#pypy IRC log for Saturday, 2011-08-13

lac_ (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) joined #pypy.00:11
lac (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) left irc: Ping timeout: 260 seconds00:12
voidspace (~anonymous@python/psf/voidspace) left irc: Quit: voidspace00:15
zk (~is@whatit.is) left irc: Read error: Connection reset by peer00:28
zk (~is@whatit.is) joined #pypy.00:30
etrepum (~bob@accessnat4.mochimedia.net) left irc: Ping timeout: 250 seconds00:36
bbot23Success: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/2800:41
kennethreitz (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) left irc: Quit: Computer has gone to sleep.00:41
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) left irc: Ping timeout: 258 seconds00:43
linq (~ident@CPE00259c448466-CM00195eda813e.cpe.net.cable.rogers.com) left irc: Ping timeout: 240 seconds00:45
kennethreitz (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) joined #pypy.00:50
mat^2 (~mathias@ joined #pypy.00:55
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/170801:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/41401:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/51101:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/17101:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/93901:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-stackless-app-level-linux-x86-32/builds/98801:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/70101:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/132801:03
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/58001:03
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-stackless-app-level-freebsd-7-x86-64/builds/45001:03
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-stackless-app-level-freebsd-7-x86-64/builds/45001:03
fenrrir (~fenrrir@ joined #pypy.01:04
derdon (~derdon@p5DE8983D.dip.t-dialin.net) left irc: Remote host closed the connection01:12
etrepum (~bob@75-101-96-144.dsl.static.sonic.net) joined #pypy.01:14
Sho_ (~EHS1@kde/hein) left irc: Quit: Stop leaking memory like it's going out of fashion.01:16
bob489json seems to run very slow in pypy01:38
zooko (~user@97-118-88-117.hlrn.qwest.net) joined #pypy.01:41
exarkunbob489: Grab jitviewer and inspect some traces :)01:42
rokujyouhitoma (~rokujyouh@p3063-ipbf1706marunouchi.tokyo.ocn.ne.jp) left irc: Ping timeout: 240 seconds01:42
Alex_Gaynorit's known, all of those are C-extensions01:42
Alex_Gaynorwe're running pure python01:42
Alex_Gaynorif you check out the github repo for simplejson there's a branch that's optimized for pypy01:42
ThantiK (~User@c-76-29-157-30.hsd1.fl.comcast.net) joined #pypy.01:46
ThantiKAwesome work on Quora Alex.  Grats.01:46
ThantiK (User@c-76-29-157-30.hsd1.fl.comcast.net) left #pypy ("Leaving.").01:46
MostAwesomeDudeAlex_Gaynor: What did you do OOB that's got everybody congratulating you? :301:46
MostAwesomeDudeVery cool; good work. :301:47
rokujyouhitoma (~rokujyouh@p3063-ipbf1706marunouchi.tokyo.ocn.ne.jp) joined #pypy.01:50
rokujyouhitoma (~rokujyouh@p3063-ipbf1706marunouchi.tokyo.ocn.ne.jp) left irc: Read error: Connection reset by peer01:50
rokujyouhitoma (~rokujyouh@240f:11:fa84:1:21f:f3ff:fed7:c020) joined #pypy.01:50
zooko (~user@97-118-88-117.hlrn.qwest.net) left irc: Ping timeout: 276 seconds01:52
fenrrir (~fenrrir@ left irc: Ping timeout: 250 seconds01:55
RhyoliteAlex_Gaynor: what's your secret plan for tuples?02:05
gorakhargosh (~user@ left irc: Quit: ERC Version 5.3 (IRC client for Emacs)02:07
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/170802:07
whitelynx (~whitelynx@li117-47.members.linode.com) left irc: Quit: Ex-Chat02:10
rokujyouhitoma (~rokujyouh@240f:11:fa84:1:21f:f3ff:fed7:c020) left irc: Quit: rokujyouhitoma02:12
JaRoel|4D (~jaroel_4d@sink.jaroel.nl) left irc: Remote host closed the connection02:30
