2014-11-13T00:01:24Z Bicyclidine quit (Ping timeout: 256 seconds)
2014-11-13T00:03:13Z Bicyclidine joined #sbcl
2014-11-13T00:04:27Z krzysz00 quit (Ping timeout: 265 seconds)
2014-11-13T00:07:41Z Bicyclidine quit (Ping timeout: 264 seconds)
2014-11-13T00:19:55Z Bicyclidine joined #sbcl
2014-11-13T00:33:01Z slyrus_ joined #sbcl
2014-11-13T00:34:16Z slyrus quit (Ping timeout: 256 seconds)
2014-11-13T00:34:36Z slyrus_ is now known as slyrus
2014-11-13T00:44:09Z krzysz00 joined #sbcl
2014-11-13T00:59:32Z kanru` joined #sbcl
2014-11-13T01:06:04Z krzysz00 quit (Ping timeout: 258 seconds)
2014-11-13T01:08:15Z slyrus quit (Ping timeout: 264 seconds)
2014-11-13T01:08:52Z Bicyclidine quit (Ping timeout: 240 seconds)
2014-11-13T01:09:42Z psilord quit (Quit: Leaving.)
2014-11-13T01:10:52Z Bicyclidine joined #sbcl
2014-11-13T01:18:04Z kanru` quit (Ping timeout: 250 seconds)
2014-11-13T01:30:09Z kanru` joined #sbcl
2014-11-13T01:34:49Z Bicyclidine quit (Ping timeout: 258 seconds)
2014-11-13T01:39:25Z kanru` quit (Ping timeout: 258 seconds)
2014-11-13T01:40:30Z scymtym_ quit (Ping timeout: 244 seconds)
2014-11-13T01:42:39Z nyef sighs.
2014-11-13T01:43:13Z nyef: There's nothing quite like getting through a build and half the test suite wondering "why is this working" before realizing that you forgot the build flag that you're working on.
2014-11-13T01:57:21Z kanru` joined #sbcl
2014-11-13T02:00:51Z brucem: nyef: yesterday, I added some debug and installed the custom compiler in a different location ... /opt/whatever-test/bin ... then I went and added /opt/whatever-test to my path without the bin and spent a long while trying to figure out why my debug stuff never hit.
2014-11-13T02:03:43Z nyef: Now I'm just having trouble getting the build to die because I haven't fixed something up, not because I've managed to screw up some x86oid assembler noise due to being badly out of practice.
2014-11-13T02:27:16Z krzysz00 joined #sbcl
2014-11-13T02:50:15Z krzysz00 quit (Ping timeout: 264 seconds)
2014-11-13T03:31:43Z foom quit (Ping timeout: 272 seconds)
2014-11-13T03:34:51Z krzysz00 joined #sbcl
2014-11-13T03:38:38Z christoph_debian quit (Ping timeout: 265 seconds)
2014-11-13T03:44:08Z foom joined #sbcl
2014-11-13T03:51:56Z christoph_debian joined #sbcl
2014-11-13T03:57:51Z psilord joined #sbcl
2014-11-13T04:09:40Z p_l_ joined #sbcl
2014-11-13T04:10:20Z nyef: Ugh. I make my changes for x86-64, I get a GC fault. I make them for x86, they work... Or I've managed to forget the build flag again.
2014-11-13T04:12:25Z Hydan quit (Ping timeout: 272 seconds)
2014-11-13T04:12:27Z p_l quit (Ping timeout: 272 seconds)
2014-11-13T04:12:27Z joshe quit (Ping timeout: 272 seconds)
2014-11-13T04:12:28Z oleo__ quit (Ping timeout: 272 seconds)
2014-11-13T04:12:29Z tmh_ joined #sbcl
2014-11-13T04:12:30Z Hydan` joined #sbcl
2014-11-13T04:12:41Z Hydan` is now known as Hydan
2014-11-13T04:12:53Z sav quit (Ping timeout: 272 seconds)
2014-11-13T04:13:47Z oleo__ joined #sbcl
2014-11-13T05:26:16Z psilord quit (Quit: Leaving.)
