00:19:43 slyrus_ [~chatzilla@adsl-71-146-132-158.dsl.pltn13.sbcglobal.net] has joined #sbcl 00:29:18 lambda-avenger [~roman@adsl-99-170-145-174.dsl.pltn13.sbcglobal.net] has joined #sbcl 01:09:59 -!- Blkt [~Blkt@dynamic-adsl-94-37-227-177.clienti.tiscali.it] has quit [Quit: Error: do not makunbound t please!] 01:10:09 -!- homie` [~levgue@xdsl-84-44-209-222.netcologne.de] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)] 01:10:58 cliftonk [~cliftonk@74.198.150.157] has joined #sbcl 01:14:37 homie [~levgue@xdsl-84-44-209-222.netcologne.de] has joined #sbcl 01:21:04 -!- cliftonk [~cliftonk@74.198.150.157] has quit [Quit: Colloquy for iPad - http://colloquy.mobi] 01:25:38 -!- tsuru [~charlie@adsl-179-29-10.bna.bellsouth.net] has quit [Remote host closed the connection] 01:27:35 -!- attila_lendvai [~attila_le@catv-80-98-24-21.catv.broadband.hu] has quit [Ping timeout: 240 seconds] 01:31:51 attila_lendvai [~attila_le@apn-94-44-3-126.vodafone.hu] has joined #sbcl 01:32:14 tsuru [~charlie@adsl-179-29-10.bna.bellsouth.net] has joined #sbcl 01:34:18 -!- slyrus_ [~chatzilla@adsl-71-146-132-158.dsl.pltn13.sbcglobal.net] has quit [Ping timeout: 250 seconds] 02:00:58 homie` [~levgue@xdsl-78-35-143-165.netcologne.de] has joined #sbcl 02:01:26 -!- attila_lendvai [~attila_le@apn-94-44-3-126.vodafone.hu] has quit [Ping timeout: 276 seconds] 02:03:30 -!- homie [~levgue@xdsl-84-44-209-222.netcologne.de] has quit [Ping timeout: 240 seconds] 02:04:19 -!- homie` [~levgue@xdsl-78-35-143-165.netcologne.de] has quit [Client Quit] 02:12:15 rpg [~rpg@216.243.156.16] has joined #sbcl 02:12:22 -!- rpg [~rpg@216.243.156.16] has quit [Client Quit] 02:23:47 homie [~levgue@xdsl-78-35-143-165.netcologne.de] has joined #sbcl 04:17:52 -!- nyef [~nyef@pool-70-20-57-204.man.east.myfairpoint.net] has quit [Quit: G'night all.] 04:48:18 slyrus_ [~chatzilla@adsl-75-36-217-249.dsl.pltn13.sbcglobal.net] has joined #sbcl 07:16:44 flip214 [~marek@unaffiliated/flip214] has joined #sbcl 08:20:04 tcr1 [~tcr@213.55.131.17] has joined #sbcl 08:35:32 -!- tcr1 [~tcr@213.55.131.17] has quit [Read error: Connection reset by peer] 08:41:56 tcr1 [~tcr@213.55.131.17] has joined #sbcl 09:05:55 -!- tcr1 [~tcr@213.55.131.17] has quit [Read error: Connection reset by peer] 09:15:30 tcr1 [~tcr@213.55.131.17] has joined #sbcl 09:17:44 stassats [~stassats@wikipedia/stassats] has joined #sbcl 10:38:24 ASau [~user@95-27-211-235.broadband.corbina.ru] has joined #sbcl 10:42:35 -!- lambda-avenger [~roman@adsl-99-170-145-174.dsl.pltn13.sbcglobal.net] has left #sbcl 10:59:41 mega1 [~quassel@catv4E5CABA2.pool.t-online.hu] has joined #sbcl 11:13:55 hargettp [~hargettp@pool-71-184-189-200.bstnma.east.verizon.net] has joined #sbcl 12:09:39 -!- tcr1 [~tcr@213.55.131.17] has quit [Read error: Connection reset by peer] 12:18:15 -!- hargettp [~hargettp@pool-71-184-189-200.bstnma.east.verizon.net] has quit [Ping timeout: 240 seconds] 12:19:21 tcr1 [~tcr@213.55.131.17] has joined #sbcl 12:33:11 -!- tcr1 [~tcr@213.55.131.17] has quit [Quit: Leaving.] 12:47:46 -!- stassats [~stassats@wikipedia/stassats] has quit [Ping timeout: 255 seconds] 13:28:16 nyef [~nyef@pool-70-20-57-204.man.east.myfairpoint.net] has joined #sbcl 13:28:30 G'morning all. 13:29:28 good morning nyef 13:29:42 stassats [~stassats@wikipedia/stassats] has joined #sbcl 13:32:15 Anything going on? 13:32:33 Besides the x86-64/darwin build apparently breaking, that is? 13:39:51 My nightly openbsd build broke last night too, but I'm not sure it isn't something at my end 13:51:43 -!