bump. 2005-03-27 03:34 gb * [r1028] : Start to bootstrap changes to reconcile OpenMCL's notion of thread-specific-data (the TCR) with Linux's. On Darwin, this means that r2 (rcontext) points #x7000 bytes beyond the TCR; on Linux, the TCR structure will be the first (real) thread-local variable, and the C runtime's thread pointer will therefore point ... #x7000 bytes beyond the TCR. (This basically means that we can keep the same value in r2/rcontext when running lisp and foreign code on Linux; Linux C and thread and debugging utilities are getting increasingly insistent about that.) This introduces binary incompatibility; there'll need to be a change in the fasl version and new images when the dust settles a bit. 2005-03-27 03:20 gb * [r1027] : Hmm. Not likely that MultiFinder/UniFinder are involved anymore. 2005-03-20 22:14 gb * [r1026] : cvs2cl --ignore ChangeLog ... 2005-03-20 22:11 gb * [r1025] : *** empty log message *** 2005-03-19 05:00 gb * [r1024] : Try to use .size directive on ELF. 2005-03-19 04:59 gb * [r1023] : Export *IGNORE-EXTRA-CLOSE-PARENTHESIS*. 2005-03-19 04:58 gb * [r1022] : Are we 0.14.3 yet ? 2005-03-19 04:57 gb * [r1021] : By default, signal errors on dangling close parens. 2005-03-19 04:57 gb * [r1020] : Doc string for TRACE: don't describe obscure SBCL options. Or obscure OpenMCL options, either 2005-03-19 04:56 gb * [r1019] : Add missing #\Newline at EOF. 2005-03-19 04:48 gb * [r1018] : no more eval, step 2005-03-19 04:07 gb * [r1017] : ignore fasls 2005-03-19 03:42 gb * [r1016] : Word-size and integer/pointer 64-bit issues. Some 64-bit Mach exception stuff (needs more work.) 2005-03-19 03:41 gb * [r1015] : Word-size and integer/pointer 64-bit issues. Use vm_region64 if applicable; someday soon, use it with reasonable parameters. 2005-03-19 03:41 gb * [r1014] : Word-size and integer/pointer 64-bit issues. 2005-03-19 03:24 gb * [r1013] : Word-size and integer/pointer 64-bit issues. 2005-03-19 03:20 gb * [r1012] : ppc64 C frame (may be the same on linux/darwin.) 2005-03-19 03:19 gb * [r1011] : ppc64 inline asm for count_leading_zeros. (Still lots of 32-bit assumptions in callers ...) 2005-03-19 03:18 gb * [r1010] : Include lisp-exceptions.h, for some obscure reason ... 2005-03-19 03:18 gb * [r1009] : Some 64-bit fixes. 2005-03-19 03:07 gb * [r1008] : some extra casts 2005-03-19 02:53 gb * [r1007] : work around some new bugs in ld64 2005-03-18 06:55 gb * [r1006] : Use printf-style '%', not format-style '~', in MACH_CHECK_ERROR. Try to get some thread state stuff for PPC64. Complain if we can't get thread state in create_thread_context_frame. When deallocate thread exception port before destroying it, just in case. Hold the mach_exception_lock_set lock before suspending in suspend_tcr. 2005-03-18 06:51 gb * [r1005] : Keep trying to get xFindSymbol to compile without warnings on current and future OS releases. 2005-03-13 11:47 gb * [r1004] : Add -maltivec to ASFLAGS. 2005-03-09 04:15 gb * [r1003] : Ensure that the XER is zeroed in the shared ff_call return code (doh!) and in a slightly more subtle case involving Linux syscall return: it seems that the XER[SO] bit can be set when the CR0[SO] bit is clear on syscall return, and we weren't zeroing the XER in that case (causing spurious arithmetic overflows on subsequent arithmetic operations.) 2005-03-09 04:05 gb * [r1002] : Move YIELD, READ-LOCK-RWLOCK, and WRITE-LOCK-RWLOCK to lisp code, mostly 'cause I don't want to have to trust inline syscalls from LAP. 2005-03-09 04:04 gb * [r1001] : Move YIELD, READ-LOCK-RWLOCK, and WRITE-LOCK-RWLOCK here (from ppc-misc), mostly 'cause I don't want to have to trust inline syscalls from LAP. 2005-03-07 03:55 gb * [r1000] : After doing inline syscalls, ensure that the XER (especially the XER[SO] bit) is cleared. 2005-03-05 09:08 bryan * [r999] : *** empty log message *** 2005-03-05 09:07 bryan * [r998] : when *print-array*, use #nA(..) syntax even when array-element-type is something other than t. 2005-03-05 08:14 bryan * [r997] : *** empty log message *** 2005-03-05 08:13 bryan * [r996] : print-unreadable-object should accept t (*terminal-io*) and nil (*standard-output*) for the stream. 2005-03-05 06:56 bryan * [r995] : *** empty log message *** 2005-03-05 06:55 bryan * [r994] : %print-unreadable-object should always return nil. when type is t, %print-unreadable-object should always print a space after the type not just when it's followed by more stuff. (seems less aesthetically pleasing, but that's what the spec says). 2005-03-04 21:50 bryan * [r993] : *** empty log message *** 2005-03-04 21:46 bryan * [r992] : ioblock-element-shift should never be < 0. 2005-03-03 06:36 bryan * [r991] : initial version of retroactive ChangeLog. 2005-03-03 06:28 bryan * [r990] : bump. 2005-03-03 03:05 bryan * [r989] : when *print-escape*, pathnames always print starting with #P. (per clhs 22.1.3.11) 2005-03-03 02:18 bryan * [r988] : when *print-array* is nil, print bit-vectors unreadably. 2005-03-03 01:04 bryan * [r987] : don't do *print-level* truncation for bit-vectors and strings. 2005-03-02 23:42 bryan * [r986] : write-a-cons: when *print-length*=0, write "..." for the entire cons. when the cons is a dotted pair, don't decrement print-length before printing the cdr. 2005-03-02 21:49 bryan * [r985] : when *print-length* would cause "..." to be printed immediately following "#(", don't print a space. thus "#(...)", not "#( ...)". 2005-03-02 20:42 bryan * [r984] : don't insert commas inbetween padchars in #\r format directive. 2005-03-01 07:41 bryan * [r983] : bump. 2005-03-01 07:40 bryan * [r982] : defclass: signal program error when a slot is initialized more than once in :default-initargs. defclass: make duplicate slot name program error a bit more verbose. 2005-03-01 06:06 bryan * [r981] : really big BIGNUMS shouldn't cause describe to FLOATING-POINT-OVERFLOW. instead just print the fact that it would if you tried. 2005-02-28 06:43 gb * [r980] : Build on 64-bit DarwinPPC. Work around some linker lossage. 2005-02-28 06:42 gb * [r979] : Remove some obsolete stuff. Try to handle cases where OS libs are gratuitously deleted and added in unnamed, undiscussed OS releases. 2005-02-28 06:35 gb * [r978] : Some word size/signedness cleanup. Use dlsym() in cases where we're sure that it exists. 2005-02-28 06:34 gb * [r977] : Rename 'dladdr' to 'darwin_dladdr', in case The OS Release That Can Not Be Discussed defines its own dladdr function. 2005-02-28 06:32 gb * [r976] : Try to work around the (old, now fixed) OSX G5 kernel signal handler bug, by flipping an unused bug in the XER. 2005-02-28 06:31 gb * [r975] : Some word size/signedness cleanup. 2005-02-28 06:30 gb * [r974] : Don't prototype print_lisp_object(). 2005-02-28 06:29 gb * [r973] : Be fussy about word size, signedness. (Needs more fussiness.) 2005-02-28 06:29 gb * [r972] : User PTR macro for imported addresses (word size issues.) 2005-02-28 06:28 gb * [r971] : Tweak FUN .stabs handling, to make fussy assemblers (that can't be discussed in public yet) happy. 2005-02-28 06:23 gb * [r970] : no more pantherg5 stuff 2005-02-28 06:20 gb * [r969] : Force 12 bit alignment. (See if this breaks Linux.) 2005-02-28 06:18 gb * [r968] : Remove obsolete INSTANCE-INITIALIZE methods. 2005-02-27 22:25 bryan * [r967] : width parameter <=0 passed to ~f format directive should act as if it never was specified. previously (when 0) would loop. 2005-02-26 03:21 gb * [r966] : Export WITH-POINTER-TO-IVECTOR from CCL; export a few missing things from OPENMCL-MOP. 2005-02-26 03:20 gb * [r965] : WITH-POINTER-TO-IVECTOR. 2005-02-26 03:18 gb * [r964] : Add UVECTORP & IVECTORP; make them type predicates. 2005-02-26 02:40 gb * [r963] : Handle the DOUBLE-FLOAT case(es) in %VECT-DATA-TO-MACPTR. 2005-02-26 02:36 gb * [r962] : %MAKE-HEAP-IVECTOR: element count not optional. 2005-02-26 01:00 gb * [r961] : Bump. 2005-02-26 00:59 gb * [r960] : Add an 'extern' to definition of debug_command_entries[]. In lisp_debugger(), define 'state' before executable body. 2005-02-26 00:57 gb * [r959] : A few name changes, for consistency with kernel notion of kernel-globals. 2005-02-26 00:56 gb * [r958] : Handle changes to DEFCLASS's processing of random slot options (:FOREIGN-TYPE). 2005-02-26 00:55 gb * [r957] : DEFCLASS: only collect random slot option values into a list if option was provided more than once. 2005-02-26 00:52 gb * [r956] : Preserve values of class slots (:ALLOCATION :CLASS) during class redefinition. 2005-02-26 00:51 gb * [r955] : Bind *LOADING-FILE-SOURCE-FILE* around calls to %FASLOAD. 2005-02-26 00:50 gb * [r954] : Define *LOADING-FILE-SOURCE-FILE* elsewhere. 2005-02-26 00:49 gb * [r953] : Define *LOADING-FILE-SOURCE-FILE* here. 2005-02-26 00:48 gb * [r952] : Kernel globals in PPC (not PPC32) package. 2005-02-23 08:01 bryan * [r951] : bump. 2005-02-22 23:24 gb * [r950] : Load f1 from the right stack offset when returning from EABI callback. 2005-02-22 23:01 gb * [r949] : All callback variants try to return values in r3, r4, and f1. 2005-02-22 22:59 gb * [r948] : Handle float returns in callbacks (needs related kernel changes; untested on Linux.) 2005-02-22 20:26 bryan * [r947] : when *pretty-print* and *print-base* > 10, use the slightly slower #'write-a-symbol to properly determine potential numbers. 2005-02-22 20:22 bryan * [r946] : use *print-base*, not *read-base*, when deciding whether to quote a potential number. 2005-02-21 01:47 bryan * [r945] : handle ~v when nil for ~{...~} 2005-02-20 23:25 bryan * [r944] : bump. use this version to check for stream-external-format being a generic function. 2005-02-20 23:24 bryan * [r943] : rewrite stream-external-format as a generic function. 2005-02-20 22:36 bryan * [r942] : treat ~vr as ~r when arg is nil. per clhs 22.3. 2005-02-20 20:33 bryan * [r941] : remove unused code: (defmethod file-stream-external-format ((s broadcast-stream)) 2005-02-20 10:08 bryan * [r940] : bump. 2005-02-20 10:05 bryan * [r939] : treat ~v[...~] as ~[...~] when arg is nil. per clhs 22.3. 2005-02-20 02:13 bryan * [r938] : somebody misses printf.. format uses ~ and not %. 2005-02-19 23:51 bryan * [r937] : reverse ~:@C printing per clhs: "~:@C prints what ~:C would, and then if the character requires unusual shift keys on the keyboard to type it, this fact is mentioned." 2005-02-19 22:34 bryan * [r936] : add Fs (file separator), Gs (group separator), Rs (record separator), and Us (unit seperator) to *name-char-alist*. #\Fs to avoid printing something unescaped #\^\, the others for completeness. the extra four entries in the alist shouldn't have any noticeable effect. probably could fix the printer (and reader?) to handle the escaped case, but i like seeing the textual representation. 2005-02-19 21:29 bryan * [r935] : when printing a number with a sign and padding, print padding first and then the sign. 2005-02-19 05:42 bryan * [r934] : bump for #A changes. 2005-02-18 23:19 bryan * [r933] : rewrite #A dispatch handler to allow for all types of sequences (bitfields, strings, vectors, lists) when they make sense for the supplied rank. 2005-02-18 00:02 gb * [r932] : Observe *READ-SUPPRESS* in reader macros. 2005-02-17 01:42 bryan * [r931] : hook asdf into *module-provider-functions*. 2005-02-16 23:15 bryan * [r930] : bump. 2005-02-16 21:09 bryan * [r929] : add docstrings to the majority of common-lisp-user symbols starting with a snapshot of those found in SBCL 0.8.18. 2005-02-16 20:34 bryan * [r928] : initial repository version from 050201 import. 2005-02-16 20:29 bryan * [r927] : unmatched ') when ccl::*ignore-extra-close-parenthesis* is nil should signal a reader-error. the default for ccl::*ignore-extra-close-parenthesis* remains t. set it to nil in your init file for maximum ansi-compliance. 2005-02-16 20:12 bryan * [r926] : sync with latest asdf-install code. 2005-02-16 19:53 bryan * [r925] : import latest ASDF from clocc repository. identical (besides revision number) to asdf.lisp found in SBCL 0.8.18. 2005-02-16 19:50 bryan * [r924] : turn on *save-doc-strings* and *fasl-save-doc-strings*. 