00:36:56 -!- nikodemus` [~nikodemus@188-67-206-250.bb.dnainternet.fi] has quit [Quit: This computer has gone to sleep] 00:40:51 nikodemus_: istm :only-one-boxed-constant-for-multiple-uses will fail if we add one slot to code component headers. 00:41:18 chp [~user@c-68-45-180-238.hsd1.nj.comcast.net] has joined #sbcl 00:41:21 (or function header (?)) either way, the leading 0 constant is padding, no? 01:06:20 ah, no. trace table noise. 01:07:19 I'm trying to figure out why a function of mine is not tail-calling 01:07:39 The disassembly is pretty large, but this seems to be the end of it: http://paste.lisp.org/display/126313 01:07:58 the RAX at the top seems to be the function I desire to tail-call to 01:08:19 but it seems that something regarding multiple return values is precluding that? 01:10:33 post the source. 01:10:41 the macro-expanded form ends up being just (let (...) (m-v-b ... (when ... (funcall ...)))) 01:10:45 can't 01:10:56 too bad. 01:11:15 is there anything that jumps out of that disasm? I can describe what you need to know for any key indicators 01:12:04 it gives no optimization nodes, and I've got speed 3 01:12:10 s/nodes/notes/ 01:13:12 specials binding. 01:14:01 You have a dynamic binding around the call; the binding must be unbound, so the call isn't in tail position. 01:14:10 that seemed to fit the pattern, but I didn't know exactly the mechanics 01:14:13 thanks, I'll look for those 01:15:21 which register holds the symbol that's being unbound, if any? 01:18:40 if R12+56 is the stack of values needing to be restored, it seems like this is just tossing 2 entries off there without actually restoring anything 01:21:21 it's something related to dynamic state, but not special binding. 01:22:52 does this look odd to you? I think I can extract this into a test case if I spend a bit of time at it 01:23:02 doesn't look odd, no. 01:24:21 so popping 2 values off it could be cleanup from any number of possible dynamic states? 01:24:36 well, any number of causes 01:24:37 Phoodus: no, it's a sentinel for debugging. 01:25:25 I've got (speed 3) (debug 0) (safety 0) 01:25:33 debug > speed or debug > space enables the insertion of instrumentation that lets you return from frames in the debugger. 01:25:36 seems not. 01:26:12 argh, you're right 01:26:28 however, the run that showed the stack full of these clauses was compiled with those flags 01:26:33 my current in-emacs session did not 01:26:48 *Phoodus* runs more junk in config coherence this time 01:29:09 rebuilt clean, I get the same exit semantics 01:29:26 and that's with the optimization flag enabled 01:30:24 I have neither source nor disassembly. 01:30:40 well, the disassembly is the same 01:30:51 I think I will try to pull this out into a smaller test case 01:31:08 because whatever assumption I'm making in here is probably also going to be causing non-tail-call issues elsewhere 01:31:51 I don't know what your source looks like, but the tail call is being compiled at debug > speed or space. 01:32:32 so that footer should not show up even at default optimization level? 01:35:04 as far as I can tell, that's an unbind-sentinel code sequence. That's only inserted when (policy fun (>= insert-debug-catch 2)) 01:35:37 and that only happens when (> debug (max speed space)), or if you explicitly ask for it. 01:39:03 is there a way in code to dump the current optimization config? 01:43:12 I don't really go there. 01:45:11 even just the standard proclamations 01:58:22 -!