zkgod pypy is so fast02:31
zkis the psycopg2 port to rpython the recommended way to work with postgresql?02:34
Da_Blitzthat is one way02:36
Da_Blitzthere is a slower all python postgres libarary as well02:37
Da_Blitzpg8000 that i have used if you dont want to retranslate pypyp02:37
zkto use the rpython port i'd have to use it with the version of pypy that's translated i guess?02:37
pjenveyzk - you'd have to build pypy with that module enabled, yea02:38
asenchiThere is also this which is translated from the rpython: https://github.com/mvantellingen/psycopg2-ctypes02:38
pjenveyAlex wrote the psycopg port, but he's hoping to kill it off actually02:38
pjenveyfor that one02:38
zkfor the ctypes version?02:38
pjenvey(which I was looking for, thanks asenchi)02:38
zkah thanks02:38
DasIch_ (~DasIch@p4FFDF0F7.dip.t-dialin.net) joined #pypy.02:38
asenchiyeah, i think it needs some work, but it's close02:39
pjenveyasenchi - do you know if that one's complete yet02:39
zkvery cool, i'll have to check it out then02:39
DasIch (~DasIch@p3E990D4C.dip.t-dialin.net) left irc: Read error: Operation timed out02:39
mwhudson (~mwh@linaro/mwhudson) joined #pypy.02:45
jared (~jared@ joined #pypy.02:47
Nick change: jared -> Guest1759702:48
zooko (~user@97-118-88-117.hlrn.qwest.net) joined #pypy.02:52
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-stackless-app-level-linux-x86-32/builds/98803:02
lucian_ (~lucian@78-86-217-168.zone2.bethere.co.uk) left irc: Ping timeout: 264 seconds03:07
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/132803:08
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/70103:11
mwhudson (~mwh@linaro/mwhudson) left irc: Ping timeout: 258 seconds03:12
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/93903:31
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/51103:37
Nick change: Gulaway -> Gulopine03:37
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/41403:37
zooko (user@97-118-88-117.hlrn.qwest.net) left #pypy ("#tahoe-lafs").03:40
davisagli (~davisagli@davisagli.com) left irc: Excess Flood03:44
davisagli (~davisagli@davisagli.com) joined #pypy.03:46
rokujyouhitoma (~rokujyouh@EM1-113-92-94.pool.e-mobile.ne.jp) joined #pypy.03:49
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) joined #pypy.03:56
Nick change: Gulopine -> Gulaway03:56
aat (~aat@cpe-72-225-174-173.nyc.res.rr.com) joined #pypy.03:58
aat (~aat@cpe-72-225-174-173.nyc.res.rr.com) left irc: Quit: Computer has gone to sleep.04:22
rokujyouhitoma (~rokujyouh@EM1-113-92-94.pool.e-mobile.ne.jp) left irc: Quit: rokujyouhitoma04:30
davisagli (~davisagli@davisagli.com) left irc: Excess Flood04:33
davisagli (~davisagli@ joined #pypy.04:33
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/58004:42
rokujyouhitoma (~rokujyouh@nttkyo1008167.tkyo.nt.ftth2.ppp.infoweb.ne.jp) joined #pypy.04:44
Niedar (bleh@ip68-99-166-222.hr.hr.cox.net) left #pypy.04:49
Niedar (~bleh@ip68-99-166-222.hr.hr.cox.net) joined #pypy.04:50
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/17105:31
gutworth (~benjamin@64-131-13-109.usfamily.net) left irc: Ping timeout: 240 seconds05:41
jack-e (~jack-e@124-171-86-33.dyn.iinet.net.au) joined #pypy.05:42
lac_ (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) left irc: Remote host closed the connection05:42
XJDHDR (c4d280b3@gateway/web/freenode/ip. joined #pypy.05:42
message144 (~message14@cpe-75-83-155-145.socal.res.rr.com) joined #pypy.05:49
message144Hi, on the pypy bitbucket hg, there is a release-1.5 tag, and on the github page, there is a 1.5-alpha tag. Are these different?05:50
Da_Blitzbitbucket is correct05:51
Da_BlitzIIRC the github page is ot offical05:51
message144Da_Blitz, ok thanks05:55
lac (~quassel@c-c4c4e055.1321-1-64736c11.cust.bredbandsbolaget.se) joined #pypy.06:20