2005-02-16 19:45 bryan * [r923] : ccl::%unreadable should signal a reader-error. ccl::%unreadable is called when the reader encounters #<, #), or #. 2005-02-16 19:44 bryan * [r922] : ## when *read-suppress* should return nil, not (values). 2005-02-16 19:43 bryan * [r921] : #nr shouldn't type-check n when *read-suppress*. 2005-02-16 19:42 bryan * [r920] : #( when *read-suppress should ignore array length problems. 2005-02-16 19:41 bryan * [r919] : rewrite #* handler #* must signal a reader-error if any of the token characters are not 0 or 1. previously it would stop at the first character that was not 0 or 1. single- and multi-escape are not allowed in the token for #*. #n* when n > 0 must have at least one element. the number of elements can not exceed n. except if *read-suppress*, then anything goes until the next whitespace or terminating macro. 2005-02-16 19:39 bryan * [r918] : more occurrences where error should be a reader-error. 2005-02-16 19:35 bryan * [r917] : fill in some missing methods for echo-streams and concatenated-streams. (defmethod stream-read-char-no-hang ((s echo-stream)) (defmethod stream-read-byte ((s echo-stream)) (defmethod stream-read-line ((s echo-stream)) (defmethod stream-read-vector ((s echo-stream) vector start end) (defmethod stream-read-char-no-hang ((s concatenated-stream)) (defmethod stream-peek-char ((s concatenated-stream)) (defmethod stream-read-line ((s concatenated-stream)) (defmethod stream-read-list ((s concatenated-stream) list count) (defmethod stream-read-vector ((s concatenated-stream) vector start end) 2005-02-16 19:31 bryan * [r916] : ccl::hash-pname loses when given a non-simple-string find-all-symbols did no checking and didn't massage. added a ccl::ensure-simple-string to both check and massage. make-symbol checked that the arg was a 'simple-string using #'require-type. replacing that with ccl::ensure-simple-string both checks and massages. ccl::pkg-arg (called by a bunch of symbol iterator macros) did check that if the arg was a string, that it was a simple string. i rewrote the initial if form as a cond that will call ccl::ensure-simple-string if the arg is indeed a string. also updated ccl::new-package-name. 2005-02-16 19:22 bryan * [r915] : comma not inside backquote should signal reader-error. 2005-02-16 19:20 bryan * [r914] : if *read-suppress*, return NIL from read(-*) forms. 2005-02-16 19:19 bryan * [r913] : adjustable arrays should print readably when *print-array* (per clhs 22.1.3.7-8) 2005-02-16 19:16 bryan * [r912] : don't declare dynamic-extent for the unused args passed to (funcall (formatter "~A") ..) which end up being returned to the caller. 2005-02-16 19:11 bryan * [r911] : "" is a substring to all strings, update %apropos-substring-p to explicitly check for that and return t before doing any more work. 2005-02-16 19:09 bryan * [r910] : a string-designator can also be a singleton character, update string-argp to reflect that. 2005-02-16 19:05 bryan * [r909] : typecase was always pushing the t clause (if present) to the end of the clauses. i suspect that most cases when it appears before the end, the user intended for it to be equivalent to a final OTHERWISE clause. hopefully the warning messages about later clauses being shadowed will be sufficient to alert them. as an added benefit it dead-code strips any cases following a t-clause. finally it now correctly raises an error if OTHERWISE (if present) is not the final clause. 2005-02-16 18:56 bryan * [r908] : ccl::simple-vector-delete loses with displaced arrays with a displaced-index-offset > 0. the second loop (looping through the scratch bit vector) to copy the items that survive into the new array misuses the offset with the bit vector. this causes an array out of bounds error. instead it should use the offset with the original array to reference the item. the third loop which copies the remaining items that were explicitly excluded from the test also needs the offset to copy the right items. 2005-02-16 18:52 bryan * [r907] : allow mincol, minpad in format directives to be negative. 2005-02-16 18:46 bryan * [r906] : non-list args in ~{..~} format directive should signal type-error. 2005-02-16 18:41 bryan * [r905] : allow #\+ in format directives. 2005-02-16 18:40 bryan * [r904] : modify defppclapfunction to accept a docstring. 2005-02-16 07:24 bryan * [r903] : don't update index if with-input-from-string exits abnormally. 2005-02-16 07:17 bryan * [r902] : spec says that the long form of defsetf encloses the body in a block named after the access function. defsetf was doing the wrong thing and putting the block around the eventual setf-expansion instead of the body that's evaluated to generate the setf-expansion. defsetf should also return the name of the access-fn and not the doc string. 2005-02-16 02:23 bryan * [r901] : accept #:. per CLHS and ansi-test SYNTAX.SHARP-COLON.7 2005-02-16 02:14 bryan * [r900] : %hash-table-equalp uses :test arg instead of equalp. spec says: equalp descends hash-tables by first comparing the count of entries and the :test function; if those are the same, it compares the keys of the tables using the :test function and then the values of the matching keys using equalp recursively. 2005-02-16 02:04 bryan * [r899] : suppress unused lexical variable warning in #: handler 2005-02-16 01:58 bryan * [r898] : clhs 22.1.3.8 says an array is printed started with #A. 2005-02-16 01:52 bryan * [r897] : clhs 22.1.3.1.4 says a complex number is printed started with #C. 2004-10-27 06:56 gb * [r896] : Hacks to keep fp exceptions enabled after pthread_kill() on Darwin. Allways enable fp exceptions when calling back into lisp. 2004-10-27 05:28 gb * [r895] : back out of earlier change, now that signal_handler calls get_interrupt_tcr 2004-10-27 05:23 gb * [r894] : Use get_interrupt_tcr() in signal_handler on linux. 2004-10-27 05:22 gb * [r893] : get_interrupt_tcr() takes a "create-p" argument, like some callers thought it did. 2004-10-27 04:51 gb * [r892] : play around with r2 when zeroing FPSCR 2004-10-06 22:44 gb * [r891] : Alex Crain's patches (paste in listener, typeout window size) 2004-10-01 11:03 gb * [r890] : if a CERROR offers to change the class of an existing GF, actually call CHANGE-CLASS to do so 2004-09-26 00:38 gb * [r889] : Use a real SHARED-INITIALIZE after CLOS is bootstrapped. Default superclasses based on metaclass, in (SHARED-INITIALIZE :AFTER) method on SLOTS-CLASS. Don't add use :INITARGS for most SLOTS-CLASS slots (the initargs to SHARED-INITIALIZE :AFTER should generally set them to computed values.) Make sure that the right version of UPDATE-SLOTS is committed. 2004-09-26 00:31 gb * [r888] : New support for funcallable instances, from Tim Moore. 2004-09-26 00:31 gb * [r887] : INSTANCE-SLOTS knows where to find slots in funcallable instances that aren't GF's (from Tim Moore.) 2004-09-25 21:38 gb * [r886] : Use %HEAP-IVECTOR-P when checking bogosity. 2004-09-25 21:37 gb * [r885] : Kinder, gentler heap-ivector stuff. Use it. 2004-09-25 21:37 gb * [r884] : Bump. 2004-09-25 21:36 gb * [r883] : Export MAKE-HEAP-IVECTOR and DISPOSE-HEAP-IVECTOR. 2004-09-25 21:36 gb * [r882] : DEFCLASS macro: handle :DOCUMENTATION as a single option, if provided. Don't default superclasses (from Tim Moore.). Pass class name, not class, in :METACLASS option. 2004-09-25 21:16 gb * [r881] : FUDGE-HEAP-POINTER: user clrrwi to clear three bits. 2004-09-25 21:15 gb * [r880] : WITH-POPUP-DISPLAY: use typeout-stream (from Alex Crain.) 2004-09-25 21:14 gb * [r879] : Indentation change. 2004-09-25 21:13 gb * [r878] : Send backtrace info to typeout window on error (from Alex Crain.) 2004-09-25 21:12 gb * [r877] : Typeout-window changes, from Alex Crain. 2004-09-25 21:08 gb * [r876] : New file, from Alex Crain 2004-09-25 21:08 gb * [r875] : Fixes to COERCE-TO-FOREIGN-TYPE, from Randall Beer. 2004-09-25 21:07 gb * [r874] : Changes from Alex Crain, for typeout window. 2004-09-25 21:07 gb * [r873] : new file 2004-09-19 23:21 gb * [r872] : use thread_abort_safely() when suspending Mach threads 2004-09-13 07:26 gb * [r871] : wait_on_semaphore: return KERN_OPERATION_TIMED_OUT if wait was aborted and we'd timed out anyway. suspend_tcr: use mach_suspend_tcr() preferentially on Darwin. 2004-09-13 07:21 gb * [r870] : mach_resume_tcr: it would be a good idea to actually resume the thread, not just do the bookkeeping (a call to thread_resume() was missing.) 2004-08-30 22:20 gb * [r869] : Lots of changes, mostly: - try to speed up callbacks that fetch characters from the virtual nsstring - kill echo-area buffers when closing windows - windows should be released when closed; documents should close when the last window closes. - try not to invalidate the buffer cache on insertions - disable background layout before closing the window (in case that's still going on ...) - clear the buffer's modified state when buffer is saved (not sure if the modeline indicator's updated correctly) - DISABLE-BLINK tells the layout manager to force redisplay of the (now non-blinking) character. 2004-08-30 21:59 gb * [r868] : Assert that SELF is an instance in instance method (should also assert that :ID-typed parameters are instances/classes.) 2004-08-30 21:58 gb * [r867] : set pointer types directly 2004-08-30 21:34 gb * [r866] : Use (THE ...) to help the bridge resolve some messages. 2004-08-30 21:31 gb * [r865] : Use DEFPARAMETER to initialize *DEFAULT-MODELINE-FIELDS*, so it's a little easier to redefine. 2004-08-30 21:30 gb * [r864] : Wrap some of the indentation code in a MODIFYING-BUFFER, to defer display updates. (This is experimental, and should probably be made more general.) 2004-08-30 21:29 gb * [r863] : Don't be quite so eager to clear the echo area, especially when it's the current buffer. 2004-08-30 21:25 gb * [r862] : Make (:* T) be shorthand for (:* :VOID). May lead to (benign) redefinition warnings. 2004-08-30 21:25 gb * [r861] : Remove some FRED-isms. 2004-08-30 21:22 gb * [r860] : DEREFERENCE-BASE-STRING: third return value is just the length (not length + displacement). 2004-08-11 00:54 gb * [r859] : Check for local macro functions (and complain if FUNCTION references them) in cheap evaluator. Pass a function and other information to *MACROEXPAND-HOOK* when expanding symbol-macros. 2004-08-11 00:19 gb * [r858] : late version of %MAKE-METHOD-INSTANCE, from HEL 2004-08-06 07:40 gb * [r857] : distinguish finding NIL from not finding symbol in MAYBE-PRINT-FAST 2004-08-03 04:56 gb * [r856] : Remove subviews from hemlock-frame (window) before closing it. 2004-08-02 00:54 gb * [r855] : INIT-WITH-TEXT-STORAGE -> SET-TEXT-STORAGE (was initing twice.) 2004-07-31 23:52 gb * [r854] : package-name-change-hook 2004-07-31 23:50 gb * [r853] : handle modeline update in %SET-BUFFER-MODIFIED 2004-07-31 23:44 gb * [r852] : remove a debugging call to FORMAT 2004-07-31 23:43 gb * [r851] : modeline buffer-update zeal 2004-07-31 23:40 gb * [r850] : *fake-stack-frames* is thread-specific, just in case 2004-07-31 23:39 gb * [r849] : %%FRAME-BACKLINK: this isn't 0S7 2004-07-31 23:38 gb * [r848] : move %NEW-PTR here 2004-07-31 23:37 gb * [r847] : %new-ptr can't exist in level-0 2004-07-31 23:36 gb * [r846] : make-load-form for foreign-variable istruct 2004-07-31 23:35 gb * [r845] : fix dangling close paren 2004-07-26 01:15 gb * [r844] : newer code from Hamilton 2004-07-23 22:15 gb * [r843] : new files 2004-06-27 08:55 gb * [r841] : early NSThread 2004-06-27 08:55 gb * [r840] : defer window creation 2004-06-27 08:55 gb * [r839] : class canonicalization, for resuming images 2004-06-27 08:55 gb * [r838] : new version from Barry Perryman 2004-06-27 08:55 gb * [r837] : canonicalize classes referenced in type signatures 2004-06-24 04:24 gb * [r836] : Remove a (debugging) call to FORMAT. 2004-06-23 17:32 gb * [r835] : New external-process waiting stuff. 2004-06-23 17:32 gb * [r834] : Initial restarts have descriptive :report clauses. 2004-06-23 17:30 gb * [r833] : Gary King's fix to FIND-LOAD-FILE. 2004-06-23 17:30 gb * [r832] : INVOKE-RESTART-INTERACTIVELY works with SIMPLE-RESTARTs again. 