- Phoodus [~foo@68.107.217.139] has quit [Read error: Connection reset by peer] 02:01:42 Phoodus [~foo@68.107.217.139] has joined #sbcl 02:02:20 pkhuong: yep, there was a rogue declare optimize deep in a function, probably old debugging cruft leftover. 02:02:53 thanks for the pointers 02:35:19 -!- ASau [~user@95-25-186-18.broadband.corbina.ru] has quit [Read error: Connection reset by peer] 02:36:59 ASau [~user@89-178-205-31.broadband.corbina.ru] has joined #sbcl 03:05:44 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 03:54:55 -!- Phoodus [~foo@68.107.217.139] has quit [Ping timeout: 252 seconds] 04:37:20 stassats` [~stassats@wikipedia/stassats] has joined #sbcl 04:41:07 saschakb_ [~saschakb@p4FEA130D.dip0.t-ipconnect.de] has joined #sbcl 04:44:47 -!- saschakb [~saschakb@p4FEA02D1.dip0.t-ipconnect.de] has quit [Ping timeout: 252 seconds] 05:12:06 Phoodus [~foo@68.107.217.139] has joined #sbcl 05:25:28 -!- antgreen [~user@bas3-toronto06-1177698504.dsl.bell.ca] has quit [Ping timeout: 240 seconds] 05:36:47 saschakb__ [~saschakb@p4FEA036B.dip0.t-ipconnect.de] has joined #sbcl 05:39:29 -!- saschakb_ [~saschakb@p4FEA130D.dip0.t-ipconnect.de] has quit [Read error: Operation timed out] 05:49:19 -!- drl [~lat@110.139.229.172] has quit [Ping timeout: 252 seconds] 06:02:27 drl [~lat@110.139.229.172] has joined #sbcl 06:14:33 I've got a memory fault at a PC address; how can I get a disassembly around there to find out what function it's in? 06:20:49 angavrilov [~angavrilo@217.71.227.181] has joined #sbcl 06:44:18 sb-ext:gc started to take too much time to complete, is that a possible regression? 06:50:01 -!- chp [~user@c-68-45-180-238.hsd1.nj.comcast.net] has quit [Remote host closed the connection] 07:12:43 pon1980 [~pon@195-67-88-105.customer.telia.com] has joined #sbcl 08:36:37 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 248 seconds] 08:39:05 nikodemus [~nikodemus@188-67-206-250.bb.dnainternet.fi] has joined #sbcl 08:39:05 -!- ChanServ has set mode +o nikodemus 08:44:46 Blkt [~user@89-96-199-46.ip13.fastwebnet.it] has joined #sbcl 08:45:56 good morning everyone 09:38:29 -!- lichtblau [~user@port-92-195-14-132.dynamic.qsc.de] has quit [Ping timeout: 248 seconds] 09:41:13 Kryztof [~user@178-83-229-138.dynamic.hispeed.ch] has joined #sbcl 09:41:13 -!- ChanServ has set mode +o Kryztof 09:58:36 saschakb_ [~saschakb@p4FEA0AB6.dip0.t-ipconnect.de] has joined #sbcl 09:58:53 -!- Phoodus [~foo@68.107.217.139] has quit [Ping timeout: 252 seconds] 10:01:57 -!- saschakb__ [~saschakb@p4FEA036B.dip0.t-ipconnect.de] has quit [Ping timeout: 252 seconds] 10:55:57 drl_ [~lat@110.139.229.172] has joined #sbcl 10:57:04 -!- pchrist [~spirit@gentoo/developer/pchrist] has quit [Quit: leaving] 10:57:36 pchrist [~spirit@gentoo/developer/pchrist] has joined #sbcl 11:00:19 I've got a question. 11:00:26 SBCL uses make internally. 11:00:40 Why doesn't it use make for top level?? 11:01:12 At least this would solve problem of finding appropriate gmake to do the job. 11:01:20 (And finding sane shell as well.) 11:02:32 why not ditch make entirely? 11:03:17 I'd rather ditch shell scripts first. 11:05:05 make has one advantage over shell. 11:05:19 Modern make programs handle parallel builds just fine. 11:06:16 And provide you with means to override various values in comprehensible way. 11:08:25 what does this have to do with sbcl ? it can't be compiled in parallel 11:13:09 Runtime can. 11:13:18 Package perhaps can too. 11:13:41 packages 11:14:17 the runtime compilation is such a small portion of the total build time that optimizing it is fairly premature 11:15:09 The main point isn't parallelism anyway. 