message144 (~message14@cpe-75-83-155-145.socal.res.rr.com) left irc: Ping timeout: 260 seconds06:25
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) left irc: Ping timeout: 250 seconds06:57
fijal (~fijal@ joined #pypy.07:09
arigato (~arigo@ joined #pypy.07:27
fijalmorning armin07:35
XJDHDRHello everyone. Is it possible to use wxPython with PyPy v1.5?07:35
fijalI think so07:36
fijalronny: ping07:37
fijalarigato: I think we're in a way good to go for 1.607:39
fijalin a way because we still have to have windows nightly for one07:40
arigatono, I'm fixing the asmgcc bug07:41
arigatoit's really a random bug that may show up anywhere, rarely07:41
arigatoso it's good to have the fix in 1.6 too07:41
fijalthat might also be worth looking07:41
arigatoor so I'm saying, but it's not really going to be a fix07:42
arigatomore like an obscure workaround that works in the common case (of this rare case)07:42
fijalthen we have the rare case of the rare case07:43
arigatowe suppose that the rare case of the rare case is not going to show up07:43
fijalso maybe shadowstack is not such a bad idea...07:43
fijalI looked at https://bugs.pypy.org/issue718 yesterday07:43
arigatohttps://bugs.pypy.org/issue799: yes, we should at least look for a bit today07:44
fijalI might be mostly off for today07:44
fijaland few more days (not sure how many)07:44
fijal[yes, I'm on holiday]07:44
fijalbut I guess sometimes mid-next-week we should get 1.6 out07:44
fijalunless more impotant tracker issues will show up07:45
__class__ (~class@99-105-57-153.lightspeed.sntcca.sbcglobal.net) left irc: Ping timeout: 258 seconds07:46
ericflo (~ericflo@c-98-234-217-1.hsd1.ca.comcast.net) left irc: Quit: ericflo07:51
fijalhm, test_posix isn't fixed apparently07:53
fijalwindows situation is not cool either07:56
fijal8h in and it's still running lib-python tests07:56
lachi all07:58
fijalhi laura07:58
XJDHDRWhen I install wxPython v2.8.12.1 unicode for python 2.7 into PyPy's site-packages folder then run a program that uses it, wxPython throws an import error. ImportError: No module named _core_08:20
fijal (~fijal@ left irc: Remote host closed the connection08:26
antocuni (~antocuni@host225-126-dynamic.14-87-r.retail.telecomitalia.it) joined #pypy.08:31
jack-e (~jack-e@124-171-86-33.dyn.iinet.net.au) left irc: Quit: jack-e08:38
iratsu (~alang@host109-150-241-124.range109-150.btcentralplus.com) joined #pypy.08:45
kkris (~kris@93-82-45-149.adsl.highway.telekom.at) joined #pypy.08:51
JaRoel|4D (~jaroel_4d@2001:980:7133:1:14d3:accc:7dc6:ccfe) joined #pypy.08:53
arigatohakanardo__: hopefully fixed, but with a hack to trackgcroot.py that can be defeated in rare cases09:15
arigatoas I said above, it's the rare case of the rare case, so hopefully good enough for now09:16
arigatoa bit hard to do really better09:16
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy.09:20
hakanardoarigato: I'm sure it is 09:24
hakanardodid you check in?09:27
arigatonot yet09:27
arigatoI'm running tests by trying to reproduce the bug09:28
rndbit (~rndbit@static. left irc: Quit: rand() & 109:28
arigatoI should fail09:28
arigatoI'm also adding a cleaner way to crash09:28
arigatowhen the root finder finds a pointer, we check that it's at least not some small integer09:29
arigatoif it is we can crash cleanly09:29
Alex_Gaynorthis won't work if pypy is used to run the kernel, I suppose :D09:29
hakanardook, so how do you define small?09:29
arigatodunno, but it's not that important09:30
arigatopointers are never < 4096, for example09:30
arigatoand integers are often09:30
Alex_Gaynorand they must be multiples of sizeof(void*)09:30
arigatowell I'm not going there09:30
arigatoah, you can have tagged pointers too09:31
Alex_Gaynordoes trackgcroot work with them?09:31
arigato...but they should not be reported by the rootfinder09:31