2004-06-23 17:29 gb * [r831] : UPDATE-SLOTS fix. 2004-06-15 09:40 gb * [r830] : added 2004-06-06 09:12 gb * [r829] : Don't sleep/poll while monitoring external process; signal completion via semaphore. 2004-06-06 09:11 gb * [r828] : Ignore NAME in SHLIB-CONTAINING-ADDRESS for Linux. 2004-06-06 06:01 gb * [r827] : Ensure TSP valid before asserting lisp-mode in ffi return. 2004-06-06 06:00 gb * [r826] : New post-gc freeing mechanism (suspened thread may own malloc lock.) Similar mechanism for TCR cleanup. 2004-06-06 05:59 gb * [r825] : New post-gc freeing mechanism (suspened thread may own malloc lock.) 2004-06-06 05:57 gb * [r824] : Do external processes differently. 2004-06-06 05:57 gb * [r823] : No periodic task for external processes. 2004-06-06 05:56 gb * [r822] : Don't pass (%CURRENT-TCR) to stack-walking functions in break loop commands. 2004-06-06 05:55 gb * [r821] : :prompt-function takes an argument. 2004-05-23 23:37 gb * [r820] : In GET-DESCRIPTOR-FOR, leave room for a #\Newline when copying an arbitrary input stream to a temporary file. 2004-05-08 20:21 gb * [r819] : Define UI-OBJECT-DO-OPERATION for *NSApp*. 2004-05-08 20:20 gb * [r818] : Use #?. Find cfstrings in __const section, too. 2004-05-08 20:19 gb * [r817] : Use #?. SET THE UI OBJECT (lost in 0.14.2 merge.) 2004-05-08 20:18 gb * [r816] : Use #?. 2004-05-07 16:56 gb * [r815] : Fix duplicate HI::SAVE-HEMLOCK-DOCUMENT. Don't parse buffer unless we're looking at a paren when trying to determine blink location. 2004-05-07 08:23 gb * [r814] : Handle new issues with :foreign-type (non-standard slot options collect values into a list.) 2004-05-07 08:22 gb * [r813] : Date tag for development. 2004-05-06 06:29 gb * [r812] : incorporate 0.14.2 changes 2004-05-02 07:28 gb * [r811] : save-as, open-document for Hemlock. Fix off-by-one in paren-blinking; lock view focus when disabling blink to fix display glitch. 2004-05-02 07:27 gb * [r810] : Use Cocoa for more file operations. 2004-05-01 02:34 gb * [r809] : SETF shows arglist in error message, if odd length. 2004-05-01 02:34 gb * [r808] : Bump the version. 2004-05-01 02:34 gb * [r807] : Font-mark/font-region stuff. 2004-05-01 02:33 gb * [r806] : Add a STREAM-WRITE-STRING method for echo area. 2004-05-01 02:32 gb * [r805] : Don't call HI::MAYBE-UPDATE-SELECTION anymore. 2004-05-01 02:32 gb * [r804] : New history, font-region stuff. 2004-05-01 02:31 gb * [r803] : Arglist stuff. 2004-05-01 02:30 gb * [r802] : Clear echo area on entry to command. 2004-05-01 02:30 gb * [r801] : Check buffer modification. 2004-05-01 02:30 gb * [r800] : Remove old selection-updating code (was unused.) 2004-05-01 02:29 gb * [r799] : Partial integration with new font-region stuff. 2004-05-01 00:53 gb * [r798] : Partial integration with Cocoa document handling. 2004-05-01 00:53 gb * [r797] : Still not working, but closer. 2004-05-01 00:52 gb * [r796] : Get rid of some (very) old code. CENTER-TEXT-PANE via Cocoa. 2004-05-01 00:51 gb * [r795] : FONT-REGION stuff. 2004-05-01 00:51 gb * [r794] : arglist-on-space, current-function-arglist. 2004-05-01 00:50 gb * [r793] : Lots-o-changes ... 2004-05-01 00:49 gb * [r792] : Synthesize obliqueness/boldness. 2004-04-30 19:58 gb * [r791] : %MAKE-CONSTANT-NSSTRING when restoring constant NSStrings. 2004-04-19 19:31 gb * [r790] : Lots of changes/fixes: selection, view-size, scrolling, paren-matching, etc. 2004-04-19 19:30 gb * [r789] : Event-queue-node constructor. 2004-04-19 06:49 gb * [r788] : Don't use %CURRENT-TCR, in %LAST-FN-ON-STACK. 2004-04-13 19:26 gb * [r787] : New fields in backtrace context. 2004-04-13 19:26 gb * [r786] : Simple loader, for now. 2004-04-13 19:25 gb * [r785] : Bump. 2004-04-13 19:25 gb * [r784] : New stack-walking scheme: pass optional backtrace context, not TCR. 2004-04-13 19:24 gb * [r783] : region now per-buffer. 2004-04-13 19:23 gb * [r782] : Less hysteria writing to echo-area stream. 2004-04-13 19:23 gb * [r781] : Remove some CLXisms. 2004-04-13 19:22 gb * [r780] : Better EVAL-REGION. 2004-04-13 19:22 gb * [r779] : Reorganize; new per-buffer active-region scheme. 2004-04-13 19:21 gb * [r778] : Immutable key-event (not dll-node.) 2004-04-13 19:21 gb * [r777] : Always delete current-region before insertion. 2004-04-13 19:20 gb * [r776] : Don't call DISPLAYED-P. 2004-04-13 19:18 gb * [r775] : Remove a few CLXisms. 2004-04-13 19:18 gb * [r774] : Mark setup for new scheme. 2004-04-13 19:17 gb * [r773] : Key events are interned, so don't make them mutable. 2004-04-13 19:16 gb * [r772] : #k"Enter" evaluates region, not "defun". 2004-04-13 19:16 gb * [r771] : Lots-o-changes: double-click, paste, etc. 2004-04-13 19:15 gb * [r770] : (Partial) fixes to struct-returning method definitions. 2004-04-13 19:14 gb * [r769] : Remove some commented-out code. 2004-04-12 19:29 gb * [r768] : Erik Pearson's SOCKET-ERROR changes/fixes. 2004-04-08 04:32 gb * [r767] : remove backup files 2004-04-08 03:49 gb * [r766] : Call #_GetCurrentEventQueue before some other thread calls it ... Make "void" NSThread early, to tell AppKit we're multi-threaded. 2004-04-08 03:46 gb * [r765] : Defaults changes. 2004-04-08 03:45 gb * [r764] : Defaults changes. Support clipview around echo-area (horizontal scrolling). 2004-04-08 03:43 gb * [r763] : Changes to support constraints, prefs dialog. 2004-04-08 03:42 gb * [r762] : New file. 2004-04-08 03:41 gb * [r761] : Require "COCOA-PREFS". Doc string on defaults; handle showPreferences: message. 2004-04-08 03:38 gb * [r760] : Handle control-g (abort-event) in GET-KEY-EVENT. 2004-04-08 03:35 gb * [r759] : Add showPreferences: action, connect Preferences item to it. 2004-04-03 06:08 gb * [r758] : Use readlink() to determine executable pathname under Linux. Use PATH_MAX on both platforms. 2004-04-03 05:53 gb * [r757] : Determine executable pathname (needs to use /proc on Linux.) 2004-04-03 05:53 gb * [r756] : Hysteresis in stack-overflow recovery. (Needs testing.) 2004-04-03 05:52 gb * [r755] : Add bt.break-condition. 2004-04-03 05:51 gb * [r754] : Break condition comes from context now. 2004-04-03 05:50 gb * [r753] : %ACTIVE-AREA is deprecated. 2004-04-03 05:49 gb * [r752] : Backtrace context now contains the break condition. 2004-04-03 05:47 gb * [r751] : Fencepost in %PTR-IN-AREA-P. Deprecate (and remove some calls to) %ACTIVE-AREA, since stacks have been contiguous for a couple of years now ... 2004-04-03 05:44 gb * [r750] : Check for circularities when checking for forward-referenced superclasses. Redefine %ADD-METHOD as ADD-METHOD when we're able to. ENSURE-CLASS-USING-CLASS of a forward-referenced class passes initargs to CHANGE-CLASS. 2004-03-28 01:51 gb * [r749] : Umm, background color components are of type :float. 2004-03-27 11:08 gb * [r748] : remove backup files from CVS 2004-03-27 11:05 gb * [r747] : Not sure what changed, but IB found inconsistencies and wanted the .nib to be saved. 2004-03-27 10:59 gb * [r746] : Not sure what changed, but IB found inconsistencies and wanted the .nib to be saved. 2004-03-27 10:58 gb * [r745] : Integration. 2004-03-27 10:58 gb * [r744] : Color stuff, redisplay changes. Use a sheet to display error messages from Hemlock command threads. 2004-03-27 10:57 gb * [r743] : Color stuff; maintain backtrace contexts in threads. Menu action functions, validation for interrupt and backtrace. 2004-03-27 10:55 gb * [r742] : Require COCOA-BACKTRACE. Move UI-DO-OPERATION here, implement backtrace operations. 2004-03-27 10:54 gb * [r741] : Don't install deallocate hook, for the time being. 2004-03-27 10:53 gb * [r740] : Update the selection before grabbing the next key-event. Timed-wait on next event, for EDITOR-SLEEP. 2004-03-27 10:52 gb * [r739] : lock/unlock the buffer (actually the gap-cache) when editing. 2004-03-27 10:52 gb * [r738] : MAYBE-UPDATE-SELECTION: doesn't do anything anymore; should be removed. 2004-03-27 10:51 gb * [r737] : Active region vars are now symbol-macros. 2004-03-27 10:50 gb * [r736] : *POP-STRING* needed a newline. 2004-03-27 10:50 gb * [r735] : Use new error handler. 2004-03-27 10:50 gb * [r734] : Make EDITOR-SLEEP kind of work. 2004-03-27 10:49 gb * [r733] : Keep active region stuff in gap-cache. 2004-03-27 10:48 gb * [r732] : Fix error message typo when allocating GFs. 2004-03-27 10:47 gb * [r731] : Don't set &allow-other-keys in GF lambda lists if method only says &key. 2004-03-27 10:45 gb * [r730] : Quote more chars when converting from native namestrings. 2004-03-27 10:45 gb * [r729] : 040326. 2004-03-27 10:44 gb * [r728] : Unknown slot options in DEFCLASS: use CDR, not CADR for option value. 2004-03-27 10:43 gb * [r727] : NANOSCECONDS: always return two integers. 2004-03-26 01:13 gb * [r726] : %SET-FPSCR-CONTROL: maintain stack discipline. Write zeroes to status fields as well, to avoid enabling a signaled exception. 2004-03-25 20:10 gb * [r725] : Lisp menu 2004-03-25 03:28 gb * [r724] : 040324. 2004-03-25 01:08 gb * [r723] : PROVIDE's been in the CL package for the last decade or so. (This (small) change was made from a Hemlock/Cocoa editor !) 2004-03-25 01:07 gb * [r722] : EVAL-SELECTION ui methods. 2004-03-25 01:06 gb * [r721] : Update the buffer-cache after processing deletion notification. MAKE-HEMLOCK-BUFFER makes the gap-context, used to make all buffers except echo area. NSSTRING-TO-BUFFER needs to set up gap context. 2004-03-25 01:04 gb * [r720] : #k"enter", #k"c-m-x" bound to "Editor Evaluate Defun". 2004-03-25 01:03 gb * [r719] : Eval-region via selection queue mechanism. 2004-03-25 01:03 gb * [r718] : Fix typo in error message (rededefining non-GF as GF.) 2004-03-24 17:26 gb * [r717] : Clean this up: put things in the right package. 2004-03-24 08:51 gb * [r716] : Use new gap-cache stuff. Keep track of editing in textstorage; don't move point when editing. 2004-03-24 08:49 gb * [r715] : Get FDs right when starting a listener process. Note package correctly. 2004-03-24 08:48 gb * [r714] : Transpose Characters - do it even if at line end (may be bugs in %SET-NEXT-CHARACTER.) 2004-03-24 08:47 gb * [r713] : New shared gap scheme. 2004-03-24 08:46 gb * [r712] : Use circumflex in mark printer. %SET-NEXT-CHARACTER changes (still bugs here.) 2004-03-24 08:44 gb * [r711] : New (shared) gap-cache stuff. 2004-03-24 08:43 gb * [r710] : SELECTION-INPUT-STREAM fixes. 2004-03-24 06:34 gb * [r709] : Skip the alignment bytes when reading a (SIMPLE-ARRAY DOUBLE-FLOAT (*)). 2004-03-24 06:33 gb * [r708] : Skip the alignment bytes when writing a (SIMPLE-ARRAY DOUBLE-FLOAT (*)), but don't act as if they were included in the value returned by SUBTAG-BYTES. 2004-03-22 17:41 gb * [r707] : Process Hemlock commands in a per-frame thread. 2004-03-22 17:40 gb * [r706] : New listener process/selection-stream stuff. 2004-03-22 17:39 gb * [r705] : Initialize UI object here. 2004-03-22 17:38 gb * [r704] : UI object will be *NSApp*. 2004-03-22 17:38 gb * [r703] : Rearrange a bit. 2004-03-22 17:38 gb * [r702] : Maintain event queue in per-frame thread. 2004-03-22 17:37 gb * [r701] : Start to make this saner. 2004-03-22 17:37 gb * [r700] : Cache buffer gap variables in buffer, when passing buffer between threads. 2004-03-22 17:36 gb * [r699] : Run a %COMMAND-LOOP (in a separate thread) after all. 2004-03-22 17:35 gb * [r698] : Major mode listed before others, always. 2004-03-22 17:35 gb * [r697] : Hook for package change. Use selection stream. 2004-03-22 17:34 gb * [r696] : Maintain key-events in doubly-linked lists. 2004-03-22 17:34 gb * [r695] : Extra slots in BUFFER, to cache gap variables. 2004-03-22 17:33 gb * [r694] : Lock rings while operating on them. 2004-03-22 17:32 gb * [r693] : MAKE-MCL-LISTENER-PROCESS changes. 2004-03-22 17:32 gb * [r692] : New file; define application class, subclasses. 2004-03-22 17:31 gb * [r691] : NEW-TCR converts pointer to fixnum, even if sign bit set. 2004-03-22 17:30 gb * [r690] : MAKE-PROCESS takes :CLASS argument. 2004-03-22 17:30 gb * [r689] : All control chars, #\Space are whitespace. 2004-03-22 17:30 gb * [r688] : toplevel loop changes. 2004-03-22 17:29 gb * [r687] : Move application stuff to l1-application. 