11:15:22 The main point is more control over various parameters. 11:15:44 in the past, people messing with the compilation parameters have been the source of support headaches 11:16:10 while the sensitivity of the runtime to exotic gcc flags is probably less than it was, we're still talking crufty 25-year-old heritage code 11:17:34 Yes, I've seen notes about NetBSD 1.4 :) 11:20:01 Right now I'm trying to find a place where I could modify CFLAGS for runtime. 11:20:10 I need debug support. 11:20:37 With make I'd just pass it on command line :) 11:21:33 isn't -g on by default? 11:22:26 I don't see enough information in debugger. 11:25:30 in the C debugger? 11:25:38 gdb 11:25:52 well I'm not sure how to get more. I suppose turning the optimization levels down 11:26:15 DGASAU`: maybe you can port https://git.overlays.gentoo.org/gitweb/?p=proj/lisp.git;a=blob;f=dev-lisp/sbcl/files/1.0.51-gentoo-fix_build_system.patch;h=f3dbbc7f37663f76ae37212c27d6faec4199955c;hb=b4f30c95c4e8af9b3be3a151ef7f2c3564011d20 11:27:48 "Port" in which sense and from where to where? 11:28:28 that patch is for 1.0.51 but IIRC doesn't apply cleanly on .54 11:29:19 Is it needed (at least partially) or is it just a homework to try that? 11:29:32 Maybe I can port it. 11:30:34 I don't understand that question 11:39:03 -!- Blkt [~user@89-96-199-46.ip13.fastwebnet.it] has quit [Remote host closed the connection] 11:39:39 I mean that if you suggest to do some work that will be committed, it's alright. 11:42:06 I tried to convince nikodemus & co. to merge it, but failed; that's why I stopped at 1.0.51 11:42:10 maybe you'll have better luck 11:42:23 Blkt [~user@89-96-199-46.ip13.fastwebnet.it] has joined #sbcl 12:00:41 what's the issue? CFLAGS being ignored, or gentoo leaving out -g, or what? 12:01:41 CFLAGS and LDFLAGS being ignored 12:02:08 please put up a bug report on launchpad, including a scenario where they're necessary 12:02:11 something like GNUMAKE="make CC='gcc -g...'" will do 12:03:19 (i freely admit to being sceptical about accepting CFLAGS, because SBCL is easy to break --or at least used to be-- with CFLAGS that don't matter for 99% of the universe 12:04:21 but if real people really need them, then that's that 12:06:25 Nobody assumes that any flags can go. 12:06:49 Still the control is needed from time to time. 12:07:45 i'm willing to believe that, but i'd like at least a single use-case that isn't better handled by making sbcl apply the flag in question by default, or via a specified build option 12:08:24 Real people really need them (same for CC). And those people are perfectly aware that incorrect flags will break the build. 12:09:00 -I/my/copy/of/zlib/headers -L/my/copy/of/zlib 12:10:13 hey, a use case :) 12:10:32 btw, asking for a specific use case here is puzzling. 99% of unix software provided as source accepts CFLAGS. That's how we do it and what we expect, dammit. 12:10:50 It accepts more than one *FLAGS variable. 12:10:56 not using autoconf is already revolutionary enough. 12:11:10 SBCL is in a unix cage, waiting to break free 12:11:15 yep, not only cflags, but LDFLAGS, CPPFLAGS, CC 12:11:23 It would be much easier if SBCL accepted them as well. 12:11:42 Rather finding out what name SBCL invented for PREFIX, I'd rather use PREFIX. 12:12:05 akovalenko: i don't dispute that at all. but having seen all kinds of crap in CFLAGS myself, i'm moderately worried that accepting them tomorrow may break sbcl 12:12:22 nikodemus: no, it will let users break SBCL 12:12:48 I don't see how it breaks SBCL. 12:12:49 nikodemus: and we can let them worry about it (again, that's a long-running tradition) 12:12:55 i'll put it on my todo for this month, but i'm happy to merge a patch as well 12:13:14 many users are clueless 12:13:35 Let them shoot their feet off. 12:13:54 Once they do, there will be less clueless users. 