2014-11-13T05:29:48Z psilord joined #sbcl
2014-11-13T05:30:38Z pranavrc joined #sbcl
2014-11-13T05:40:12Z nyef quit (Quit: G'night all)
2014-11-13T05:43:05Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-13T06:29:45Z oleo__ quit (Quit: Verlassend)
2014-11-13T06:52:31Z slyrus joined #sbcl
2014-11-13T07:19:05Z edgar-rft joined #sbcl
2014-11-13T07:33:17Z fridim_ joined #sbcl
2014-11-13T07:46:04Z slyrus quit (Ping timeout: 256 seconds)
2014-11-13T08:15:23Z slyrus joined #sbcl
2014-11-13T08:16:45Z karswell quit (Read error: Connection reset by peer)
2014-11-13T08:18:14Z karswell joined #sbcl
2014-11-13T08:19:15Z karswell quit (Read error: Connection reset by peer)
2014-11-13T08:21:00Z karswell joined #sbcl
2014-11-13T09:09:14Z leo2007 quit (Ping timeout: 265 seconds)
2014-11-13T09:16:04Z leo2007 joined #sbcl
2014-11-13T10:18:09Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-13T10:21:02Z DGASAU joined #sbcl
2014-11-13T10:34:05Z stassats` joined #sbcl
2014-11-13T10:40:10Z attila_lendvai joined #sbcl
2014-11-13T11:00:38Z stassats` quit (Remote host closed the connection)
2014-11-13T11:04:07Z stassats joined #sbcl
2014-11-13T11:19:50Z DGASAU quit (Remote host closed the connection)
2014-11-13T11:20:52Z kanru` quit (Ping timeout: 264 seconds)
2014-11-13T11:21:10Z DGASAU joined #sbcl
2014-11-13T11:35:18Z p_l_ is now known as p_l
2014-11-13T12:04:23Z stassats: turns out, i can disabled that useless reader normalization thing in the readtable
2014-11-13T12:04:29Z stassats: but why is it on by default?
2014-11-13T12:09:22Z stassats: and disabling it is rather peculiar
2014-11-13T12:09:26Z stassats: (setf (sb-impl::%readtable-normalization *readtable*) nil (sb-impl::%readtable-normalization sb-impl::*standard-readtable*) nil)
2014-11-13T12:14:37Z stassats: but if it's still the default, i guess i will have to make it faster, so that people who don't know any better don't suffer
2014-11-13T12:16:38Z kanru` joined #sbcl
2014-11-13T12:25:24Z Krystof: http://www.unicode.org/reports/tr31/#normalization_and_case
2014-11-13T12:25:33Z Krystof: I don't really understand why we're doing NFKC rather than NFC
2014-11-13T12:27:06Z Krystof: maybe my alter ego can comment
2014-11-13T12:40:58Z eudoxia joined #sbcl
2014-11-13T12:53:11Z _8hzp joined #sbcl
2014-11-13T12:56:32Z hzp quit (Ping timeout: 245 seconds)
2014-11-13T13:03:59Z Hache_ joined #sbcl
2014-11-13T13:25:24Z psy_ joined #sbcl
2014-11-13T13:33:37Z nyef joined #sbcl
2014-11-13T13:34:42Z nyef: stassats: I have undefined_tramp and closure_tramp in read-only space on x86/linux with only two debug.impure.lisp test failures most likely due to backtrace breakage.
2014-11-13T13:34:49Z stassats: nyef: what do we use reg_CODE again?
2014-11-13T13:34:50Z stassats: for
2014-11-13T13:35:00Z stassats: constant access, what else?
2014-11-13T13:35:43Z nyef: Access to constants, and letting the GC know which heap object contains the program counter (or interior-pointer).
2014-11-13T13:36:15Z stassats: constants could be accessed relative to PC, but the limited offsets make it a bit more expensive
2014-11-13T13:37:10Z stassats: the GC use case i don't quite understand yet
2014-11-13T13:37:54Z nyef: It's actually pretty well commented. Have a look for scavenge_interrupt_contexts in, IIRC, gc-common.c.