2004-03-22 17:28 gb * [r686] : Selection-stream stuff, apparently still buggy. 2004-03-22 17:27 gb * [r685] : Load l1-application. 2004-03-22 17:27 gb * [r684] : 040322. 2004-03-22 17:27 gb * [r683] : Add l1-application. 2004-03-22 17:26 gb * [r682] : Change document classes to Hemlock*. 2004-03-22 11:12 gb * [r681] : restored 2004-03-22 11:12 gb * [r680] : accidental removal 2004-03-20 23:38 gb * [r679] : Bump version to 040320. 2004-03-20 07:38 gb * [r678] : Handle key events in echo area. 2004-03-20 07:37 gb * [r677] : Revive kbdmac. 2004-03-20 07:37 gb * [r676] : Remove more clx/tty stuff. 2004-03-20 07:36 gb * [r675] : Revive, but don't create stream yet. 2004-03-20 07:36 gb * [r674] : More command-interpreter changes. 2004-03-19 18:39 gb * [r673] : set foreign thread tcr back to NULL in tsd on exit 2004-03-19 00:48 gb * [r672] : Bind current-buffer. No real difference in output-buffering stuff. 2004-03-19 00:47 gb * [r671] : Remove/comment-out some unused stuff. 2004-03-19 00:47 gb * [r670] : Use MESSAGE for editor-errors. Call command-interpreter-info-function (for extended input stuff.) 2004-03-19 00:46 gb * [r669] : Bind current-buffer when clearing echo area. 2004-03-19 00:44 gb * [r668] : Quoted-insert sets up command-info function. 2004-03-19 00:43 gb * [r667] : Command-information struct moves here; add function slot. 2004-03-19 00:43 gb * [r666] : Add echo-area stuff, start to use it. 2004-03-18 00:19 gb * [r665] : Foreign thread exit on first cleanup call. (Should all threads do so, after all ?) 2004-03-18 00:19 gb * [r664] : Return NIL (not 0/%null-ptr) if no underlying pthread. 2004-03-18 00:18 gb * [r663] : Window-close on listener thread exit: don't wait around (deadlocks during QUIT.) 2004-03-17 15:43 gb * [r662] : Suspend other threads on entry to debugger (not on entry to Bug.) 2004-03-17 15:29 gb * [r661] : Suspend other threads on entry to debugger (not on entry to Bug.) 2004-03-17 15:29 gb * [r660] : Backtrace-info changes. 2004-03-17 15:28 gb * [r659] : tsp/vsp ranges, for bogus-object detection. 2004-03-17 15:26 gb * [r658] : Backtrace/break-loop changes. 2004-03-17 15:26 gb * [r657] : LLGPL. Fix some dynamic-binding stuff; ad-hoc scheme for bogus-thing check in (some) other threads. 2004-03-17 15:24 gb * [r656] : %CATCH-TOP is always in the same place. 2004-03-17 15:23 gb * [r655] : No more *OPEN-EDITOR-DOCUMENTS*. 2004-03-17 15:22 gb * [r654] : Getting closer ... 2004-03-17 15:19 gb * [r653] : Create an NSThread early. 2004-03-17 15:18 gb * [r652] : %make-nsstring/%make-constant-nsstring. 2004-03-17 15:17 gb * [r651] : Smaller, striped ... 2004-03-12 13:24 gb * [r650] : add backtrace nib 2004-03-12 13:23 gb * [r649] : use nib 2004-03-08 19:36 gb * [r648] : Fix fatal_mach_error() prototype. 2004-03-08 14:02 gb * [r647] : Fix MACH_CHECK_ERROR. 2004-03-08 14:01 gb * [r646] : bt.stack-group -> bt.tcr 2004-03-08 14:00 gb * [r645] : Add stack-inspector stuff. 2004-03-08 13:59 gb * [r644] : SLOTS-CLASS has to look for primary-p-ness. 2004-03-08 13:58 gb * [r643] : Set GF lfun-bits if continued after congruency error. SLOTS-CLASS has to look for primary-p-ness. 2004-03-08 13:56 gb * [r642] : Factor in window creation. 2004-03-08 13:56 gb * [r641] : Factor out window creation. 2004-03-08 13:55 gb * [r640] : New file. Doesn't work ... 2004-03-07 19:53 gb * [r639] : Parameterize MAXIMUM_MAPPABLE_MEMORY. 2004-03-07 19:53 gb * [r638] : No EVAL-REDEF for %newgotag. 2004-03-07 19:52 gb * [r637] : No DEFNX1 for %newgotag. 2004-03-07 08:03 gb * [r636] : Note that use of go/block tags is deprecated. 2004-03-07 08:03 gb * [r635] : Don't use dynamic block/go tags. Forget about event-polling, etc. 2004-03-07 08:02 gb * [r634] : Don't use dynamic block/go tags. 2004-03-07 08:01 gb * [r633] : Try to bring inline with constants64.[sh] 2004-03-07 07:55 gb * [r632] : Echo areas on windows. Start the UI here when asked to. Scroll windows. 2004-03-07 07:54 gb * [r631] : Only synch if we added a new key. 2004-03-07 07:53 gb * [r630] : APPLICATION-WILL-FINISH-LAUNCHING: start the UI. 2004-03-07 07:53 gb * [r629] : Window-scrolling changes. 2004-03-07 07:53 gb * [r628] : Let Cocoa scroll windows. 2004-03-07 07:52 gb * [r627] : Per-thread stack-space calculated on a per-process basis. 2004-03-07 07:52 gb * [r626] : BREAK-LOOP creates backtrace context, tells the application when it's valid. 2004-03-07 07:50 gb * [r625] : *BACKTRACE-CONTEXTS* is bound per-thread. 2004-03-07 07:50 gb * [r624] : Foreign thread initialization needs to emulate shallow per-thread binding. 2004-03-07 07:49 gb * [r623] : 040307 2004-03-07 07:49 gb * [r622] : Lock the area_lock while threads are suspended. TCR cleanup frees tlb_pointer. 2004-03-07 07:48 gb * [r621] : darwin_exception_cleanup() frees native_thread_info (foreign exception state). 2004-03-07 07:37 gb * [r620] : Careful with the use of ptr_from_lispobj in add_lisp_base_string. Careful with the use of immheader_tag_p when printing vectors. 2004-03-06 10:10 gb * [r619] : Lots of PPC64 changes. Move image_base to #x200000 on Darwin; reserve a little under 2GB by default. Don't abort() if Mach exception-port stuff fails. 2004-03-06 10:04 gb * [r618] : Move image-base to #x200000 on Darwin. 2004-03-06 09:47 gb * [r617] : Random changes; still barely functional. 2004-03-03 23:27 gb * [r616] : Define and use COMPUTE-CLASS-PRECEDENCE-LIST. Flush initargs caches whenever class is finalized. 2004-03-03 22:20 gb * [r615] : Uninstall dealloc-hook on *SAVE-EXIT-FUNCTIONS* (run before SAVE-APPLICATION), not *LISP-CLEANUP-FUNCTIONS* (so slot-vectors stay around during shutdown.) 2004-03-03 22:18 gb * [r614] : CRLF -> LF. 2004-03-03 22:01 gb * [r613] : Require (and use) "COCOA-DEFAULTS". 2004-03-03 22:00 gb * [r612] : Wait until initial thread does CLOSE during listener thread cleanup. 2004-03-03 22:00 gb * [r611] : CLOSE message: close document before invoking SUPER method, not after. Insertion/deletion fixes. 2004-03-03 21:56 gb * [r610] : New file: maintain "defaults" via NSUserDefaults. 2004-03-03 21:53 gb * [r609] : fix %MAKE-DISPLACED-ARRAY call 2004-03-03 02:39 gb * [r608] : Fix in destructuring-bind code. 2004-03-02 00:52 gb * [r607] : Fix typos, start to introduce package-changed hooks. 2004-03-01 18:04 gb * [r606] : Declare the right function in new_tcr(). 2004-03-01 17:08 gb * [r605] : Use AREA_LOCK when accessing all_areas, unless we're early in bootstrapping. Change the API, such as it is. Need to ensure that readers/GC obey locking, as well. 2004-03-01 17:06 gb * [r604] : APPLICATION-UI-OBJECT stuff. 2004-03-01 17:06 gb * [r603] : Notify "application-ui" on package change. 2004-03-01 17:05 gb * [r602] : Don't terminate LISP-THREADs. 2004-03-01 17:05 gb * [r601] : DEFAULT-INITARGS: function follows form ... 2004-03-01 17:04 gb * [r600] : Use application-ui stuff to note initial package. 2004-03-01 17:03 gb * [r599] : Invert order of initform/initfunction (FORM follows FUNCTION ? Nah ..) in canonicalized direct-default-initargs. 2004-03-01 17:02 gb * [r598] : Fix SIZE-VECT in bignum printer. 2004-03-01 17:01 gb * [r597] : More changes. Every day, more changes. (Listeners kind of work now.) 2004-02-29 14:56 gb * [r596] : Some listener-mode bindings. 2004-02-29 01:40 gb * [r595] : Derived from lispbuf.lisp; barely differs yet. Start to bring this code into the 90s. 2004-02-29 01:39 gb * [r594] : Some commented-out Listener-mode stuff. 2004-02-29 01:38 gb * [r593] : CR-terminated files have :EXTERNAL-FORMAT :MACOS, not :MAC. 2004-02-29 01:36 gb * [r592] : Various changes. Listeners still aren't functional. 2004-02-29 01:36 gb * [r591] : "listener" replaces "lispbuf". 2004-02-29 01:35 gb * [r590] : Bryan O'Connor's fix to COMP-MAKE-DISPLACED-ARRAY (make displacement explicit.) 2004-02-28 23:59 gb * [r589] : Fix typo, add more cases in CACHEABLE-CTYPE-P. 2004-02-28 09:24 gb * [r588] : Merge into cocoa-editor.lisp, remove this file. 2004-02-28 09:23 gb * [r587] : More changes for Cocoa Hemlock. 2004-02-28 09:22 gb * [r586] : Don't cache things involving certain MEMBER-CTYPEs. 2004-02-28 09:21 gb * [r585] : stream ELEMENT-SHIFT fixes (from Bryan O'Connor.) 2004-02-28 09:20 gb * [r584] : INVOKE-RESTART recognizes simple restarts (from Bryan O'Connor.) 2004-02-28 09:19 gb * [r583] : MAKE-ARRAY (and variants) track explicit array displacement (from Bryan O'Connor). 2004-02-28 09:16 gb * [r582] : ARRAY-DISPLACEMENT checks for explicit displacement (from Bryan O'Connor.) 2004-02-28 09:14 gb * [r581] : Require NUMBER-MACROS at compile-time. 2004-02-28 09:14 gb * [r580] : 0.14.2-040228. 2004-02-28 09:13 gb * [r579] : ADJUST-ARRAY fixes (from Bryan O'Connor). 2004-02-28 09:12 gb * [r578] : WITH-SIMPLE-RESTART change (from Bryan O'Connor). 2004-02-28 09:11 gb * [r577] : Define $arh_exp_disp_bit (from Bryan O'Connor). 2004-02-28 02:24 gb * [r576] : Try to persuade system to use openmcl icon for lisp documents. 2004-02-27 16:32 beer * [r575] : ObjC generic functions and methods for predefined ObjC messages mostly seem to work; support for conversion between ObjC records and Lisp structs (mostly from Gary) 2004-02-27 16:27 beer * [r574] : *TYPE-SIGNATURE-TABLE* now contains message descriptors; some new automatic type coercions for message arguments 2004-02-27 16:25 beer * [r573] : UPDATE-TYPE-SIGNATURES-FOR-METHOD now takes a class as a second argument 2004-02-27 16:23 beer * [r572] : Yet a few more new ObjC special words 2004-02-26 19:59 gb * [r571] : (Hopefully) return NULL if semaphore_create() fails. 2004-02-26 19:58 gb * [r570] : Check return value from semaphore creation. 2004-02-26 19:58 gb * [r569] : QUIT typechecks arg before it's too late to do so. 2004-02-24 20:31 gb * [r568] : Start to use Hemlock for documents. 2004-02-24 20:30 gb * [r567] : Use some other obscure way of initializing class before looking for methods (Panther). 2004-02-24 01:41 gb * [r566] : more changes. 2004-02-24 01:40 gb * [r565] : method lookup: ensure class is initialized, via fairly obscure means. 2004-02-24 01:40 gb * [r564] : Add COCOA-HEMLOCK. 2004-02-24 01:39 gb * [r563] : require HEMLOCK-TEXTSTORAGE. (Blows up during load ...). 2004-02-23 17:20 gb * [r562] : Lots-o-changes. 2004-02-23 17:18 gb * [r561] : Reorganize; use FASL-CONCATENATE. 2004-02-23 17:18 gb * [r560] : Bugs in PPC2-%SET-BIT. 2004-02-22 01:41 gb * [r559] : Learn to spell. Learn what package we're in. 2004-02-22 01:40 gb * [r558] : Learn to spell. 2004-02-22 01:27 gb * [r557] : PPC64 changes (some of them rather suspect ...). 32-bit kernel may be a little funky ... 2004-02-22 01:21 gb * [r556] : Lots of changes. Still needs event-translation work (and lots of other things.) 2004-02-22 01:20 gb * [r555] : Fewer files. More warnings ... 2004-02-22 01:19 gb * [r554] : Maybe update the selection if the mark we move is the buffer's point. 2004-02-22 01:16 gb * [r553] : Process single events. 2004-02-22 01:16 gb * [r552] : Cocoa, not an RT. (Still needs some work.) 2004-02-22 01:15 gb * [r551] : Buffers maybe have text-storage, but no windows list. 2004-02-22 01:15 gb * [r550] : Move some modeline stuff here from window.lisp. 2004-02-22 01:14 gb * [r549] : Handle INITFUNCTION/INITARG case in SHARED-INITIALIZE. 2004-02-18 18:51 beer * [r548] : More (and different!) support for ObjC generic functions and methods 2004-02-16 10:41 gb * [r547] : ACOS of DOUBLE-FLOAT wants to (%MAKE-DFLOAT) the result. Again. Still. 2004-02-15 16:25 gb * [r546] : Foreign threads linger, too. 2004-02-15 16:23 gb * [r545] : Remove DBG trap in NS-EXCEPTION method. 2004-02-15 16:23 gb * [r544] : Metaclass of user-defined class derived from class-name symbol, not foreign string. 2004-02-15 16:22 gb * [r543] : ALLOCATE-INSTANCE et al: install the instance's slot-vector if it has one. Error out (don't just return nil) if instance slot-vector is missing. Randall's fix to WRITER-METHOD-CLASS. 2004-02-15 16:20 gb * [r542] : Fix PROVIDE call. 2004-02-15 16:19 gb * [r541] : Change PROCESS-INTERRUPT mechanism, since postEvent:atStart: doesn't seem to always do so from subthreads. 2004-02-15 16:17 gb * [r540] : SINGLE-FLOAT, in *TYPE-TO-TYPECODE*. 