12:14:11 -!- Cryotank2011 [~Cryotank2@c-24-17-62-152.hsd1.wa.comcast.net] has quit [Quit: Cryotank2011] 12:14:27 easy for you, you don't have to deal with all "sbcl won't compile anymore" 12:14:35 nikodemus: if you accept patches, I may find time to prepare them. 12:14:41 stassats`: I have to deal with it anyway. 12:16:15 DGASAU`: sure. put one up launchpad with the "review" tag, or it to sbcl-devel. one request, though: echo all outside flags affecting build at the start of the build (and possibly at the end for a failed build) 12:17:01 I'm not sure gmake supports the latter, not off top of my head. 12:17:19 DGASAU`: via make.sh 12:17:21 But I'll take it into account, alright :) 12:18:38 -!- saschakb_ [~saschakb@p4FEA0AB6.dip0.t-ipconnect.de] has quit [Ping timeout: 252 seconds] 12:18:44 (i'm also not categorically opposed to refactoring the build to be driven from the top down using make, and adding a ./configure script, but i would rather gouge my eyes out with a wooden spoon than admit M4 or autotools to the build) 12:19:45 it's just been easier on a day-to-day basis so far to evolve things using the scripts instead of doing a big refactoring 12:19:47 what about ASDF? 12:19:55 what about it? 12:20:08 build SBCL using ASDF 12:20:18 i don't really see the benefit 12:20:33 own dog food 12:20:47 it's not my dog food 12:21:09 i gave my asdf commit bit away a long time ago, and would not take it back unless i was paid well to do it 12:21:40 or rather, building it completely using lisp, be it ASDF or whatever 12:21:54 that way it's more portable and doesn't depend on different shells or makes 12:22:24 yeah, instead we have to rely on lisps implementing file-write-date correctly, and relative pathname merging correctly 12:22:26 that sounds way more fun 12:22:43 I bet we'd end up writing a function called NEW-APPEND-DIRECTORIES 12:22:44 also, we're still going to maintain a big linear build order for the lisp files with :host-only and :target-only markings, etc, and running different processes -- so ASDF wouldn't really do much to help us with either 12:22:45 we already depend on lisps and workaround things 12:23:00 yes, but why would we want to do more of that? 12:23:47 but i don't quite remember why we can't standardize on eg. bash instead of vanilla sh 12:25:09 so that one can build SBCL on a toaster? 12:26:10 i guess die hard BSD people won't be happy with bash 12:26:20 I think it's solaris which doesn't put it in a sensible place 12:26:58 Yes, we'd rather see ksh instead. 12:27:13 For Solaris we provide ksh. 12:28:00 And, honestly, I don't remeber anyone really using Solaris sh. 12:28:07 In most cases people use ksh or bash. 12:28:50 Joyent uses pkgsrc, and it builds ksh for such modern platforms like Solaris and HP-UX. 12:29:03 interactively, sure. For scripting, really? (Particularly, scripts intended to be portable to other Unixes and Windows?) 12:29:18 Yes, for scripting. 12:29:35 so bash gets installed as /bin/bash these days? 12:29:37 I don't remember what we do for Interix. 12:29:41 No. 12:30:12 "we" as in? 12:30:20 /usr/bin/env bash ? 12:30:22 As in pkgsrc. 12:30:51 nikodemus: no, scripts are passed as argument to interpreter. 