arigatoyes, trackgcroot only finds registers that will contain gc pointers, so it should work09:31
arigatomoreover it's fine to ignore registers that contain prebuilt pointers09:32
arigatoif you produce a register by doing some arithmetic, it's not a gc pointer09:33
arigatoif that was making a tagged pointer, then too bad, it's not reported as containing a gc pointer, which is fine09:33
arigatoif you merge the path with another path that can also contain a real gc pointer, then it's going to be reported09:33
arigatowell it should kind of work, but maybe there are corner cases again09:33
Alex_Gaynorthis is stuff that's already implemented?09:34
oal (~olav@5.79-160-122.customer.lyse.net) joined #pypy.09:37
kenaan12arigo default 112ffe3d85f32e 15/pypy/translator/c/gcc/: Test and (90% of a) fix.09:38
kenaan12arigo default 119741bff7ba67 15/pypy/: Add a debugging check earlier, to make it easier to track errors in asmgcc.09:38
rndbit (~rndbit@static. joined #pypy.09:46
hakanardo (hakan@h59ec009e.seluldx.dyn.perspektivbredband.net) left #pypy.09:48
arigato? :-)09:50
ronnyhmm, the biy could have some extra tools09:50
ronnywell, i fixed the bugtracker duplicate show thing09:50
arigatoah, good09:50
ronnyturns out i searched for it in the completely wrong place09:51
rokujyouhitoma (~rokujyouh@nttkyo1008167.tkyo.nt.ftth2.ppp.infoweb.ne.jp) left irc: Quit: rokujyouhitoma10:07
arigatotests of trackgcroot passed on tannit10:12
ojii (~ojii@ joined #pypy.10:12
kenaan12arigo release-1.6.x 116fa38d806a56 15/pypy/translator/c/gcc/: Test and (90% of a) fix.10:21
arigatoantocuni: replying to your suggestion here:10:37
arigatointeresting too, at least for a hackish solution like this:10:37
arigatoit generates C code and then you have to edit it and paste it in some more complete C project10:37
antocuniyes, something like that10:38
antocuniI was also thinking of something which tells you "ok, but I can't compile because you miss this bridge"10:38
antocuniand then what you do is to improve your tests to cover also that case10:38
antocuniso the JIT sees all the "relevant" bridges and can emit C code for them10:39
arigatohum, there are a lot of guards that are never taken10:39
arigatowhat about them?10:39
antocuninot sure10:40
antocunisome of them are guards that can be "statically removed" in the context of the C code10:40
antocunie.g., guards on class versioning10:40
antocuniacutally, I did not relly think deeply about it10:41
antocuniit's just something that popped out of my mind10:41
Action: antocuni --> lunch10:42
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) joined #pypy.11:25
arigato (~arigo@ left irc: Quit: See you11:27
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) left irc: Ping timeout: 264 seconds11:31
gorakhargosh (~user@ joined #pypy.11:32
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 246 seconds11:36
iratsu (~alang@host109-150-241-124.range109-150.btcentralplus.com) left irc: Ping timeout: 250 seconds11:43
kenaan12hakanardo jit-short_from_state 11407843026514 15/: hg merge default11:53
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/512 [12hakanardo, jit-short_from_state]11:54
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) joined #pypy.12:05
ita (~ita@kde/developer/tnagy) joined #pypy.12:05
ita^o^  hi all !!! ^o^ :-)12:05
Da_Blitz nyo12:06
itapypy on linux64 is totally foobar on the nightly build here, does anyone want a testcase?12:07
Da_Blitznot sure, try lodging a bug on the bug tracker12:07
Da_Blitzit may however already be fixed, have you tried the most recent nightly12:07
itayes, and several other nightly builds, and the release 1.5 (which leaks file descriptors so much that it is not usable ... here at least)12:08
Da_BlitzFD leaages are normmaly programmer error12:09
Da_Blitzit uses a proper GC, not ref counting12:09