2004-02-15 16:16 gb * [r539] : WITH-OUTPUT-TO-STRING: don't quote element-type. 2004-02-15 16:14 gb * [r538] : OPCODE is in the ARCH package. 2004-02-15 16:10 gb * [r537] : No more quit-acknowledgement. 2004-02-15 16:10 gb * [r536] : Date suffix for bleeding-edge version. 2004-02-15 16:10 gb * [r535] : No more "quit acknowledgement". 2004-02-15 16:08 gb * [r534] : pick up acos/asin fixes 2004-02-13 18:18 beer * [r533] : Some initial steps toward integrating ObjC methods into CLOS generic functions 2004-02-13 18:17 beer * [r532] : Added a few more special ObjC words I've come across Removed old generic function name-mangling code, since we're no going that route 2004-02-12 07:22 gb * [r531] : ACOS: double-float case uses %MAKE-DFLOAT to allocate result. 2004-02-12 03:19 gb * [r530] : Start to conditionalize for PPC64, almost. 2004-02-12 03:17 gb * [r529] : Fix a lot of the cases broken by removal of extract_2_lisptags_. 2004-02-12 03:16 gb * [r528] : More portability stuff; still a ways to go. 2004-02-12 03:15 gb * [r527] : Flesh out a bit. 2004-02-12 03:15 gb * [r526] : Don't suppress PPC64 defs of eabi stack frames. 2004-02-12 01:23 gb * [r525] : Hoist register names, etc into main file. 2004-02-11 19:01 gb * [r524] : Copy of regular Darwin Makefile, for starters. 2004-02-11 19:00 gb * [r523] : Beg the assembler to be AltiVec-aware. Probably a better way. 2004-02-11 19:00 gb * [r522] : Partly convert to 64-bit-maybeness. 2004-02-11 18:59 gb * [r521] : Parameterize more for ppc32/ppc64. 2004-02-11 18:59 gb * [r520] : 32-bit platform constants. 2004-02-11 18:58 gb * [r519] : Include platform-specific constants file. 2004-02-11 18:58 gb * [r518] : Assume that the assembler is AltiVec aware in vector save/restore. (Do we need vector save/restore ?) 2004-02-11 08:54 gb * [r517] : add asdf, defsystem 2004-02-10 19:06 gb * [r516] : VARIABLE-INFORMATION returns :SPECIAL if visible binding is proclaimed (but not declared) SPECIAL. 2004-02-10 18:39 gb * [r515] : 0.14.1 2004-02-10 17:58 gb * [r514] : was never used 2004-02-09 16:26 gb * [r513] : ABORT-BREAK restart around process's initial function. 2004-02-09 16:22 gb * [r512] : Aargh. 2004-02-09 16:22 gb * [r511] : WITH-STANDARD-ABORT-HANDLING around the idle loop. 2004-02-09 08:34 gb * [r510] : egc back on 2004-02-09 08:33 gb * [r509] : fix error message on REQUIRE failure 2004-02-09 02:11 gb * [r508] : %%SCALE-DFLOAT reserved 16 bytes on tsp, but used 24. 2004-02-08 06:08 gb * [r507] : Add "ccl:tools;" to *MODULE-SEARCH-PATH*. 2004-02-08 06:08 gb * [r506] : 040208, Alpha->Beta. 2004-02-08 05:58 gb * [r505] : New file. 2004-02-08 05:01 gb * [r504] : imported from SourceForge clocc project HEAD as of 040207 2004-02-08 05:01 gb * [r503] : imported from SourceForge cclan project HEAD as of 040207 2004-02-08 01:06 gb * [r502] : Don't include "ccl:" or "home:" on *module-search-path*. Do include "ccl:openmcl;modules;". 2004-02-08 01:01 gb * [r501] : *MODULE-FILE-ALIST* is gone. Honest. 2004-02-08 01:00 gb * [r500] : Add a doc string for *MODULE-PROVIDER-FUNCTIONS*; reference it in the error message if REQUIRE fails. 2004-02-08 00:11 gb * [r499] : Incorporate Bryan O'Connor's changes to REQUIRE (with some changes.) 2004-02-08 00:10 gb * [r498] : Remove *MODULE-FILE-ALIST*. Add *MODULE-PROVIDER-FUNCTIONS*. 2004-02-08 00:09 gb * [r497] : Don't reference *MODULE-FILE-ALIST* anymore. 2004-02-07 22:20 gb * [r496] : Define CCL::CONDITIONAL-STORE, for use by CLX (and maybe other things, if it learns how to decode more "places".) 2004-02-07 22:19 gb * [r495] : Add STORE-GVECTOR-CONDITIONAL, to hide some represenatation details. 2004-02-07 21:59 gb * [r494] : %GET-DOUBLE-FLOAT and %GET-SINGLE-FLOAT: offset defaults to 0. 2004-02-07 21:58 gb * [r493] : bump. 2004-02-07 21:58 gb * [r492] : %GET-DOUBLE-FLOAT/%SET-DOUBLE-FLOAT were shadowed by later versions. 2004-02-07 21:57 gb * [r491] : Initial listener doesn't claim to be *INTERACTIVE-ABORT-PROCESS*. 2004-02-07 21:56 gb * [r490] : Don't call CHOOSE-FILE-DIALOG when looking for a module. 2004-02-07 21:55 gb * [r489] : New scheme to select process for SIGINT. 2004-02-07 21:54 gb * [r488] : Revive objc-classes/methods when restarting an image. Check appkit/foundation versions, print a message & exit if mismatch. Find cfstring-containing library sections after loading appkit/foundation, and again on image startup. 2004-02-07 21:50 gb * [r487] : Change comments. 2004-02-06 19:43 gb * [r486] : REVIVIE-OBJC-CLASSES (mostly working, but needs to happen at the right time and needs to handle version mismatches.) 2004-02-06 19:42 gb * [r485] : Don't call REMAP-OBJC-CLASSES on saved-application startup. 2004-02-06 19:41 gb * [r484] : No more REMAP-ALL-LIBRARY-CLASSES. 2004-02-06 19:40 gb * [r483] : Replace *TERMINAL-INPUT-LOCK* with the new shared-resource scheme. 2004-02-06 19:39 gb * [r482] : MAKE-MCL-LISTENER-PROCESS: spawned process becomes "primary owner" of input stream, if it has an attached SHARED-RESOURCE. 2004-02-06 19:38 gb * [r481] : Probably better to have PROCESS-WAIT sleep explicitly, so that it doesn't spend all of its time yielding. 2004-02-06 19:36 gb * [r480] : Tweaks to the :Y command, to handle new shared-resource scheme. Marco Baringer's :SET breakloop command, from almost a year ago. 2004-02-06 19:35 gb * [r479] : TOPLEVEL-FUNCTION manages the "shared resource" attached to the shared terminal-input stream. 2004-02-06 19:33 gb * [r478] : INPUT-STREAMs may have a SHARED-RESOURCE attached to them; use it to manage terminal-input requests. 2004-02-06 19:32 gb * [r477] : Slot type predicates use CTYPEP, to (kind of) deal with (FUNCTION ...) types. 2004-02-06 19:30 gb * [r476] : Load DLL-NODE a little earlier. 2004-02-06 19:30 gb * [r475] : SHARED-RESOURCEs, mostly for arbitrating access to CCL::*TERMINAL-INPUT*. (May not work well, but neither did the old scheme.) 2004-02-06 19:28 gb * [r474] : 040206. 2004-02-06 19:27 gb * [r473] : Do handshaking with initial thread before shutting down for SAVE-APPLICATION. 2004-02-06 19:27 gb * [r472] : WITH-TERMINAL-INPUT changed slightly. 2004-02-06 19:26 gb * [r471] : Keep the tcr around as long as possible (count down from PTHREAD_DESTRUCTOR_ITERATIONS). Handle the case when a pthread can no longer reveive signals but the underlying Mach thread still exists, in suspend_tcr/resume_tcr. 2004-02-06 19:24 gb * [r470] : Provide mach_suspend_tcr(), mach_resume_tcr() so that threads can be suspended between pthread_exit() and their actual termination. 2004-02-06 18:57 gb * [r469] : Add shutdown_count field to tcr. 2004-02-02 16:22 gb * [r468] : New OBJC-[CLASS,METACLASS,INSTANCE]-p, OBJC-OBJECT-P uses them. Install a callback to take over _dealloc responsibilities. Doing so seems to sometimes trigger a hang on QUIT, where the initial thread never gets an interrupt. Maybe memory corruption, maybe something else: need to check that (free) is the right way to free the object. 2004-02-02 16:15 gb * [r467] : Bryan O'Connor's change to INVOKE-RESTART: throw NIL if simple-restart/null fn. 2004-02-02 16:14 gb * [r466] : FUNCTION-IS-CURRENT-DEFINITION? knows about these new-fangled SETF function names. 2004-02-02 16:12 gb * [r465] : Disable interrupts when accessing the termination queue; may need to inhibit GC as well. 2004-02-02 16:11 gb * [r464] : REQUIRE-GLOBAL-SYMBOL checks the environment, too. 2004-02-02 16:11 gb * [r463] : Handshaking during QUIT/PREPARE-TO-QUIT; this is intended to work around a bug where #_pthread_kill seems to not deliver the signal to the target thread if the signaling thread (a) exits and (b) runs certain Cocoa tsd-cleanup routines. (I.e., it's a mystery.) 2004-02-02 16:08 gb * [r462] : Bryan O'Connor's change to FILE-LENGTH. Default methods on some things that are supposed to TYPE-ERROR, not get a NO-APPLICABLE-METHOD. 2004-02-02 16:07 gb * [r461] : 040202. 2004-02-02 16:06 gb * [r460] : export %STR-FROM-PTR. Don't export %STR-FROM-PTR-IN-SCRIPT, which doesn't exist. 2004-02-02 16:06 gb * [r459] : Bryan O'Connor's fix to WITH-SIMPLE-RESTART. LET-GLOBALLY moved here. 2004-02-02 16:01 gb * [r458] : Remove old vector-copying stuff. 2004-02-01 19:41 beer * [r457] : PRINT-OBJECT shouldn't die when given an OBJC-OBJECT whose ObjC part has "gone away" 2004-02-01 19:38 beer * [r456] : Use DEFCLASS and MAKE-INSTANCE 2004-02-01 19:37 beer * [r455] : Move toward the use of FIND-CLASS and CLASS-OF 2004-02-01 18:09 gb * [r454] : Experimental change, to defer tcr cleanup. 2004-02-01 16:50 beer * [r453] : REMOVE-SLOT-INITARGS was dropping the last KEY/VALUE pair 2004-02-01 15:42 gb * [r452] : Ivar parsing follows different ruls for STRUCT than for *STRUCT; #\B is for :. Now seems able to parse all Foundation/AppKit classes in 10.3.2. 2004-02-01 13:18 gb * [r451] : NS-LISP-EXCEPTION has a lisp slot. Now, to figure out a way to get rid of that ... 2004-02-01 13:17 gb * [r450] : More than one class can share the same metaclass. 2004-02-01 03:05 gb * [r449] : Remove stale comment. 2004-02-01 03:04 gb * [r448] : Fix missing close-paren. 2004-02-01 02:55 gb * [r447] : Unparse :VOID, use keywords more consistently when unparsing. 2004-02-01 02:54 gb * [r446] : Fix DEF-ACCESSORS for VALUES-CTYPE. 2004-02-01 02:53 gb * [r445] : PARSE-DEFGENERIC fix from Bryan O'Connor. 2004-01-31 04:34 gb * [r444] : PARSE-FOREIGN-STRUCT-OR-UNION-SPEC allows for anonymous fields. 2004-01-30 23:58 gb * [r443] : Don't enable EGC by default (so I can work on tuning it some ...) 2004-01-30 23:57 gb * [r442] : Handle BROADCAST-STREAMs, error out on other non-FILE-STREAMs. 2004-01-30 23:56 gb * [r441] : Lots of nit-picking; BROADCAST-STREAM methods (at least partly) fixed to return value from last stream in list. 2004-01-30 23:54 gb * [r440] : Fix CLOSE call to LOCAL-SOCKET-FILENAME. 2004-01-30 23:53 gb * [r439] : WHITESPACEP looks in readtable. 2004-01-30 23:52 gb * [r438] : FRESH-LINE uses STREAM-FRESH-LINE. 2004-01-30 23:51 gb * [r437] : IF-EXISTS: dialog nonsense. PROBE-FILE checks WILD-PATHNAME-P, which moved here. 2004-01-30 23:49 gb * [r436] : WILD-PATHNAME-P now earlier. 2004-01-30 23:49 gb * [r435] : Bump fasl version. 2004-01-30 20:09 gb * [r434] : Lots of changes: foreign slot definitions have a :FOREIGN-TYPE; initarg is used to recognize them. Foreign direct slot offset is (slightly encoded, for bitfields) relative slot offset, can be set by parsing ivars of predefined classes or from :FOREIGN-TYPE when setting (CLASS-DIRECT-SLOTS). Effective-slot location set from class's own direct slot or from superclass direct slot. The foreign "init.." method is called as part of ALLOCATE-INSTANCE, not (again) in SHARED-INITIALIZE. ENSURE-CLASS-USING-CLASS tries to do what we mean, not what we say. Hook into the guts of NSObjectDeallocate to note when an instance is deallocated (and remove its lisp slots-vector); still needs work. Note that saving images with foreign classes in them still doesn't work. 2004-01-30 19:57 gb * [r433] : Lots of changes: get rid of old DEF-OBJC-CLASS support, initialize predefined classes using CLOS (mostly). 2004-01-30 19:55 gb * [r432] : Install predefined classes here. Use DEFCLASS to create NS-LISP-EXCEPTION. 2004-01-30 19:51 gb * [r431] : MAP-OBJC-CLASS -> INSTALL-FOREIGN-OBJC-CLASS. 2004-01-30 19:49 gb * [r430] : Used DEFCLASS to creat classes, SLOT-VALUE or an accessor to access slots. 2004-01-30 19:48 gb * [r429] : Parse instance-var type strings differently from arg type strings. 2004-01-30 19:46 gb * [r428] : Use CMPD vice CMPW. (1 down, a few hundred to go.) 2004-01-30 19:45 gb * [r427] : 040130 2004-01-30 19:44 gb * [r426] : %CSTRLEN. 2004-01-30 19:44 gb * [r425] : CLOSE just does CALL-NEXT-METHOD, which might work now. 2004-01-30 19:43 gb * [r424] : CLOSE: set the closed slot to ... something non-NULL. STREAM-IO-ERROR: so things like sockets can specialize it. Need to test ELEMENT-SHIFT stuff; not sure if I merged Bryan's patch correctly. 2004-01-30 19:36 gb * [r423] : SOCKET-ERROR is (simple) STREAM-ERROR. Change how it's signaled, and define STREAM-IO-ERROR for SOCKET. 2004-01-30 19:34 gb * [r422] : Lock on *TERMINATION-POPULATION* access. 