- flip214 [~marek@unaffiliated/flip214] has quit [Remote host closed the connection] 13:57:17 hargettp [~hargettp@dhcp-162.mirrorimage.net] has joined #sbcl 14:15:19 -!- hargettp [~hargettp@dhcp-162.mirrorimage.net] has quit [Quit: hargettp] 14:16:01 tcr1 [~tcr@213.55.131.17] has joined #sbcl 14:32:03 hargettp [~phil@dhcp-162.mirrorimage.net] has joined #sbcl 14:32:40 -!- hargettp [~phil@dhcp-162.mirrorimage.net] has quit [Remote host closed the connection] 14:33:09 hargettp [~hargettp@dhcp-162.mirrorimage.net] has joined #sbcl 14:52:36 -!- homie [~levgue@xdsl-78-35-143-165.netcologne.de] has quit [Read error: Connection reset by peer] 14:54:09 homie [~levgue@xdsl-78-35-166-152.netcologne.de] has joined #sbcl 14:55:50 -!- homie [~levgue@xdsl-78-35-166-152.netcologne.de] has quit [Client Quit] 14:57:32 homie [~levgue@xdsl-78-35-166-152.netcologne.de] has joined #sbcl 15:35:14 -!- jiacobucci1 [~jiacobucc@gw-asdl.ae.gatech.edu] has quit [Read error: Connection reset by peer] 15:39:53 flip214 [~marek@unaffiliated/flip214] has joined #sbcl 15:45:10 -!- flip214 [~marek@unaffiliated/flip214] has quit [Ping timeout: 240 seconds] 15:50:58 flip214 [~marek@unaffiliated/flip214] has joined #sbcl 16:25:31 Blkt [~user@net-93-151-248-118.cust.dsl.teletu.it] has joined #sbcl 16:25:44 ok, getpagesize is wrong 16:26:06 can somebody test the osf1 build with BACKEND_PAGE_BYTES for os_vm_page_size? 16:26:08 :) 16:26:49 so... based on martin cracauer's extensive testing of backend page size on linux, are we just going to use 32k everywhere? 16:28:39 and why don't we call make-config.sh for grovel-features.sh? 16:28:44 I'm going to abstain from this one. 16:30:27 osf1 build? 16:30:49 well, whatever OSes use osf1-os.c anyway 16:30:51 The question of page size. 16:31:07 foom: I was mostly joking 16:31:53 but I'm serious about not knowing how best to get -ldl added to the make command in grovel-features.sh... 16:33:21 I think you can do it unconditionally 16:33:41 my linux, which is really not natty nifty numinous nincompoop, has a man page which says "link with -ldl" 16:33:52 linux only? 16:33:53 Unconditionally on Linux, right? 16:35:21 ITA has been using the 32k page-size for a long time by now, and martin recently actually tested its relative speed vs 4/16/64k sizes. That decision might need to be changed if the GC ever gets fixed to be better, of course. 16:36:01 uh, where "recently" is over a year ago by now, apparently. :) 16:36:41 joshe: that's what I'm thinking 16:37:05 Well unconditionally on all OSes is clearly not the thing to do :) 16:37:05 and darwin x86-64 builds with 32k pagesize now 16:38:20 I'm trying an OpenBSD build with the page size fix to bsd-os.c right now 16:38:21 oh lameness 16:38:32 ok, I return you to your wondering about -ldl 16:39:33 ok, I see the grovel-features picks up Config.* 16:39:52 bah. x86-64 everywhere but linux where its x86_64. 16:39:55 it's 16:40:32 oh, but that's -ldl is already in there... hrmm... 16:40:56 It's amd64 on the (real) BSDs 16:42:27 boinkmarks is showing some regressions with the 32kb patch -- mostly with the arrays -- and some speed-ups...case of microbenchmarks not reflecting real world usage? 16:46:19 jiacobucci [~jiacobucc@gw-asdl.ae.gatech.edu] has joined #sbcl 16:48:20 huh? 16:48:34 cc -ldl -lpthread os-provides-dladdr-test.c -o os-provides-dladdr-test # breaks 16:48:47 cc os-provides-dladdr-test.c -o os-provides-dladdr-test -ldl -lpthread # works 16:48:59 OpenBSD/amd64 builds fine with os_vm_page_size = BACKEND_PAGE_BYTES; 16:49:07 where's my copy of the unix hater's handbook? 