Da_Blitzso it means you are not closing files corectly12:09
itahaha no, the files *are* being closed properly12:10
Da_Blitzyou are explicitly calling .close?12:10
itai saw a report on the tracker for it by the way12:10
Da_Blitzah np12:10
itatry-finally everywhere12:10
Da_Blitzthen a bug report for that would be apreciated12:10
antocuniita: indeed, if you call .close and still leak FDs, then we would like a bug report12:13
antocunialthough it's a bit weird, most of us use linux64 and we never noticed such problems12:14
fenrrir (~fenrrir@ joined #pypy.12:19
CIA-5503ita 07roundup * 10#828/subprocess on nightly builds, linux 64: 12:21
CIA-55[new] I have tried several nightly builds, and it seems they have been broken for a12:21
CIA-55while. At least, I cannot get the behaviour of cPython with ... * 14https://bugs.pypy.org/issue82812:21
itahaha there it is12:21
antocuniita: any chance to have a smaller test case?12:22
itaantocuni: no12:22
ita(maybe, i am searching)12:23
itacould you confirm that you are getting a different behaviour? this way i wont have to search for the bug if the problem comes from external libraries12:25
antocuniita: the build fails but not for leaks of FDs, it seems: http://paste.pocoo.org/show/457699/12:25
antocuniuhm, but with CPython it works,, indeed12:26
antocunianyway, this program seems impossible to debug12:26
itaantocuni: at least i know that the problem is real12:26
antocunicould we have a version which has plain python files, instead of everything inside "waf"?12:27
itaantocuni: look in the folder .waf-xyz...12:27
Action: ita is trying to reproduce the problem with subprocess12:27
ita.oO(it happens even without threading at least)12:30
Da_Blitzwhats the binary stuff at the end of the file12:30
itaDa_Blitz: something that makes redistribution easier (works on python 2, python 3, pypy, ironpython and jython unmodified)12:31
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/512 [12hakanardo, jit-short_from_state]12:32
Da_Blitzita: got a link to whatever it is?12:33
CIA-5503guibou 07roundup * 10#826/Decimal seems slower on pypy than cpython: 12:33
CIA-55justin: your are right about the fact that this code should be write12:33
CIA-55differently. When I had the float overflow in this function, I do not though ... * 14https://bugs.pypy.org/issue82612:33
itaDa_Blitz: http://code.google.com/p/waf/12:33
antocuni (~antocuni@host225-126-dynamic.14-87-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds12:34
derdon (~derdon@p5DE897D1.dip.t-dialin.net) joined #pypy.12:42
antocuni (~antocuni@host11-57-dynamic.56-82-r.retail.telecomitalia.it) joined #pypy.12:47
LoganLK (~LoganLK@ left irc: Quit: LoganLK12:49
itabah, my testcase eats an exception, but it is the same problem : too many open files12:50
Action: ThomasWaldmann plays with pypy nightly and moin212:55
ThomasWaldmannita: "with" might also be an option12:56
bob489 (~Bob@ool-43539e48.dyn.optonline.net) left irc: Quit: Leaving.12:56
Action: ThomasWaldmann had also a file leaks issue12:56
ThomasWaldmannthat's incomplete13:01
itaThomasWaldmann: sure, i am trying to complete it13:01
Da_Blitzhmm, think i found it13:06
itaDa_Blitz: ?13:07
neppord (~neppord@251dsl84-sala.soderhamn-net.com) joined #pypy.13:07
Da_Blitzi think subprocess relies on cpython's refcounting to have files closed13:10
Da_Blitzhence if you do Popen() enough times without a gc collectio you will run out of FD's13:10
Da_Blitzi assume it is running a subprocess once per file or somthing like that13:11
itasince to be the case, there is only one point in my app where files are read, and they remain visible in "lsof"13:11
Da_Blitzita: i dont think the problem is with your app13:11
Da_Blitzbut the subprocess lib itself13:11
neppordantocuni:  is there any known bugs with the jit viewer?13:12