2004-01-30 19:34 gb * [r421] : ADD-READER/WRITER-METHOD: less specialized. Turn a few early accessors into GFs; writer methods on some slots. 2004-01-30 19:32 gb * [r420] : CLASS-SLOT-INITARGS, CREATE-READER/WRITER-METHOD-FUNCTION: less specialized. 2004-01-30 19:30 gb * [r419] : WITH-STRING-FROM-CSTRING: stack allocate a lisp-string, getting bytes from c-string. 2004-01-30 19:28 gb * [r418] : BIT/BITFIELD foreign integer types. 2004-01-29 17:45 gb * [r417] : CTYPE-SUBTYPE: no specialized (COMPLEX ) arrays, so check NUMERIC-CTYPE-COMPLEXP first. 2004-01-29 00:06 gb * [r416] : kinder, gentler do_tsp_overflow 2004-01-27 01:18 gb * [r415] : SUBSEQ: Array-headers aren't sequences; vector headers are. 2004-01-26 16:51 gb * [r414] : (SETF %GET-BITFIELD) becomes %SET-BITFIELD, which is what the setf inverse thought it was anyway. 2004-01-26 16:50 gb * [r413] : %DECODE-FOREIGN-NAME handles 0-length names correctly. 2004-01-26 12:09 gb * [r412] : Kinder, gentler %SHARED-INITIALIZE uses INSTANCE-CLASS-WRAPPER. 2004-01-26 12:08 gb * [r411] : INSTANCE-CLASS-WRAPPER (how many versions of this are there ?) looks at pointer-based classes. 2004-01-26 12:06 gb * [r410] : Bryan O'Connor's fix to %ACTIVE-RESTART. 2004-01-26 12:05 gb * [r409] : So much for the 040124 idea ... 2004-01-26 12:04 gb * [r408] : Initialize ordinal->type vector elements to NIL. 2004-01-25 22:16 gb * [r407] : Linux "as" doesn't like semicolon comments, so use /* */. 2004-01-25 22:16 gb * [r406] : GET-INTERNAL-xxx-TIME moved here. 2004-01-25 22:16 gb * [r405] : WITH-STACK-SHORT-FLOATS is PPC32-specific. 2004-01-25 22:15 gb * [r404] : Start to make RLET assert foreign-type ordinals, then give up on the idea. (FOREIGN-TYPES would need to work earlier in the cold load than they do.) 2004-01-25 22:14 gb * [r403] : Try to introduce "foreign type ordinals", which can be embedded in MACPTRs (this is hard to bootstrap for many reasons.) (PREF foo :thing.embedded-record) expands into some SETFable noise around %INC-PTR. 2004-01-25 22:12 gb * [r402] : Bump. 2004-01-25 22:11 gb * [r401] : MAKE-ARRAY earlier. 2004-01-25 22:11 gb * [r400] : Move some time stuff to time.lisp. 2004-01-25 22:11 gb * [r399] : %COMPOSITE-POINTER-REF, so embedded foreign structs can be SETF-able. 2004-01-25 22:10 gb * [r398] : Don't REQUIRE at loadtime, so we can maybe move earlier in the cold load. 2004-01-25 22:10 gb * [r397] : MACPTRs may say what they're pointing to, though this idea is hard to bootstrap. 2004-01-25 22:09 gb * [r396] : Move LOGICAL-PATHNAME-P earlier. 2004-01-25 22:08 gb * [r395] : Say "target:" a few times. 2004-01-25 22:07 gb * [r394] : Fix %SVAR-BINDING-ADDRESS for the nth time. 2004-01-25 22:07 gb * [r393] : Some conditionalization, mostly dealing with immediate/non-immediate SHORT-FLOATs. 2004-01-25 22:06 gb * [r392] : Conditionalize use of WITH-STACK-SHORT-FLOATS, sortof. 2004-01-25 22:01 gb * [r391] : No low (half-) word of lisp-fpscr. 2004-01-25 22:00 gb * [r390] : WITH-STACK-SHORT-FLOATS moved here. 2004-01-25 22:00 gb * [r389] : Halfhearted attempt at #+ppc32/#+ppc64 conditionalization. Will need to split this in half. 2004-01-25 21:59 gb * [r388] : %composite-pointer-ref. 2004-01-24 23:45 gb * [r387] : -some- PPC64 backend stuff. 2004-01-24 23:43 gb * [r386] : Vector case of SUBSEQ had better be faster ... 2004-01-24 23:43 gb * [r385] : Start to associate "ordinal" numbers with types (for typed pointers.) 2004-01-24 23:41 gb * [r384] : Return was-open-p from CLOSE method. 2004-01-24 23:38 gb * [r383] : UPDATE-SLOTS: assume that slotds are ordered, don't order them by location. 2004-01-24 23:37 gb * [r382] : %COPY-GVECTOR-TO-GVECTOR. 2004-01-24 23:36 gb * [r381] : Array element type of strings is CHARACTER, not that it matters whether it's that or BASE-CHAR. 2004-01-24 23:34 gb * [r380] : Work harder to determine if an instance really is an instance. 2004-01-24 23:34 gb * [r379] : Lots of changes from Randall Beer. 2004-01-24 23:33 gb * [r378] : Preliminary NSBox stuff. Not much of it, even. 2004-01-24 23:32 gb * [r377] : Some of these actually work ... 2004-01-24 23:31 gb * [r376] : target-lisp-node-size is 8. Don't require vinsns (break circle). 2004-01-24 23:30 gb * [r375] : Some fixes; call it fulltag-misc after all. 2004-01-24 23:30 gb * [r374] : Rename some things to (start to) match ppc64. 2004-01-24 23:28 gb * [r373] : target-lisp-node-size in backend. Make default if :ppc32-target, not :ppc-target. 2004-01-24 23:27 gb * [r372] : Define target-most-[positive,negative]-fixnum 2004-01-24 23:27 gb * [r371] : Don't use LOAD-TIME-VALUE to reference vinsn-template cell; look up the template at runtime. Change some vinsn usage. 2004-01-24 23:25 gb * [r370] : Fix 64-bit rotates, define clrrdi instead of defining clrrdi. twice. 2004-01-24 23:23 gb * [r369] : Yawn. Update spec class alist for new storage-class names. 2004-01-24 23:22 gb * [r368] : Match against :u64/:s64 storage class names. 2004-01-24 23:22 gb * [r367] : New constants for 64-bit storage classes. 2004-01-24 23:21 gb * [r366] : New GPR mode-names for 64-bit storage classes. 2004-01-24 23:20 gb * [r365] : 64-bit storage classes. 2004-01-22 10:53 gb * [r364] : Um, COPY-FPR sets (and uses) an FPR ... 2004-01-22 10:52 gb * [r363] : MAKE-VINSN zeros the GPRS-SET & FPRS-SET fields when pulling a vinsn off of the freelist. 2004-01-22 10:51 gb * [r362] : Try to un-botch SPbuiltin_length, for about the third time in a week. 2004-01-22 06:11 gb * [r361] : New --version command-line argument; new APPLICATION-VERSION-STRING method. 2004-01-22 06:10 gb * [r360] : Bump, again. 2004-01-22 06:10 gb * [r359] : Correct TARGET-LEVEL-1-MODULES, to reflect new target names. 2004-01-21 08:32 gb * [r358] : Default methods (specialized on T) of OPEN-STREAM-P, INTERACTIVE-STREAM-P, INPUT-STREAM-P, OUTPUT-STREAM-P complain that argument isn't a stream. This is OK, as long as users only specialize these methods on subclasses of FUNDAMENTAL-STREAM, and they shouldn't be doing otherwise anyway. 2004-01-21 07:37 gb * [r357] : Bryan O'Connor's fix to STREAM-UNREAD-CHAR method for ECHO-STREAM. 2004-01-21 07:30 gb * [r356] : NX1-ASH: where did those numbers come from ? 2004-01-21 07:29 gb * [r355] : (SIGNED-BYTE 1) is perfectly legal ... 2004-01-20 23:54 gb * [r354] : Define FIXUP-PPC-BACKEND in terms of target. 2004-01-20 23:54 gb * [r353] : :PPC -> :PPC32. 2004-01-20 23:43 gb * [r352] : Backend name changes. 2004-01-20 23:43 gb * [r351] : More PPC/PPC32 changes. (Not generalized yet ...) 2004-01-20 23:41 gb * [r350] : Bump again. 2004-01-20 23:41 gb * [r349] : Bryan O'Connor's fix to PROBE-FILE. 2004-01-20 23:40 gb * [r348] : Don't refer to *PPC-VINSN-TEMPLATES*; use the target backend instead. 2004-01-20 23:39 gb * [r347] : New backend names. 2004-01-20 23:38 gb * [r346] : backend changes 2004-01-20 23:37 gb * [r345] : new file 2004-01-20 23:17 gb * [r344] : meet the new files, same as the old files 2004-01-20 23:16 gb * [r343] : new file, sorta 2004-01-20 23:15 gb * [r342] : removed, now PPC32-specific 2004-01-20 23:15 gb * [r341] : backend changes 2004-01-20 00:41 gb * [r340] : Bryan O'Connor's fix to force RETURNing clauses into the main body. 2004-01-20 00:40 gb * [r339] : Allow (EVAL-WHEN () ,@body). 2004-01-20 00:40 gb * [r338] : PARSE-DEFGENERIC allows multiple DECLARE forms, passes them as :DECLARATIONS option. HANDLER-BIND a little less jumpy. 2004-01-20 00:37 gb * [r337] : (Finally) try to handle odd poweropen struct alignment rules. 2004-01-20 00:37 gb * [r336] : (Finally) try to handle odd poweropen struct alignment rules. Pad :DBM_CONSTANT to naturally align double-float field, since unnatural alignment is now the norm. 2004-01-20 00:35 gb * [r335] : SIGNAL keeps looking in the same cluster when a handler declines. 2004-01-20 00:34 gb * [r334] : Allowing INVOKE-DEBUGGER to build a condition from extra arguments would clearly lead to increased street crime. 2004-01-20 00:33 gb * [r333] : Define (and signal) some new condition classes. 2004-01-20 00:30 gb * [r332] : Map kernel "can't construct arglist" to a (new) TYPE-ERROR subclass. 2004-01-20 00:29 gb * [r331] : GF's need to check keywords after all. Recognize :ALLOW-OTHER-KEYS T sooner, don't treat :ALLOW-OTHER-KEYS NIL as an unknown argument. 2004-01-20 00:27 gb * [r330] : Paranoia in CTYPE-SUBTYPE: allow float aliases. 2004-01-19 04:12 gb * [r329] : Don't reference *objc-readtable*. 2004-01-18 04:00 gb * [r328] : From Bryan O'Connor: "LOOP FOR x ..." uses destructuring mechanism on X, so "LOOP FOR NIL ..." is accepted. 2004-01-18 03:58 gb * [r327] : More pathname info for DESCRIBE. Use INSTANCE-SLOTS vice INSTANCE.SLOTS, so .accessors will work on other instance types. 2004-01-18 03:57 gb * [r326] : PPRINT-DISATCH: symbol is external in CL, so don't use it as a structure name. 2004-01-18 03:55 gb * [r325] : RATIONALIZE comments; extra typechecking in BOOLE. 2004-01-18 03:55 gb * [r324] : WITH-HASH-TABLE-ITERATOR allows declarations. 2004-01-18 03:54 gb * [r323] : Handle FOREIGN-ARRAY-TYPE. (Note: this is Darwin-specific.) 2004-01-18 03:53 gb * [r322] : SBIT, %SBITSET hande 0-dimensional bitvector case. 2004-01-18 03:52 gb * [r321] : Bump. 2004-01-18 03:52 gb * [r320] : TYPE-OF changes. Blame CLHS. 2004-01-18 03:51 gb * [r319] : More PROGRAM-ERRORs signaled during destructuring. 2004-01-18 03:50 gb * [r318] : Lots more changes. 2004-01-18 03:49 gb * [r317] : MAKE-STRING-OUTPUT-STREAM: no such thing as a STRING-CHAR anymore. 2004-01-18 03:48 gb * [r316] : WARN has to be more anal-retentive about the condition and any args it gets. 2004-01-18 03:47 gb * [r315] : Handle bytespecs with size 0. 2004-01-18 03:46 gb * [r314] : Quote dots in namestring components. Catch a few bogus cases, like (:ABSOLUTE :UP) in directory. 2004-01-18 03:44 gb * [r313] : %%NTH-ARG-DCODE: signal program error if 0 args. 2004-01-18 03:40 gb * [r312] : Use safer accessors. (May revert back some cases, since .accessors should now use INSTANCE-SLOTS rather than INSTANCE.SLOTS.) 2004-01-18 03:38 gb * [r311] : Define more accessors; INSTANCE-SLOTS is now used in most "early" .accessors. Don't look at GF-AOK bit; whatever the problem with initarg caching is, that's not the solution. 2004-01-18 03:36 gb * [r310] : NRECONC uses ENDP. %MAKE-DISPLACED-ARRAY always displaces to what the user said to displace to. 2004-01-18 03:35 gb * [r309] : DISPLACED-ARRAY-P didn't handle transitive displacement. 2004-01-18 03:34 gb * [r308] : Handle denorms in INTEGER-DECODE-SHORT-FLOAT. 2004-01-18 03:33 gb * [r307] : (EXPT b e) is (COERCE 1 (TYPE-OF B)) if (ZEROP E). 2004-01-18 03:32 gb * [r306] : *BASE-CHAR-CLASS* may or may not be the same as *CHARACTER-CLASS*. (integer 1 1) denotes a BIT. 2004-01-18 03:30 gb * [r305] : Handle negative bytespecs; don't define synonyms where they may be involved. Open-code INTEGERP. 2004-01-15 08:34 gb * [r304] : Remove a stale comment. Might consider a less-hairy typespec. 2004-01-15 08:33 gb * [r303] : (INTEGER 0 0) is a subtype of (INTEGER 0 1), aka BIT. 2004-01-15 08:32 gb * [r302] : MAKE-INSTANCES-OBSOLETE: methods return class. 2004-01-15 08:29 gb * [r301] : %%CHECK-KEYWORDS is (a) slow (b) wrong (c) probably unnecessary. If (c) is true, there's a similar case where the same things are probably true as well, and callers should be changed. (c) is certainly true in some cases (when the "combined method" is a simple method. 2004-01-15 08:26 gb * [r300] : READER-ERROR inherits from PARSE-ERROR and STREAM-ERROR. 2004-01-15 08:25 gb * [r299] : PATHNAME-DEVICE typechecks its arg before ignoring it. 2004-01-15 08:20 gb * [r298] : Default version of STREAM-FILENAME signals an error; streams that aren't FILE-STREAMs aren't pathname-designators. 2004-01-15 08:18 gb * [r297] : Bump up the (global) type-cache size. Fixes in CLASS type methods. 2004-01-15 08:16 gb * [r296] : A million monkeys and a million typewriters: sooner or later, one of them will get ASH right ... 2004-01-15 03:59 gb * [r295] : Special-case (TRUNCATE MOST-NEGATIVE-FIXNUM x). 