16:49:25 I'd try the test suite too but I have to run 16:49:29 where's the barf-bag? 16:51:24 slyrus_: Let me guess... it's an ordering dependency, the first version effectively ignores -ldl because none of its symbols are yet in need of definition by the linker? 16:51:31 slyrus_: well obviously! Link order has always mattered. :) 16:51:41 nyef: exactly so. 16:52:01 foom: not until recently -- or at least the OSes used to bail us out on this one 16:52:24 It didn't (doesn't?) matter with dynamic libraries. 16:52:31 it has always mattered with static libs 16:52:40 so this line in src/runtime/GNUmakefile should probably change: $(CC) ${LINKFLAGS} -o $@ $^ $(LIBS) 16:52:54 that looks fine: LIBS is at the end 16:52:55 foom: I think the point was that stuff like -ldl used to get pulled in be default 16:53:08 foom: ah, but -ldl isn't in LIBS 16:53:18 oh, yes it is 16:53:20 bnvm 16:53:22 nvm 16:53:41 slyrus_: yes, the linker behavior changed recently to require explicitly specifying dependencies instead of pulling them implicitly from other linked libraries. 16:54:08 a.k.a. --as-needed 16:54:23 ok, well perhaps we can't just use the default CC build rule in tools-for-build/Makefile anymore then 16:54:30 or we need to better specify the args for it 16:54:50 no, not --as-needed. --as-needed is a totally different thing. :) 16:55:29 you mean: *add-needed* 16:55:47 ah, LDLIBS is perhaps what we should have been using 16:56:33 yay success 16:57:32 foom: now can you explain the waitpid thing to me? :) 16:58:45 foom: nope, that's also one of the effects of --as-needed, not linking transitively 17:00:15 fe[nl]ix: You sure? Note that the default behavior changed to --no-add-needed recently. 17:02:39 yay. success on linux(2.6.37-12-generic)/x86-64. 17:04:26 *slyrus_* is amazed that he got by hacking on SBCL for so long without git stash 17:05:16 Mmm. Git just makes it all easier, doesn't it? 17:05:35 I'm a big fan of git rebase --interactive, myself. 17:05:44 yea, wasn't the official repository going to switch to git at some point? :) 17:05:46 foom: you're right, and that flag now got renamed to --copy-dt-needed-entries 17:07:40 And IMO that change (the behavior, not the name) is pretty retarded, but nobody asked me. :) 17:09:09 nyef: sub 5-minute builds are pretty nice too :) 17:09:14 foom: it's pretty useful 17:09:27 slyrus_: Gah! 17:09:32 I get half-hour builds. 17:09:34 It really doesn't work out very well for C++ libraries, whose headers often include uses of the indirect dependencies. 17:09:48 just enough time to get some coffee, rather than walk to the coffee shop, read the paper, drink the coffee, etc... 17:10:07 I'm thinking that my tablet (RIP) did a build in about 10 minutes. 17:10:25 Ok, the lappy is still clocking in at around 9:30, but it's not too bad... 17:11:07 foom: the use case is when the dependency of a widely-used library(gtk+ for instance) changes soname 17:11:22 And the nvidia settlement designated replacement is functionally identical to my current netbook, which clocks 30 minute builds. 17:11:23 fe[nl]ix: incorrect. sonames got correctly encoded into the binary in any case. 17:11:30 foom: if DT_NEEDED entries are copied, all users of that library must be recompiled 17:12:03 fe[nl]ix: except it used to only copy indirect ones as-needed. 