12:31:05 Like "$(BASH) configure" 12:31:36 saschakb_ [~saschakb@p4FEA0D01.dip0.t-ipconnect.de] has joined #sbcl 12:33:06 The main motivation for bootstrapping ksh is that configure scripts 12:33:06 need even saner shell than Solaris' /usr/xpg4/bin/sh 12:33:20 not to mention /bin/sh 13:10:32 -!- nikodemus [~nikodemus@188-67-206-250.bb.dnainternet.fi] has quit [Quit: This computer has gone to sleep] 13:14:56 -!- pon1980 [~pon@195-67-88-105.customer.telia.com] has left #sbcl 14:03:07 -!- udzinari [90a0621f@gateway/web/freenode/ip.144.160.98.31] has quit [Ping timeout: 265 seconds] 14:33:41 nyef [~nyef@c-174-63-105-188.hsd1.ma.comcast.net] has joined #sbcl 14:33:46 G'morning all. 14:56:27 slime causes SBCL to deadlock and I'm not sure how to debug it 14:56:38 strace shows many futex(0x1008af1a28, FUTEX_WAIT_PRIVATE 14:57:12 homie [~levgue@xdsl-78-35-168-212.netcologne.de] has joined #sbcl 15:00:15 attila_lendvai [~attila_le@87.247.10.185] has joined #sbcl 15:00:15 -!- attila_lendvai [~attila_le@87.247.10.185] has quit [Changing host] 15:00:15 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 15:06:50 -!- drl [~lat@110.139.229.172] has quit [Quit: Leaving] 15:10:46 antgreen [user@nat/redhat/x-nfoebqwgonmoohyz] has joined #sbcl 15:29:23 nikodemus_: What happens if... Two threads compete on a WP violation for the same page, but the loser is interrupted by a GC before it obtains the free_pages_lock? AIUI, the page doesn't get reprotected because it will be pinned, but the GC wipes the /cleared/ flag... 15:30:06 Hrm. Or maybe not. 15:31:49 When is the cleared flag ever cleared? 15:32:02 There's something wrong in that logic. 15:37:44 mensch [~mensch@c-67-189-241-178.hsd1.ma.comcast.net] has joined #sbcl 16:05:25 nikodemus_: http://ftp.netbsd.org/pub/NetBSD/misc/asau/sbcl/patch-destdir-prefix.diff 16:05:36 nikodemus_: what do you think? 16:15:07 -!- Kryztof [~user@178-83-229-138.dynamic.hispeed.ch] has quit [Ping timeout: 240 seconds] 16:16:35 interesting suggestion from Martin. 16:17:21 Interesting indeed. 16:21:39 Okay, that's the regressions I know about fixed. 16:44:45 lichtblau [~user@port-92-195-58-29.dynamic.qsc.de] has joined #sbcl 16:49:42 homie` [~levgue@xdsl-78-35-166-146.netcologne.de] has joined #sbcl 16:51:52 -!- homie [~levgue@xdsl-78-35-168-212.netcologne.de] has quit [Ping timeout: 240 seconds] 16:52:20 -!- homie` [~levgue@xdsl-78-35-166-146.netcologne.de] has quit [Client Quit] 16:54:41 homie [~levgue@xdsl-78-35-166-146.netcologne.de] has joined #sbcl 17:18:06 nikodemus_: http://ftp.netbsd.org/pub/NetBSD/misc/asau/sbcl/patch-destdir-prefix-r2.diff 17:22:20 -!- Blkt [~user@89-96-199-46.ip13.fastwebnet.it] has quit [Read error: Connection reset by peer] 17:22:25 drl__ [~lat@110.139.229.172] has joined #sbcl 17:23:11 -!- drl_ [~lat@110.139.229.172] has quit [Ping timeout: 240 seconds] 17:29:16 -!- cmm [~cmm@bzq-79-183-68-139.red.bezeqint.net] has quit [Ping timeout: 255 seconds] 17:30:48 cmm [~cmm@109.67.199.25] has joined #sbcl 17:35:09 -!- cmm [~cmm@109.67.199.25] has quit [Ping timeout: 252 seconds] 17:36:05 cmm [~cmm@109.67.199.25] has joined #sbcl 17:45:28 ... did Jim Wise just make --fancy attempt to build threaded darwin/ppc? 17:46:02 Ah, I see... He didn't, he prevented it on x86-64/sunos. 17:58:26 beslyrus [~Brucio-12@adsl-99-49-14-228.dsl.pltn13.sbcglobal.net] has joined #sbcl 17:58:49 milanj [~milanj_@178-223-191-12.dynamic.isp.telekom.rs] has joined #sbcl 18:08:16 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 252 seconds] 18:16:40 -!