2004-01-14 04:50 gb * [r294] : forgot to commit this earlier 2004-01-14 01:13 gb * [r293] : Signal a PROGRAM-ERROR or two. Better checking of SYMBOL-MACROs in AUGMENT-ENVIRONMENT. 2004-01-14 01:12 gb * [r292] : Allow NIL to be a -local- function name. Don't tail-call SIGNAL-PROGRAM-ERROR. 2004-01-14 01:11 gb * [r291] : Add a (long-)missing IN-PACKAGE. 2004-01-14 01:11 gb * [r290] : DISASSEMBLE's output should go to *STANDARD-OUTPUT*. 2004-01-14 01:10 gb * [r289] : FILL-POINTER's lament. 2004-01-14 01:10 gb * [r288] : maybe bignum-bignum-gcd really works now. 2004-01-14 01:09 gb * [r287] : IMAGPART uses * vice FLOAT, to get the sign of 0.0 right. PHASE of a negative rational returns single-float PI. 2004-01-14 01:08 gb * [r286] : ALT-LIST-LENGTH here. LAST allows positive integers. GET-PROPERTIES uses a TYPE-ERROR to complain about bad plists. 2004-01-14 01:06 gb * [r285] : Recognize that SLOT-MISSING might return if it's specialized. Call it with the symbol SETF, not '(SETF SLOT-VALUE). 2004-01-14 01:04 gb * [r284] : Check initargs via DESTRUCTURING-BIND in %SHARED-INITIALIZE. ENSURE-CLASS-FOR-DEFCLASS and *defclass-redefines-improperly-named-classes-pedantically*. (SETF CLASS-NAME) isn't an accessor; the real method uses REINITIALIZE-INSTANCE. 2004-01-14 01:02 gb * [r283] : Bad keywords trigger PROGRAM-ERRORs. 2004-01-14 01:01 gb * [r282] : PARSE-ERROR isn't a STREAM-ERROR. Tail-call ERROR from SIGNAL-PROGRAM-ERROR, to keep it out of backtrace. UNDEFINED-FUNCTION-CALL: lose FUNCTION-NAME slot, use CELL-ERROR-NAME instead. 2004-01-14 00:58 gb * [r281] : Can't have a function called CL:RANDOM-STATE, so change its name and its only caller. 2004-01-14 00:58 gb * [r280] : SYMBOL-MACROLET in the cheap evaluator: process declarations along with symbol-macros, so that AUGMENT-ENVIRONMENT can complain about SPECIAL decls easier. 2004-01-14 00:56 gb * [r279] : Incorporate lots of changes (most of the type system, in fact) from recent CMUCL sources. Thank Cristophe Rhodes and others for their work. Warning to fix an EVAL-WHEN; can't define some stuff at compile-time while bootstrapping from the old type system. 2004-01-14 00:55 gb * [r278] : Fix ASSOC and friends, broken just a few days ago. Signal PROGRAM-ERROR when detecting argument mismatches in macros. 2004-01-14 00:54 gb * [r277] : TYPE-OF's a little closer to being right (checks for PROPERly-named classes.) 2004-01-14 00:53 gb * [r276] : STRUCTURE isn't a type specifier; STRUCTURE-OBJECT is. Use the right one in COPY-STRUCTURE. 2004-01-14 00:52 gb * [r275] : Use ENDP to detect bad alists in COPY-ALIST. Move ALT-LIST-LENGTH elsewhere (and fix it ...) 2004-01-14 00:50 gb * [r274] : DEFCLASS: allow NIL as slot :type, expand into (new) ENSURE-CLASS-FOR-DEFCLASS. Signal a PROGRAM-ERROR in DEFINE-METHOD-COMBINATION. Can't have a macro named CL:METHOD, so change its name to CCL::REFERENCE-METHOD for the time being. 2004-01-14 00:48 gb * [r273] : LCM-2, SIGNUM fixes. 2004-01-14 00:47 gb * [r272] : (SATISFIES (OUTPUT-STREAM-P)). (The thing that's catching these things is a half-bootstrapped type system, but I'd have liked to have seen that one a lot earlier ...) 2004-01-14 00:46 gb * [r271] : Add an IN-PACKAGE. PARSE-INTEGER signals PARSE-ERRORs. 2004-01-14 00:45 gb * [r270] : The type system is/was choking on unknown types; it choked on (NUMBER 0 *), which should have been (REAL 0 *) in the first place. 2004-01-14 00:43 gb * [r269] : New CTYPEs for negation, intersection. 2004-01-14 00:43 gb * [r268] : Restore SPbuiltin_assq (ignore NILs in the alist.) (ash x y) returns -1 on (<= n -29); need to find a way to parameterize this while keeping M4 happy. 2004-01-12 15:20 beer * [r267] : Partial support for MAKE-INSTANCE on ObjC classes 2004-01-12 07:31 gb * [r266] : Too-few args to generic-function -> PROGRAM-ERROR, not just ERROR. 2004-01-12 07:30 gb * [r265] : BUTLAST, NBUTLAST handle dotted lists again. 2004-01-12 07:29 gb * [r264] : Handle the case where &WHOLE is a structured lambda-list in %destructure-lambda-list. 2004-01-12 01:54 gb * [r263] : Start to replace some old (old-binding scheme) entries with &restv support. SPsvar_progvsave: save the old binding, not the new one. 2004-01-12 01:53 gb * [r262] : Start to replace some old (old-binding scheme) entries with &restv support. 2004-01-12 01:53 gb * [r261] : ARRAY-IN-BOUNDS-P allows bignum subscripts (which are never in bounds, of course.) 2004-01-12 01:52 gb * [r260] : ASSEQL was misparenthesized. 2004-01-12 01:51 gb * [r259] : rename some macros: ppc->ppc64. 2004-01-12 01:50 gb * [r258] : NX-TAG-INFO compares tagbody tags with EQL. 2004-01-11 06:24 gb * [r257] : (SETF FIND-CLASS) checks more carefully for "protected classes". 2004-01-11 06:14 gb * [r256] : Vinsn renames; slight change to remove backend support for "handle"-based memory accessors. 2004-01-11 06:13 gb * [r255] : Change some vinsn names, eliminate some unused vinsns. This'll be "ppc32-vinsns.lisp" soon. 2004-01-11 06:08 gb * [r254] : Fix nil-value constant. 2004-01-11 06:08 gb * [r253] : New file; not close to correct yet. 2004-01-11 06:07 gb * [r252] : Make *DEFAULT-TEMP-STACK-SIZE* too big. 2004-01-11 06:06 gb * [r251] : Fix LCM/LCM-2 error checking. 2004-01-11 06:06 gb * [r250] : _SPbuiltin_length handles the LIST-LENGTH case inline. 2004-01-09 22:09 gb * [r249] : Back out of the attempt to make LIST-LENGTH a builtin. 2004-01-09 21:36 gb * [r248] : Another typo in BUTLAST. 2004-01-09 21:29 gb * [r247] : Typo in ASSEQL. 2004-01-09 21:27 gb * [r246] : LIST-LENGTH will soon be builtin. 2004-01-09 21:27 gb * [r245] : [N]BUTLAST fixes. 2004-01-09 21:26 gb * [r244] : BAD-SLOT-TYPE, IMPROPER-LIST conditions. 2004-01-09 21:25 gb * [r243] : Handle IMPROPER-LIST 2004-01-09 21:25 gb * [r242] : BAD-SLOT-TYPE errors. 2004-01-09 21:24 gb * [r241] : LENGTH changes. 2004-01-08 00:30 gb * [r240] : A little bit of conditionalization. 2004-01-08 00:29 gb * [r239] : TYPEP returns 1 value. 2004-01-08 00:29 gb * [r238] : Detect when destructive right shifts yield a fixnum more accurately (this was causing the bignums to be passed to internal functions that expect fixnums in some cases). 2004-01-07 23:40 gb * [r237] : ASSQ is builtin, so just EVAL-REDEF it. 2004-01-07 23:39 gb * [r236] : ASSOC variants don't allow NIL in an alist. ASSEQL typechecks CAR/CDR. MEMEQL uses ENDP, doesn't use %CAR/%CDR in unsafe contexts. 2004-01-07 23:38 gb * [r235] : Do an EVAL-REDEF for ASSQ. 2004-01-07 23:37 gb * [r234] : Fix (at least the function case) of :ALLOW-OTHER-KEYS NIL. 2004-01-07 15:42 beer * [r233] : Added support for foreign slot definitions 2004-01-07 03:20 gb * [r232] : Assume that elaboration will call #'%SLOT-ID-REF-MISSING, #'%SLOT-ID-SET-MISSING as constant functions (not as symbols.) 2004-01-06 05:21 gb * [r231] : Print class addresses when generating slots, in the commented-out FORMAT calls. 2004-01-06 05:19 gb * [r230] : Derive the names of superclasses when recursively mapping classes; don't use the name MAP-OBJC-CLASS was called with/defaulted to over and over. 2004-01-06 03:05 gb * [r228] : Say "4" instead of "(ash 1 ppc::word-shift)", since this is platform-dependent anyway. 2004-01-06 03:04 gb * [r227] : Use "TARGET" package, not "ARCH". 2004-01-06 02:33 gb * [r226] : Preliminary support for statically-linked kernel. 2004-01-06 02:30 gb * [r225] : FIXNUMSHIFT in TARGET package. 2004-01-06 02:30 gb * [r224] : SUBTAG-INSTANCE in TARGET package. 2004-01-06 02:29 gb * [r223] : Intern accessors in "TARGET" package. 2004-01-06 02:29 gb * [r222] : EXTRACT-DEFAULT function's in PPC package now. 2004-01-06 02:12 gb * [r221] : Error codes in ARCH package. 2004-01-06 02:11 gb * [r220] : Nilreg-relative-symbols are PPC common. 2004-01-06 02:11 gb * [r219] : Set lisp_global(STATICALLY_LINKED) if STATIC is defined. 2004-01-06 02:10 gb * [r218] : Don't assume that dladdr exists if STATIC is defined. 2004-01-06 02:09 gb * [r217] : STATICALLY_LINKED global. 2004-01-06 02:09 gb * [r216] : PPC regs are in PPC package. 2004-01-06 02:08 gb * [r215] : Opcode/operand stuff's in ARCH package. 2004-01-06 02:08 gb * [r214] : 040106, for a while. 2004-01-06 02:07 gb * [r213] : Error codes are in ARCH package. 2004-01-06 02:07 gb * [r212] : TCR flags are in ARCH. (Strange, but true.) 2004-01-06 02:06 gb * [r211] : All PPCs have an FPSCR with the same bits in it ... 2004-01-06 02:06 gb * [r210] : Fix (bad) typo in %SHARED-INITIALIZE. 2004-01-06 02:05 gb * [r209] : Some changes to initarg checking/caching; may still not be correct. 2004-01-06 02:04 gb * [r208] : Internal-package dependency changes. 2004-01-06 02:03 gb * [r207] : Errors in ARCH package. 2004-01-06 02:03 gb * [r206] : Some support for static linking under Linux, and changes to the startup code for when *STATICALLY-LINKED* is in effect. 2004-01-06 02:02 gb * [r205] : Should be close to having PPC64 stuff (tags, object layout) here. 2004-01-06 02:02 gb * [r204] : Should be close to having PPC32 stuff (tags, object layout) here. May need to be reorganized to cross-compile cleanly. 2004-01-06 02:01 gb * [r203] : Mostly PPC (not PPC32 or PPC64) specific, nominally. (Undoubtedly still some PPC32 dependencies.) 2004-01-06 01:59 gb * [r202] : Mostly PPC (not PPC32 or PPC64) specific, nominally. 2004-01-06 01:59 gb * [r201] : Mostly PPC (not PPC32 or PPC64) specific, nominally. Opcode/operand lap stuff's in ARCH. 2004-01-06 01:58 gb * [r200] : Most stuff that's truly PPC-specific should be here (in the PPC package) now. 2004-01-06 01:56 gb * [r199] : Most stuff that's truly PPC-specific should be here (in the PPC package) now. 2004-01-06 01:54 gb * [r198] : Storage classes (s64/u64 need to be added, at least) are in the ARCH package. 2004-01-06 01:53 gb * [r197] : (generic) opcode/operand stuff is in the ARCH package. 2004-01-06 01:53 gb * [r196] : Builtin-functions list is in the ARCH package. 2004-01-06 01:52 gb * [r195] : Back in the ARCH package, but there's a lot less stuff here. 2004-01-03 23:36 gb * [r194] : Try again, using WITH-NEGATED-BIGNUM-BUFFERS. 2004-01-03 23:23 gb * [r193] : %BIGNUM-BIGNUM-GCD wasn't zeroing out the buffers it (re-)uses, so use a simpler algorithm (that conses more) until this is fixed. 2004-01-03 19:52 gb * [r192] : Slight change to requirements. 2004-01-03 19:51 gb * [r191] : New file. 2004-01-03 19:51 gb * [r190] : New (empty, in fact) file. 2004-01-03 19:50 gb * [r189] : Try to be less ppc32-specific. 2004-01-03 19:49 gb * [r188] : Try to be more ppc32-specific. 2004-01-03 19:49 gb * [r187] : (Mostly) work with gnu-objc; may have broken apple-objc ... 2004-01-03 19:48 gb * [r186] : special-case GS under GNUstep. 2004-01-03 19:48 gb * [r185] : More gnu-objc conditionalization. Take user-specified class-name when mapping objc-class. Make CLASS-OF objc-metaclasses be OBJC-METACLASS. 2004-01-03 19:45 gb * [r184] : More gnu-objc conditionalization. 2004-01-03 19:43 gb * [r183] : Generalize the notion of packages whose symbols are always exported. 2004-01-03 19:42 gb * [r182] : Move command-line-arg stuff to l1-pathnames. 2004-01-03 19:42 gb * [r181] : Install the kernel error callbacks, late in the cold load. 2004-01-03 19:41 gb * [r180] : Move FOREIGN-OBJECT-DOMAIN accessors elsewhere. Make slot accessors work on pointer-based instances. 2004-01-03 19:40 gb * [r179] : Add a VALIDATE-SUPERCLASS method for (FOREIGN-CLASS STANDARD-CLASS), just in case. 2004-01-03 19:39 gb * [r178] : Move COMMAND-LINE-ARGUMENTS stuff here; use it to find the CCL directory, maybe. 2004-01-03 19:39 gb * [r177] : Rename the %ERR-DISP callback to %X-ERRDISP, so the kernel doesn't try to call out to %ERR-DISP too early in the cold load. 2004-01-03 19:38 gb * [r176] : Rename the CMAIN callback to XCMAIN, so the kernel doesn't try to call out to CMAIN too early in the cold load. 2004-01-03 19:37 gb * [r175] : Add PPC-ARCH. (This is all still a moving target.) 2004-01-03 19:37 gb * [r174] : Integrate a little better with the new debugger. 2004-01-03 19:36 gb * [r173] : According to my calculations, you need 5 bits to express a 5-bit register field (RT, RA, RB, etc.) 2004-01-03 19:35 gb * [r172] : Print MACPTRs, get method names/specializers half-right. 