2014-11-13T13:38:27Z kanru` quit (Ping timeout: 272 seconds)
2014-11-13T13:39:22Z stassats: maybe because i still don't really understand how LRA works
2014-11-13T13:41:38Z pranavrc quit
2014-11-13T13:41:56Z nyef: LRA? LRA is simple. Do you mean LIP?
2014-11-13T13:42:33Z nyef: LRA works like a SIMPLE-FUN, only without the whole linked-list thing in the code header.
2014-11-13T13:42:39Z stassats: either
2014-11-13T13:42:49Z stassats: the whole function call sequence
2014-11-13T13:43:42Z nyef: Ah, yeah, the function call sequence is obnoxiously densely packed. /-:
2014-11-13T13:44:13Z stassats: i just need to poke around it
2014-11-13T13:44:58Z stassats: do we need it because of the precision?
2014-11-13T13:45:06Z nyef: Which "it"?
2014-11-13T13:45:12Z stassats: LRA
2014-11-13T13:45:26Z stassats: x86 doesn't have it
2014-11-13T13:45:55Z nyef: Yes, it's a tagged pointer to an "internal" header which says where to find the start of the actual heap object.
2014-11-13T13:46:02Z nyef: If we weren't precise, we could drop it.
2014-11-13T13:46:19Z stassats: what about component_from_pc?
2014-11-13T13:46:32Z stassats: ptr
2014-11-13T13:46:42Z stassats: too slow?
2014-11-13T13:46:43Z nyef: For an arbitrary fixnum on the controls stack?
2014-11-13T13:47:10Z nyef: And to "know" that we're looking at a return address requires reliable backtrace, which we still don't have.
2014-11-13T13:47:52Z stassats: that rings a bell
2014-11-13T13:51:29Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-13T13:53:13Z nyef: Anyway, this weekend or next week I'm hoping to get "read-only tramps" working on SPARC, which should be on par with doing so on ARM, and harder than doing so on PPC. After that, I'll let you have at. (-:
2014-11-13T13:54:08Z stassats: so, how does it work? you aren't just rewriting the assem.S tramps into assembly VOPs?
2014-11-13T13:55:56Z psilord quit (Quit: Leaving.)
2014-11-13T13:57:24Z nyef: Genesis needs to be able to find at least undefined_tramp before loading the main series of fasls, so there's a hook there.
2014-11-13T13:57:40Z DGASAU joined #sbcl
2014-11-13T13:58:19Z nyef: x86 is easiest, no headers, no alignment requirement, just an address and trap.
2014-11-13T13:58:51Z nyef: PPC has an alignment requirement because there's a header.
2014-11-13T13:59:25Z nyef: SPARC and ARM use tagged tramps, so they need a displaced pointer on the assembly-routine.
2014-11-13T14:00:05Z nyef: But yeah, it's rewriting the arch-assem.S tramps into assembly-routines (not VOPs), and updating the three or so references from various VOPs in the backend.
2014-11-13T14:00:12Z nyef: Plus possibly some impact to the debugger.
2014-11-13T14:00:25Z nyef: ... And maybe the GC as well.
2014-11-13T14:18:33Z stassats: why is it named interior?
2014-11-13T14:20:12Z nyef: Because it points to the interior of a heap object, not to its header.
2014-11-13T14:20:35Z stassats: when does this happen?
2014-11-13T14:21:28Z nyef: Depends on the register.
2014-11-13T14:21:34Z nyef: For the program counter, most of the time.
2014-11-13T14:21:51Z nyef: For LIP, typically during array access and function call/return.
2014-11-13T14:22:06Z stassats: do we not access arrays with offsets?
2014-11-13T14:22:34Z stassats: and why should PC be handled specially?
2014-11-13T14:23:00Z nyef: Because it usually points to within a CODE-OBJECT.
2014-11-13T14:23:10Z stassats: what about reg_CODE then?
2014-11-13T14:23:24Z nyef: It usually points to the CODE-OBJECT in question, but not always.