- stassats` [~stassats@wikipedia/stassats] has quit [Ping timeout: 240 seconds] 19:34:41 Qworkescence [~quad@unaffiliated/quadrescence] has joined #sbcl 19:37:33 sdemarre [~serge@91.176.40.160] has joined #sbcl 19:47:42 Phoodus [~foo@68.107.217.139] has joined #sbcl 20:10:22 Cryotank2011 [~Cryotank2@c-24-17-62-152.hsd1.wa.comcast.net] has joined #sbcl 20:13:06 prxq [~mommer@mnhm-590c17cf.pool.mediaWays.net] has joined #sbcl 20:35:13 -!- angavrilov [~angavrilo@217.71.227.181] has quit [Ping timeout: 252 seconds] 20:43:30 slyrus [~chatzilla@adsl-99-49-14-228.dsl.pltn13.sbcglobal.net] has joined #sbcl 21:39:19 LiamH [~none@pdp8.nrl.navy.mil] has joined #sbcl 21:47:00 So, if we're getting decent threading behavior on OSX now, how are we doing in terms of UI / ObjC bindings? 21:54:15 -!- sdemarre [~serge@91.176.40.160] has quit [Ping timeout: 240 seconds] 22:05:27 nikodemus [~nikodemus@188-67-206-250.bb.dnainternet.fi] has joined #sbcl 22:05:27 -!- ChanServ has set mode +o nikodemus 22:06:35 Hello nikodemus. 22:20:15 hello 22:20:49 i'm completely overfed with a wide variety of cakes. *oof* 22:21:29 -!- homie [~levgue@xdsl-78-35-166-146.netcologne.de] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)] 22:21:40 Sounds like you might have had fun, though. 22:22:14 that's a distinct possibility :) 22:22:36 What did you think of Martin's idea to release 1.0.55 sooner rather than later? 22:23:30 ambivalent 22:23:59 Mmm. I could go either way, I suppose. I would have been against it this morning, though. 22:24:19 "But I still have bugs to fix!" (-: 22:24:19 while we have a lot of bugfixes, we also have a lot of changes that are a bit more interesting than that and could really use some shakedown time 22:25:01 I can see that, yeah. 22:25:27 I know threading doesn't work on darwin/ppc, but I'm not sure about the other non-linux ppc targets. 22:25:47 besides, i was sort of thinking we could call the next one 1.1 or even 1.2 22:26:07 But the only os-specific check for threading for --fancy is for sunos/x86-64. 22:26:24 Hrm. Is there anything else we want for 1.1 or 1.2, though? 22:26:46 pthreads on NetBSD. :) 22:26:59 i have a vague list, but i'm starting to think we should just timebox it once per year 22:27:14 (Alright, alright, I know that I'm supposed to do it myself.) 22:27:15 I suggested that back when we released 1.0, remember? 22:27:40 (Admittedly, I just liked the idea of 1.0 being the ten year anniversary release, the 1.1 being the eleven year anniversary release, and so on.) 22:27:46 i remember it's been talked about previously 22:29:07 if we jumped now to 1.2, we could make 1.2.0 the january 2012 release, so 1.YEAR-2010.MONTH-1 :) 22:29:53 And revisit the entire discussion in 2019? 22:30:06 Psh. That's 2.0.0, obvio. :) 22:30:08 then it would be time for 2.0, obviously 22:30:29 (Did I say 2019? I meant 19119, obviously.) 22:31:39 Ha! 22:31:42 Even CL folks are susceptible to version race! 22:32:11 After some thought on the early release matter, I'd wait too, if only to give people the time to find bugs in the bugfixes. 22:33:39 since the biggest issue is a simple build buglet, i maybe we should just announce the patch for that on sbcl-announce 22:35:07 ... We're up to .48 already, and the release was two days ago? 22:35:25 the thing that irks me about that buglet is that i think i had it fixed at some point. i distinctly remember looking at a patch that had what seemed like whitespace changes during some rebase of mine last month, and deciding "nah, i'm not going to manually re-indent that to minimize the diff" or something like that 22:36:01 wonders of git! 22:36:20 Mmm. Something like that, at least. 