2004-01-03 19:35 gb * [r171] : Move FOREIGN-OBJECT-DOMAIN accessors here. 2004-01-03 10:49 gb * [r170] : Zero the domain/type fields when making a stack-block. 2003-12-31 11:12 gb * [r169] : Use STANDARD-OBJECT-P to find instance class-wrappers. 2003-12-31 11:11 gb * [r168] : Fix an EQL-SPECIALIZER botch. Lots of stuff to support foreign classes/foreign objects. 2003-12-31 11:09 gb * [r167] : Use %class-slots instead of %CLASS.SLOTS when CLASS-SLOTS won't do. Move CONSTANTLY here, remove some outdated comments. 2003-12-31 11:08 gb * [r166] : Careful about getting wrappers in dispatch. Use %class-slots instead of %CLASS.SLOTS when CLASS-SLOTS won't do. 2003-12-31 11:06 gb * [r165] : Write the address involved in a #, as if that helps ... Use CLASS-NAME vice %CLASS.NAME. 2003-12-31 11:04 gb * [r164] : Macptr domain/type accessors. In LAP to simplify treating them as (SIGNED-BYTE 32), but it was easier to make domain unsigned. 2003-12-31 11:03 gb * [r163] : Half-hearted change to %TYPEP, to handle foreign-typed macptrs. 2003-12-31 11:02 gb * [r162] : Don't call MAP-OBJC-CLASS when walking modules. 2003-12-31 11:02 gb * [r161] : Start to integrate CLOS & ObjC. 2003-12-31 11:01 gb * [r160] : New file: start to integrate CLOS/ObjC. 2003-12-31 11:01 gb * [r159] : MACPTR->STACK zeroes domain, type fields. 2003-12-29 08:19 gb * [r158] : SUPER's gensym'ed in methods; FLET SUPER accesses it. Remove the CFBUNDLE stuff, so this file can be loaded when it isn't necessary. 2003-12-29 08:17 gb * [r157] : Access SUPER via (SUPER). 2003-12-29 08:17 gb * [r156] : Put the CFBUNDLE nonsense here, independent of the rest of objc-support. 2003-12-29 08:17 gb * [r155] : Get rid of *DEFAULT-BUNDLE-PATH* and friend; require FAKE-CFBUNDLE-PATH and use the function of that name explicitly. 2003-12-29 08:15 gb * [r154] : Use (SUPER) in OBJC-MESSAGE-SEND-SUPER call. 2003-12-21 22:40 gb * [r153] : revive class info, autorelease-pool when reiniting CCL directory 2003-12-20 12:18 gb * [r152] : Comment out things that don't (yet) work. 2003-12-20 12:17 gb * [r151] : Punt on primary accessor methods. 2003-12-20 12:16 gb * [r150] : Add newline before EOF. 2003-12-20 12:16 gb * [r149] : Fix dangling close paren. 2003-12-20 10:45 gb * [r148] : Alanr's fix to closure-inspector. 2003-12-20 07:48 gb * [r147] : Remove inline loginfo. 2003-12-20 07:48 gb * [r146] : Add newline before EOF. 2003-12-20 00:29 gb * [r145] : Replaced by objc-runtime.lisp 2003-12-19 23:44 gb * [r144] : A little clearer to say OPTION-NAME instead of (CADR OPTION) in DEFCLASS misc-option case. 2003-12-19 23:35 gb * [r143] : Change debugger prompt. Still confuses the hell out of ILISP. 2003-12-19 23:34 gb * [r142] : Keep count of splay-tree nodes; PRINT-OBJECT methods. 2003-12-19 23:33 gb * [r141] : Bump the version number. 2003-12-19 23:33 gb * [r140] : Call MAP-OBJC-CLASS on (true) classes; MAP-OBJC-CLASS will handle the metaclass. 2003-12-19 23:32 gb * [r139] : Add (commented-out) DBG call. 2003-12-19 23:31 gb * [r138] : Some fixes: ensure that the Cocoa demo works under OSX. 2003-12-19 22:09 gb * [r137] : New & improved debugger. 2003-12-19 21:47 gb * [r136] : map-splay-tree 2003-12-19 21:37 gb * [r135] : This is all #+apple-objc. 2003-12-19 21:36 gb * [r134] : Move readtable stuff out of main path; preserve (for a while) for backward- compatibility. 2003-12-19 21:36 gb * [r133] : Require (new) OBJC-RUNTIME. Conditionalize a little for runtime differences (needs more work.) 2003-12-19 21:33 gb * [r132] : Define bundle/executable paths here. 2003-12-19 21:31 gb * [r131] : New file, intended to replace apple-objc.lisp. Try to hide the differences between Apple/GNU ObjC runtimes here. 2003-12-19 14:05 beer * [r130] : Added missing comma to SELF in OBJC-METHOD-DEFINITION-FORM 2003-12-16 18:23 gb * [r129] : use compiler frontend hooks 2003-12-16 18:22 gb * [r128] : integrate XREF changes 2003-12-16 18:13 gb * [r127] : minor changes 2003-12-15 22:16 gb * [r126] : new file 2003-12-15 21:12 gb * [r124] : start to conditionalize for runtime 2003-12-15 21:11 gb * [r123] : get more of this working 2003-12-13 21:53 gb * [r122] : Revive efforts to get this working. 2003-12-13 21:51 gb * [r121] : DIRECTORYP. 2003-12-12 23:32 gb * [r120] : SETUP-SLOT-LOOKUP does nothing if no slots. 2003-12-12 23:30 gb * [r119] : STRUCTURE isn't a type-specifier; use STRUCTURE-OBJECT instead. 2003-12-12 23:29 gb * [r118] : STRUCTURE isn't a type-specifier. 2003-12-12 23:29 gb * [r117] : TYPE-SPECIFIER does some typechecking. 2003-12-12 23:28 gb * [r116] : FAKE-STACK-FRAME isn't a type specifier. 2003-12-12 23:27 gb * [r115] : STANDARD-INSTANCE isn't a type specifier. 2003-12-12 23:26 gb * [r114] : COMPLEX-ARRAY, DISPLACED-ARRAY aren't type specifiers. 2003-12-12 21:12 gb * [r113] : New wrapper layout. 2003-12-12 21:12 gb * [r112] : Export METHOD-NAME from CCL, OPENMCL-MOP. 2003-12-12 21:12 gb * [r111] : Set TYPE-PREDICATE slot of standard slot definition. 2003-12-12 21:10 gb * [r110] : New slot-value scheme. No more reader/writer dcode. 2003-12-12 21:10 gb * [r109] : New slot-value scheme. 2003-12-12 21:09 gb * [r108] : New scheme. 2003-12-10 08:25 gb * [r107] : Don't free() a pointer that was just realloc()ed. 2003-12-10 08:25 gb * [r106] : Don't reference tlb_pointer until we're sure it's big enough. 2003-12-10 08:24 gb * [r105] : More wrapper slots for faster slot lookup. 2003-12-10 08:24 gb * [r104] : New date. 2003-12-10 08:23 gb * [r103] : Set things up for faster slot lookup. 2003-12-10 08:22 gb * [r102] : SLOT-ID stuff. 2003-12-10 08:21 gb * [r101] : Some LAP for slot lookup. 2003-12-10 08:20 gb * [r100] : New PPC32, MOP features. 2003-12-07 05:25 gb * [r99] : Move SLOT-ID-VALUE/SETF-SLOT-ID-VALUE from l1-clos.lisp to l1-clos-boot.lisp. 2003-12-07 05:24 gb * [r98] : There -will- be more fields in a class wrapper soon. 2003-12-07 04:11 gb * [r97] : First cut at SLOT-VALUE. 2003-12-07 01:44 gb * [r96] : PRINT-OBJECT method for SLOT-ID. 2003-12-07 01:44 gb * [r95] : Start to implement new slot-lookup scheme. 2003-12-07 00:25 gb * [r94] : Return value of insertNewline: method is void. 2003-12-07 00:16 gb * [r93] : Be more careful about recognizing EQL specializers. 2003-12-07 00:15 gb * [r92] : COLLECT macro moved to CCL package. 2003-12-01 21:44 gb * [r91] : %foreign-type-or-record: recognize (:struct :foo) off the bat. 2003-12-01 21:44 gb * [r90] : Better %%rusage, since Darwin mucks up system/user time. 2003-12-01 21:42 gb * [r89] : %SVAR-BINDING-ADDRESS fix. 2003-11-30 08:55 gb * [r88] : ensure_gc_structures_writable: account for entire heap, not just current dynamic area. Don't try to commit a 0-length page range. 2003-11-30 08:54 gb * [r87] : Distinguish between read errors (especially EINTR) and EOF; terminate on true EOF. 2003-11-19 08:52 gb * [r85] : remove sparc lap; move all PPC lap code to ccl:level-0;PPC; 2003-11-18 15:32 gb * [r84] : PPC lap macros hash table in backend 2003-11-18 10:10 gb * [r83] : ppc-arch -> ppc32-arch, lap-macros slot in backend 2003-11-18 09:15 gb * [r82] : rename of ppc-arch.lisp 2003-11-18 08:25 gb * [r81] : missed a few DEFENUM stragglers 2003-11-18 07:51 gb * [r80] : move ONCE-ONLY, COLLECT. ITERATE to CCL package 2003-11-18 07:51 gb * [r79] : bump version 2003-11-18 06:59 gb * [r78] : rename arch:: ppc:: refs to ppc32::, for now 2003-11-18 06:52 gb * [r77] : start big package rename scheme 2003-11-18 03:52 gb * [r76] : hasn't been used in a long time ... 2003-11-13 09:19 gb * [r75] : PPC2-LOCAL-GO: need to clean up the vstack regardless of XFER token. 2003-11-10 01:15 gb * [r74] : fix handling of immediate operands in vector instructions 2003-11-08 19:45 gb * [r73] : Clear *INTERACTIVE-ABORT-PROCESS* in KILL-LISP-POINTERS. 2003-11-08 19:44 gb * [r72] : Don't try to interrupt *INTERACTIVE-ABORT-PROCESS* if it's null. 2003-11-08 19:39 gb * [r71] : Bump the version. 2003-11-06 08:57 gb * [r70] : Use symbolic constants in a few places. Check for recent CREATE_LISP_FRAME_INSTRUCTION before pc_luser-ing through stack frame initialization, since it may be re-initialization. 2003-11-06 08:56 gb * [r69] : Define a few opcodes, and define CREATE_LISP_FRAME_INSTRUCTION. 2003-11-04 19:04 gb * [r68] : More structure-class accessor stuff. 2003-11-04 19:04 gb * [r67] : Guess ... 2003-11-04 17:15 gb * [r66] : Some more work to make STRUCTURE-CLASS have real slot definitions. 2003-11-04 17:15 gb * [r65] : Define a MAKE-INSTANCES-OBSOLETE method for STRUCTURE-CLASS. Next step: make it do something, like warning ... 2003-11-04 17:13 gb * [r64] : PRINT-OBJECT method for SLOT-DESCRIPTOR, whether standard or not. 2003-10-30 00:47 gb * [r63] : Fix initialization of early effective-slotds. Start to integrate STRUCTURE-CLASS a bit better. 2003-10-29 23:18 gb * [r62] : Distinguish between inherited structure slots and direct ones. 2003-10-29 23:16 gb * [r61] : Introduce SLOTS-CLASS; make STD-CLASS and STRUCTURE-CLASS (for now) inherit from it. 2003-10-29 01:00 gb * [r60] : More changes/fixes: line-buffered-p stuff. 2003-10-29 00:59 gb * [r59] : new file 2003-10-27 23:48 gb * [r58] : Suppress some compiler warnings. 2003-10-27 23:47 gb * [r57] : Declare some special variables. 2003-10-27 22:42 gb * [r56] : Use asterisks in (more) special variable names. 2003-10-27 22:28 gb * [r55] : More #-clx stuff, fix typos. 2003-10-27 13:22 gb * [r54] : Fix botch in mark_tcr_tlb(). 2003-10-27 13:21 gb * [r53] : Handle tlb extension 2003-10-27 13:21 gb * [r52] : Bump the fasl version. 2003-10-27 13:20 gb * [r51] : Handle XSYMNOBIND. 2003-10-27 11:24 gb * [r50] : Maintain (weak) inverse mapping from idx to svar. 2003-10-27 11:23 gb * [r49] : nth-value-in-frame, etc: new binding scheme. 2003-10-27 11:18 gb * [r48] : Bump. 2003-10-26 10:12 gb * [r47] : Fix typo. 2003-10-26 10:10 gb * [r46] : Check for non-bindable symbols without trapping. 2003-10-26 10:09 gb * [r45] : Try to special-case NO_THREAD_LOCAL_BINDING_MARKER when walking tlb. 2003-10-26 10:08 gb * [r44] : XSYMNOBIND 2003-10-26 10:06 gb * [r43] : Look for svar-idx in %RESET-OUTERMOST-BINDING. 2003-10-24 16:40 gb * [r42] : (Hopefully) fix %RESET-OUTERMOST-BINDING. 2003-10-24 07:33 gb * [r41] : Enable/fix shallow-binding stuff. 2003-10-24 07:33 gb * [r40] : trlle. 2003-10-24 07:32 gb * [r39] : GC tcr->tlb. 2003-10-24 07:31 gb * [r38] : Make initial tlb a little bigger. 2003-10-24 07:29 gb * [r37] : #$-reader takes more care with constants. 2003-10-24 07:28 gb * [r36] : Cheap evaluator takes more care with constants. 2003-10-24 07:28 gb * [r35] : Avoid consing SVARS unnecessarly when transforming symbols. 2003-10-23 21:46 gb * [r34] : UNDEFINE-CONSTANT. 2003-10-23 21:46 gb * [r33] : Low-level symbol/svar accessors. 2003-10-23 21:45 gb * [r32] : KEYWORD interning sets global value. 2003-10-23 21:45 gb * [r31] : Need an SVAR for symbol-value in general; global versions for when we don't. 2003-10-23 21:43 gb * [r30] : Use UNDEFINE-CONSTANT vice MAKUNBOUND 2003-10-23 19:52 gb * [r29] : New binding stuff. 2003-10-23 19:51 gb * [r28] : Rename old binding subprims. 2003-10-23 19:51 gb * [r27] : SVAR-related tcr fields. 2003-10-23 13:37 gb * [r26] : backend SVAR support 2003-10-23 12:59 gb * [r25] : SVAR support. 2003-10-23 12:57 gb * [r24] : More SVAR type stuff. 2003-10-23 12:57 gb * [r23] : PRINT-OBJECT for svars. 2003-10-23 12:57 gb * [r22] : SVAR class. 2003-10-23 12:57 gb * [r21] : Special-case SVARs in fasl-gvect. 2003-10-23 12:56 gb * [r20] : SVAR binding stuff. 2003-10-23 12:56 gb * [r19] : SVAR type support. 2003-10-23 09:51 gb * [r18] : New svar stuff. (will need GC changes.) 2003-10-23 09:47 gb * [r17] : svar progv support. 2003-10-23 09:47 gb * [r16] : svar tag, struct. 2003-10-23 09:47 gb * [r15] : New subprims for svar binding scheme. 2003-10-22 00:09 gb * [r14] : recover darwin-headers sources 2003-10-22 00:07 gb * [r13] : recover headers sources 2003-10-22 00:00 gb * [r12] : Handle new gc-traps for egc state changes. 2003-10-22 00:00 gb * [r11] : CLASS-SLOT bugs in dcode/reader-trap. 2003-10-21 23:59 gb * [r10] : Serialize EGC state changes. 2003-10-19 08:57 gb * [r6] : Initial revision