17:13:46 foom: most gtk/qt apps don't use any of libX11's symbols directly 17:14:00 kclifton [~kclifton@s198-166-45-245.ab.hsia.telus.net] has joined #sbcl 17:15:17 fe[nl]ix: so the combination of --as-needed and --no-add-needed helped avoid a great many compilations to distros 17:15:25 and Gentoo users(like me) 17:17:40 fe[nl]ix: Nope. --no-add-needed does nothing for you there. -lqt-mt wouldn't add a DT_NEEDED for libX11.so.6 unless the program actually used a symbol from it. --as-needed of course helps by omitting explicitly -l'd libraries. 17:18:06 -!- tcr1 [~tcr@213.55.131.17] has quit [Quit: Leaving.] 17:18:22 (Most of the excessive -l options came from stupid gnu autotools' libtool) 17:20:34 -!- mega1 [~quassel@catv4E5CABA2.pool.t-online.hu] has quit [Remote host closed the connection] 17:24:08 foom: just tested it: http://pastebin.com/beVjaCrU 17:25:33 fe[nl]ix: ldd recursively trawls dependencies. use readelf -d instead 17:35:36 foom: oops, you're right 17:58:18 Fare [~Fare@ita4fw1.itasoftware.com] has joined #sbcl 17:58:45 is the default allocation page size still 4KiB on amd64 ? 17:59:50 |3b|`` [foobar@cpe-72-179-19-4.austin.res.rr.com] has joined #sbcl 18:00:05 -!- |3b|` [foobar@cpe-72-179-19-4.austin.res.rr.com] has quit [Remote host closed the connection] 18:02:21 Fare: not as of yesterday. 18:03:55 yay 18:04:04 what's the new default? 18:04:17 It's whatever martin set it to. 18:04:25 -!- hargettp [~hargettp@dhcp-162.mirrorimage.net] has quit [Remote host closed the connection] 18:04:27 32K 18:04:28 32K, isn't it? 18:04:28 32K? 18:04:41 better. 18:04:49 hargettp [~hargettp@dhcp-162.mirrorimage.net] has joined #sbcl 18:06:00 can a committer update asdf to 2.012 while we're at it? 18:06:55 nothing urgent, but a few changes that should make asdf more future-proof 18:08:55 any objections to adding waitpid to the ldso stubs and the undefineds? 18:09:00 Hunh. LVARs aren't the only things in IR1 with a "USE" slot that is in some sense the 'source' for something (set of sources in the case of LVARs). 18:09:15 slyrus_: No objections here. 18:21:50 ok, darwin and recent linux changes pushed. 18:22:08 no back to regularly scheduled real work. 18:22:11 s/no/now/ 18:22:36 feel free to test on various platforms (and older linuxen) 18:27:29 -!- slyrus_ [~chatzilla@adsl-75-36-217-249.dsl.pltn13.sbcglobal.net] has quit [Ping timeout: 276 seconds] 18:31:52 rmarynch [~roman@88.135.194.233] has joined #sbcl 18:32:06 Good evening all 18:32:26 Hello rmarynch. 18:34:20 Does anybody know the rough estimate of SBCL porting to Win64, including threads? 18:34:37 In man-months 18:39:03 I don't know, but I'd expect that estimating them as two separate tasks would be best. 18:41:45 rmarynch: I've just been re-reading your compiler internals documentation. I'm badly unconvinced about the presentation, but it's helped to clarify some things for me. 18:42:24 nyef: I am glad to hear that:) What is bad with the presentation? 18:44:22 Blkt` [~user@net-188-153-223-94.cust.dsl.teletu.it] has joined #sbcl 18:44:57 It seems to be more an explanation of reverse-engineering the compiler than a presentation of the principles and models by which the compiler operates. 18:46:06 nyef: this is because I didn't have the "big picture" myself. I just wrote the paragraphs along with the movement through the sources 18:46:12 -!- Blkt [~user@net-93-151-248-118.cust.dsl.teletu.it] has quit [Ping timeout: 276 seconds] 18:46:20 That's about what I expected, yes. 18:47:40 nyef: BTW, do you know whether Win32 threads from Dmitry K will be merged in the near future? 