2014-11-13T14:23:25Z stassats: it points to the header
2014-11-13T14:23:44Z stassats: so, when it was interrupted half way through the call, we need to consult the PC?
2014-11-13T14:24:02Z oleo joined #sbcl
2014-11-13T14:24:53Z nyef: If it's interrupted halfway through a call or return sequence, then some other register (not reg_CODE) can be pointing to the nearest available header for PC.
2014-11-13T14:24:59Z kanru` joined #sbcl
2014-11-13T14:25:31Z nyef: Consider what happens on entry before the COMPUTE-CODE pseudo-instruction runs?
2014-11-13T14:27:28Z stassats: and we need all this because we move the code object during gc
2014-11-13T14:28:49Z nyef: Right.
2014-11-13T14:29:29Z nyef: And we can't necessarily STOP moving code objects because of the fun-end-breakpoint noise, not that normal breakpoints work correctly anyway.
2014-11-13T14:29:44Z stassats: does fun-end-breakpoint even work?
2014-11-13T14:29:55Z nyef: On ARM? Probably not.
2014-11-13T14:30:13Z nyef: But I went to quite a bit of trouble to make sure that it works on PPC, for example.
2014-11-13T14:30:26Z stassats: does it?
2014-11-13T14:30:38Z nyef: It did at one point.
2014-11-13T14:30:53Z jlarocco joined #sbcl
2014-11-13T14:30:59Z nyef: The entire breakpoint mechanism is conceptually broken, though.
2014-11-13T14:31:30Z stassats: is it ever only used from lisp?
2014-11-13T14:31:50Z stassats: that is fun_end_breakpoint_trap
2014-11-13T14:32:28Z stassats: i'm asking because it uses reg_ZERO
2014-11-13T14:32:30Z nyef: It's not even used directly from Lisp, the debugger allocates a new code-object and copies the trap in, then uses the copy.
2014-11-13T14:33:12Z nyef: But yeah, the commentary makes it quite clear that it's an LRA without a corresponding function, and we only ever use those from lisp.
2014-11-13T14:33:36Z nyef: IIRC, the PPC version is the one with the explanation, as well.
2014-11-13T14:34:56Z stassats: so, how does it affect code object movement?
2014-11-13T14:36:43Z nyef: It's a purely-temporary code object, only ever returned to once.
2014-11-13T14:37:00Z nyef: And they're small.
2014-11-13T14:37:36Z nyef: On the other hand, they're all fixed-size, so mark/sweep + freelist could work quite well for them.
2014-11-13T14:38:42Z stassats: i don't quite get what fun_end_breakpoint_trap does
2014-11-13T14:39:06Z stassats: if it's an unrelated code-object, what does it break?
2014-11-13T14:39:51Z nyef: Umm... maybe it doesn't at this point. I'm possibly confusing myself as well as you.
2014-11-13T14:40:47Z stassats: can gengc survive some objects not being moved?
2014-11-13T14:41:44Z stassats: because i don't think the whole LRA dance is good for performance, especially on ARM
2014-11-13T14:42:47Z nyef: It can survive some objects being pinned, with a few caveats.
2014-11-13T14:45:37Z stassats: so, what are the downsides to pinning code objects on the stack?
2014-11-13T14:46:46Z nyef: Hrm. You're still talking about conservatism, but a substantially reduced conservatism, and you can guarantee that it only applies to fixnums...
2014-11-13T14:47:40Z stassats: i don't think it's conservatism, you wouldn't gc those objects anyway, just move
2014-11-13T14:48:28Z nyef: It's conservatism: You can construct a fixnum that is, in fact, a fixnum, that happens to share an unboxed value with the insides of a code-object.
2014-11-13T14:48:40Z stassats: of course, the other way is safepoints , but it's still not know how well safepoints can be implemented on ARM
2014-11-13T14:49:41Z stassats: nyef: if the code objects doesn't move, why does it matter?
2014-11-13T14:49:58Z DGASAU quit (Remote host closed the connection)
2014-11-13T14:50:45Z nyef: It's a false positive on a liveness test.