22:36:27 i think last week was one of the more productive ones for me, ever, in terms of raw issues dealt with 22:36:38 "of number of raw issues" 22:37:37 is there a way to ask the OS what the protection flags on some page are on posixoid systems? (or mach...) 22:38:20 On linux, I might try /proc/self/maps. Perhaps the wine project has a better option (for VirtualQueryEx)? 22:38:31 mach has something. 22:38:49 But it works on a different granularity than posix, so there's probably a couple hoops to go through. 22:39:58 i'm trying to figure out what those /really rare/ memfaults on darwin could be. i've seen things like that before, but never managed to get my hands on while the process was still up 22:40:34 What's up with that _cleared flag, anyway? 22:40:48 It looks like it's only ever cleared itself in debugging-only code. 22:41:07 Which makes it essentially useless after two flips. 22:41:50 what? 22:42:08 -!- akovalenko [~akovalenk@95.73.48.236] has quit [Ping timeout: 240 seconds] 22:42:29 page_table[whatever].write_protected_cleared. 22:42:48 I vote we spend the time testing the new threads stuff instead of rushing 1.0.55 out the door. 22:42:51 fwiw 22:42:53 it for threaded code. page is protected. T1 hits it, goes to unprotect it. before it unprotects it, T2 hits the same page. so T1 in addition to unprotecting the page also marks it as "previously protected", so T2 knows the fault was not bogus, and that it doesn't need to do anything 22:42:56 nyef: mm... the clearing logic seems buggy. 22:43:10 and nikodemus, I think I said it before, but lemme say it again: "yay!" 22:43:13 nikodemus: when is the _cleared bit reset? 22:43:52 good question 22:44:57 ooh... 22:45:24 I think we want a barrier too 22:45:27 -!- slyrus [~chatzilla@adsl-99-49-14-228.dsl.pltn13.sbcglobal.net] has quit [Ping timeout: 240 seconds] 22:45:43 i bet the bug becomes a lot easier to reproduce if i #define SC_GEN_CK 22:45:49 pkhuong: there are locks around it 22:46:12 ah, ok. 22:46:28 Okay, I need to get ready to get going. 22:47:34 -!- prxq [~mommer@mnhm-590c17cf.pool.mediaWays.net] has quit [Quit: Leaving] 22:47:46 Good luck with the OSX page tables. 22:47:49 -!- nyef [~nyef@c-174-63-105-188.hsd1.ma.comcast.net] has quit [Quit: G'night all.] 22:48:17 akovalenko [~akovalenk@95.72.42.228] has joined #sbcl 22:49:38 beslyrus: so, my current theories re. those memory faults: 22:49:49 (1) our initial mmap of dynamic space overlaps something else (maybe something randomized, so it's nondeterministic), so /mysterious foreign code/ sometimes write-protects parts of our heap that should not be protected. 22:50:10 (2) it is completely bogus and the page isn't even write protected, and darwin is just messing with us 22:50:46 (3) gremlins 22:51:09 nikodemus: 22:51:40 pkhuong: lovely, thank you 22:52:07 we can run that *before* mapping our core in, as well. 22:52:30 very true 22:58:57 -!- antgreen [user@nat/redhat/x-nfoebqwgonmoohyz] has quit [Remote host closed the connection] 22:59:10 it's only 260 MBs in the heap... could be a random stray pointer. 23:00:37 pkhuong: by why do we get the memory fault? the page should either not be protected, or the cleared flag should be set 23:01:13 don't we protect empty pages? 23:02:45 i don't think so 23:03:11 mm, nope. 23:03:11 gc_find_freeish_pages seems to agree with that 23:10:47 Kryztof [~user@77-58-246-74.dclient.hispeed.ch] has joined #sbcl 23:10:47 -!- ChanServ has set mode +o Kryztof 23:11:14 -!- Cryotank2011 [~Cryotank2@c-24-17-62-152.hsd1.wa.comcast.net] has quit [Quit: Cryotank2011] 23:51:45 -!- nikodemus [~nikodemus@188-67-206-250.bb.dnainternet.fi] has quit [Quit: This computer has gone to sleep]