18:48:29 I don't know, but I doubt it. 18:48:48 you do not like pthreads layer? 18:49:20 Indeed not. And even granting the pthreads layer, what on /earth/ is up with the signal emulation? 18:49:41 *rmarynch* sighs 18:50:43 At the same time, I don't have a windows box for serious hacking on anymore. 18:51:10 Blkt`` [~user@net-188-153-220-217.cust.dsl.teletu.it] has joined #sbcl 18:52:34 -!- Blkt` [~user@net-188-153-223-94.cust.dsl.teletu.it] has quit [Ping timeout: 250 seconds] 19:05:56 rpg [~rpg@mpls.sift.info] has joined #sbcl 19:07:38 One more question. Is there any article about CL tree shakers design? Is GC the only mandatory part of the smallest CL image? For example, when I only have an application which does (print "hi") ? 19:10:04 -!- flip214 [~marek@unaffiliated/flip214] has quit [Quit: Leaving] 19:11:20 GC isn't mandatory for CL at all 19:27:32 It's an implicit requirement for indefinite operation on a finite machine, but not part of the language specification, and some applications may not require it at all. 19:29:02 There are probably articles out there about tree shakers, but I don't know of any offhand. 19:31:17 The thing is that Win64 port and three shaker are the only things which stop my managers from using SBCL into production. I am trying to tell them that our company should fund SBCL development, but they do not think so:) 19:31:29 "tree shaker" :) 19:31:44 -!- ASau [~user@95-27-211-235.broadband.corbina.ru] has quit [Remote host closed the connection] 19:32:13 ASau [~user@93-80-210-26.broadband.corbina.ru] has joined #sbcl 19:32:33 There's a proof-of-concept tree shaker out there already. 19:33:16 yep, by jsnell. But I miss the theory basis to finish it (and the free time too) 19:33:50 It's basically a specialized application of GC. 19:36:29 Blkt``` [~user@net-93-151-249-215.cust.dsl.teletu.it] has joined #sbcl 19:38:20 -!- Blkt`` [~user@net-188-153-220-217.cust.dsl.teletu.it] has quit [Ping timeout: 276 seconds] 19:50:38 -!- rmarynch [~roman@88.135.194.233] has quit [Quit: Leaving] 20:03:29 slyrus_ [~chatzilla@173-228-44-88.dsl.static.sonic.net] has joined #sbcl 20:19:07 -!- rpg [~rpg@mpls.sift.info] has quit [Quit: rpg] 20:29:14 -!- hargettp [~hargettp@dhcp-162.mirrorimage.net] has left #sbcl 20:41:47 -!- Blkt``` is now known as Blkt 21:19:56 -!- Blkt [~user@net-93-151-249-215.cust.dsl.teletu.it] has quit [Ping timeout: 265 seconds] 22:47:23 After having written the easy half of a tree shaker, the problems with small cl are CLOS (MOP, really), redefinitions and error handling 22:47:43 give that up, and the rest isn't too bad to patch out. 22:49:51 So, I think I might have an angle on being able to use specialized storage for value cells in d-x closures. 22:50:25 Which would mean that closing over floats or word-sized integers would suck even less. 22:59:08 tcr1 [~tcr@213.55.131.17] has joined #sbcl 23:03:13 -!- tcr1 [~tcr@213.55.131.17] has quit [Read error: Connection reset by peer] 23:13:03 tcr1 [~tcr@213.55.131.17] has joined #sbcl 23:14:36 hargettp [~hargettp@pool-71-184-191-214.bstnma.east.verizon.net] has joined #sbcl 23:22:55 -!- hargettp [~hargettp@pool-71-184-191-214.bstnma.east.verizon.net] has quit [Quit: Linkinus - http://linkinus.com] 23:26:09 -!- tcr1 [~tcr@213.55.131.17] has quit [Read error: Connection reset by peer] 23:48:39 hargettp [~hargettp@96.237.131.147] has joined #sbcl