2014-11-13T14:50:58Z nyef: It may not matter in the end, but it's still a conservatism.
2014-11-13T14:51:07Z stassats: liveness of what?
2014-11-13T14:51:33Z psy_ quit (Remote host closed the connection)
2014-11-13T14:51:37Z DGASAU joined #sbcl
2014-11-13T14:52:23Z nyef: The code object that's referenced only by a fixnum on the stack that just happens to coincide with its address on the heap.
2014-11-13T14:52:54Z nyef: And at this point I need to get some actual work done, so I'm going to be a lot less responsive for a bit. (-:
2014-11-13T14:53:50Z stassats: but we know it's a fixnum
2014-11-13T14:54:30Z nyef: How, when the other option is an unboxed, aligned, return address?
2014-11-13T14:54:50Z stassats: because it's in the right register?
2014-11-13T14:55:18Z nyef: No, it's on the control stack.
2014-11-13T14:55:37Z psy_ joined #sbcl
2014-11-13T14:56:03Z nyef: And your most-likely next objection is answered by "because we don't have reliable backtraces to find the stack frames".
2014-11-13T14:56:19Z stassats: no, i know that
2014-11-13T14:57:50Z stassats: so, we need both non moving code and safepoints
2014-11-13T14:58:03Z psy__ joined #sbcl
2014-11-13T14:58:17Z psy__ quit (Read error: Connection reset by peer)
2014-11-13T14:59:22Z stassats: i guess i can implement safepoints for arm, shouldn't be to hard
2014-11-13T14:59:28Z stassats: just to see how well they work
2014-11-13T15:00:51Z stassats: but as i said before, it's easier to get rid of reg_NULL
2014-11-13T15:00:57Z stassats: than reg_CODE
2014-11-13T15:02:26Z nyef: ... What happens if you move reg_CODE to inside the thread structure?
2014-11-13T15:03:02Z stassats: more expensive function calls?
2014-11-13T15:03:35Z nyef: Well, yes, but it would also free up the underlying register to hold the thread structure pointer in the first place.
2014-11-13T15:03:41Z stassats: but with immobile stack code objects and safepoints, we can get rid of both reg_CODE and _NULL
2014-11-13T15:04:44Z psilord joined #sbcl
2014-11-13T15:05:20Z stassats: and then use on of them for reg_THREAD
2014-11-13T15:24:04Z krzysz00 quit (Ping timeout: 244 seconds)
2014-11-13T15:32:20Z slyrus quit (Ping timeout: 244 seconds)
2014-11-13T15:38:34Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-13T15:39:01Z slyrus joined #sbcl
2014-11-13T15:40:04Z Hache_ quit (Ping timeout: 258 seconds)
2014-11-13T15:43:38Z DGASAU joined #sbcl
2014-11-13T15:48:39Z slyrus quit (Remote host closed the connection)
2014-11-13T16:07:53Z eudoxia quit (Quit: Lost terminal)
2014-11-13T16:27:02Z White_Flame quit (Read error: Connection reset by peer)
2014-11-13T16:28:15Z White_Flame joined #sbcl
2014-11-13T16:39:33Z rszeno joined #sbcl
2014-11-13T16:44:32Z eudoxia joined #sbcl
2014-11-13T16:55:09Z stassats: (write (char-name #\a) :readably t) alright, but (write (char-name #\Nul) :readably t) isn't
2014-11-13T16:55:33Z stassats: because the latter comes from SB-IMPL::*BASE-CHAR-NAME-ALIST*
2014-11-13T16:56:07Z stassats: is that alright? i don't know. does that mean i can make all other char names to be simple-base-strings as well?
2014-11-13T17:06:43Z krzysz00 joined #sbcl
2014-11-13T17:07:00Z nyef: You're asking about similarity between base string and string?
2014-11-13T17:07:11Z stassats: base string are not printable
2014-11-13T17:07:19Z stassats: by sbcl, for arbitrary reasons
2014-11-13T17:07:33Z stassats: so, some results from char-name are printable, and some are not
2014-11-13T17:07:46Z stassats: printable readably
2014-11-13T17:07:59Z stassats: read printable? print readable?
2014-11-13T17:08:09Z stassats: preadable?
2014-11-13T17:08:19Z nyef: readably printable?
2014-11-13T17:09:20Z stassats: also, those names are also capitilzied
2014-11-13T17:09:29Z stassats: (char-name #\ack) => "Ack"
2014-11-13T17:09:34Z stassats: (char-name #\a) => "LATIN_SMALL_LETTER_A"
2014-11-13T17:10:38Z nyef: Hrm. This is interesting. CLHS 3.2.4.2.2 doesn't define a special rule for strings distinct from arrays.
2014-11-13T17:10:54Z stassats: CLHS 3.2.4.2.2
2014-11-13T17:10:54Z specbot: Definition of Similarity: http://www.lispworks.com/reference/HyperSpec/Body/03_bdbb.htm
2014-11-13T17:11:17Z stassats: nyef: this is similarity, not pread read
2014-11-13T17:11:20Z stassats: argh
2014-11-13T17:11:21Z stassats: print read
2014-11-13T17:11:52Z stassats: does "Ack" look like a base string or a string to you?
2014-11-13T17:12:25Z Krystof: yes
2014-11-13T17:13:01Z stassats: #.(coerce 'simple-base-string "Nul") would be, but i'm not asking about that, really that point
2014-11-13T17:13:11Z stassats: i'm asking what should be expected from CHAR-NAME
2014-11-13T17:13:25Z nyef: stassats: Print-read is defined in terms of similarity.
2014-11-13T17:13:30Z stassats: if nothing can be expected, then i can make it more efficient by always returning s-b-s
2014-11-13T17:13:36Z stassats: nyef: similarity is for fasls
2014-11-13T17:14:21Z nyef: Variable *PRINT-READABLY* defines that it affects the reader in terms of similarity.
2014-11-13T17:14:40Z stassats: i guess we can't print arrays either then
2014-11-13T17:15:21Z stassats: damn slime is wacky again
2014-11-13T17:17:07Z stassats: of course, slime-autodoc-use-multiline-p disappeared
2014-11-13T17:17:18Z stassats: the bloody thing is getting on my nerves
2014-11-13T17:19:12Z nyef: There's a reason why I can't be bothered to use SLIME anymore. (-:
2014-11-13T17:19:40Z stassats: i didn't want to updated it, but then emacs 24.4 broke things
2014-11-13T17:20:25Z stassats: and that's why i don't contribute to slime anymore, but my slime replacement is progressing really slowly (as in not at all)
2014-11-13T17:22:35Z stassats: (slime-autodoc-use-multiline-p): Deleted. Use eldoc-display-message-no-interference-p instead.
2014-11-13T17:22:56Z stassats: there's no damn eldoc-display-message-no-interference-p variable
2014-11-13T17:24:40Z stassats: it meant eldoc-echo-area-use-multiline-p, and it has to be set to NIL, not truncate-sym-name-if-fit
2014-11-13T17:26:50Z stassats: (write (make-array 10 :element-type 'fixnum) :readably t) => #.(COERCE #(0 0 0 0 0 0 0 0 0 0) '(SIMPLE-ARRAY FIXNUM (*)))
2014-11-13T17:32:17Z stassats: clhs 22.1.3.4
2014-11-13T17:32:17Z specbot: Printing Strings: http://www.lispworks.com/reference/HyperSpec/Body/22_acd.htm
2014-11-13T17:32:21Z stassats: doesn't mention *print-readably*
2014-11-13T17:32:57Z eudoxia quit (Quit: Lost terminal)
2014-11-13T17:35:19Z attila_lendvai quit (Quit: Leaving.)
2014-11-13T17:35:52Z attila_lendvai joined #sbcl
2014-11-13T17:35:52Z attila_lendvai quit (Changing host)
2014-11-13T17:35:52Z attila_lendvai joined #sbcl
2014-11-13T17:36:21Z attila_lendvai quit (Client Quit)
2014-11-13T17:38:11Z rszeno quit (Ping timeout: 265 seconds)
2014-11-13T17:47:33Z stassats: target-unicode raises a lot of questions
2014-11-13T17:47:37Z stassats: target-unicode.lisp
2014-11-13T17:54:58Z rszeno joined #sbcl
2014-11-13T18:01:07Z kanru` quit (Ping timeout: 245 seconds)
2014-11-13T18:04:05Z krzysz00 quit (Ping timeout: 264 seconds)
2014-11-13T18:31:30Z stassats: got (time (loop for i below 10000 do (unicode< "abcювфыв" "ювфывabc"))) from 0.241 seconds 230,086,448 bytes consed, to 0.134 seconds 98,851,616 bytes
2014-11-13T18:33:05Z stassats: although if i understood what it's doing i could do more
2014-11-13T18:33:34Z stassats: i just removed questionable things like generic-apply and all the frivolous consing with subseq and others
2014-11-13T18:44:05Z psilord quit (Quit: Leaving.)
2014-11-13T18:58:22Z les quit (Ping timeout: 240 seconds)
2014-11-13T19:06:59Z rszeno quit (Quit: Leaving.)
2014-11-13T19:20:42Z stassats: argh, &optional in public API
2014-11-13T19:20:56Z stassats: and now i can't do anything about it
2014-11-13T19:21:01Z stassats: (at least without annoying users)
2014-11-13T19:30:47Z prxq joined #sbcl
2014-11-13T19:31:43Z stassats: not good, GC invariant lost, gencgc.c:860
2014-11-13T19:39:47Z les joined #sbcl
2014-11-13T19:40:04Z les quit (Changing host)
2014-11-13T19:40:04Z les joined #sbcl
2014-11-13T19:59:31Z stassats: (let ((x (* a b))) (declare (ignore x)) ...) am i missing something?
2014-11-13T20:11:52Z antoszka quit (Read error: Connection reset by peer)
2014-11-13T20:12:04Z antoszka joined #sbcl
2014-11-13T20:16:57Z antoszka quit (Ping timeout: 245 seconds)
2014-11-13T20:24:55Z antoszka joined #sbcl
2014-11-13T20:25:04Z antoszka quit (Changing host)
2014-11-13T20:25:04Z antoszka joined #sbcl
2014-11-13T20:52:04Z eudoxia joined #sbcl
2014-11-13T20:56:12Z eudoxia quit (Client Quit)
2014-11-13T21:29:02Z Bicyclidine joined #sbcl
2014-11-13T21:34:49Z DGASAU quit (Ping timeout: 245 seconds)
2014-11-13T21:38:30Z DGASAU joined #sbcl
2014-11-13T22:16:06Z scymtym_ joined #sbcl
2014-11-13T22:16:48Z nyef: Is it safe to presume that the only code in read-only space is assembly-routines, or is anyone likely to start bodging more code in there at runtime?
2014-11-13T22:23:31Z prxq quit (Remote host closed the connection)
2014-11-13T22:47:43Z nyef: Backtrace for: #<SB-THREAD:THREAD "main thread" RUNNING {AF1E5F9}>
2014-11-13T22:47:43Z nyef: 0: ("assembly-routine UNDEFINED-TRAMP")
2014-11-13T22:48:47Z nyef: So, the question is, do I leave it at this, or do I specifically check for UNDEFINED-TRAMP and bounce it to "undefined function", which should cause the affected test cases to pass as-is?
2014-11-13T22:50:43Z nyef: Seeing "assembly-routine UNWIND" for a throw to NO-SUCH-TAG is also nice. (-:
2014-11-13T22:54:28Z fridim_ quit (Ping timeout: 264 seconds)
2014-11-13T22:56:03Z edgar-rft quit (Quit: activity vanished by mental contermination)
2014-11-13T23:08:14Z krzysz00 joined #sbcl
2014-11-13T23:22:44Z krzysz00 quit (Ping timeout: 245 seconds)
2014-11-13T23:48:54Z ehaliewicz joined #sbcl