antocunineppord: "it usually works"13:13
neppordokey, cause i get a simple keyerror in the parser, in the parer_log_count13:15
antocunineppord: are you sure to have a very recent version of pypy and an updated checkout?13:15
neppordnew pypy-lib or new interpreter?13:16
neppordhmmm, well i have te newest on the download page for the mac, but i guess thats not enught.13:16
antocunino, you need a nightly build13:17
neppordok thx13:17
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) left irc: Ping timeout: 250 seconds13:18
ThomasWaldmanncannot find gc roots!13:18
Action: ThomasWaldmann crashed it, again :)13:18
Da_BlitzThomasWaldmann: have you looked under the couch13:18
ThomasWaldmannjust used moin2 and fired ab -n 1000 -c 10 at it13:19
arigato (~arigo@ joined #pypy.13:22
ThomasWaldmannarigato: moin :)13:26
ThomasWaldmann(15:19) < ThomasWal> cannot find gc roots!13:26
ThomasWaldmann(15:20) < ThomasWal> just used moin2 and fired ab -n 1000 -c 10 at it13:26
Da_Blitzita: yep, think i found it13:26
Da_Blitzexec_command in Context.py13:26
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1709 [12arigo]13:27
Da_Blitzdefinetly looks like subprocess assuming refcounting13:27
itaDa_Blitz: so, a bug in subprocess affecting only pypy?13:27
Da_Blitzwell not exactly a bug13:27
Da_Blitzbut yes13:27
ronnyThomasWaldmann: i wonder if it works better with shadowstack?13:28
itaDa_Blitz: so, is it a one-liner fix or something bigger? :-)13:28
Da_Blitzbasically the issue is that subprocess never closes files13:29
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/162 [12fijal, release-1.6.x]13:29
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/16313:29
Da_Blitzit expects the python implementation to do it for it13:29
Da_Blitzwhich is fine if you have refcounting13:29
ThomasWaldmannronny: no idea, i just used the nightly 32bit binary13:30
ThomasWaldmanntw@x300:~$ ab -n 1000
Da_Blitzbut if you delay collecting garbage a bit then its posible to open enough files  to run out of FD's13:30
Da_Blitzbefore you release some13:30
ThomasWaldmannafter 159 requests, it crashes13:30
antocuni (~antocuni@host11-57-dynamic.56-82-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds13:30
arigatoThomasWaldmann: to debug it we would need either the exact exact exact same configuration as you (including set of modules), or access to your machine :-/13:31
ramusara_ (~ramusara@ joined #pypy.13:31
gutworth (~benjamin@64-131-13-109.usfamily.net) joined #pypy.13:32
JaRoel|4D (~jaroel_4d@2001:980:7133:1:14d3:accc:7dc6:ccfe) left irc: Remote host closed the connection13:32
ThomasWaldmannarigato: we have a quickinstall script in our repo that just installs everything to a VE13:32
ThomasWaldmann(and the script can be modified easily to use pypy as python)13:33
fenrrir (~fenrrir@ left irc: Quit: Saindo13:33
arigatoI wouldn't mind a step-by-step guide that includes urls13:34
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/581 [12hakanardo, jit-short_from_state]13:42
hakanardo (~hakan@h59ec009e.seluldx.dyn.perspektivbredband.net) joined #pypy.13:42
neppord (neppord@251dsl84-sala.soderhamn-net.com) left #pypy.13:49
ronnyarigato: hg clone http://bitbucket.org/thomaswaldmann/moin-2.0;cd moin-2.0;./quickinstall13:50
arigatoronny: thanks, Thomas gave me more detailled instructions already13:50
ronnyThomasWaldmann: you guys have ben eager in the xstatic department13:51
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/415 [12hakanardo, jit-short_from_state]13:52
ThomasWaldmannronny: just what we needed ourselves13:54
ThomasWaldmannronny: i am sure the api can be improved. need feedback.13:54
timonatorŸ/win 1714:07
jerithYou win 17. Would you like to enter again?14:07
CIA-5503dablitz 07roundup * 10#828/subprocess on nightly builds, linux 64: 14:07
CIA-55[chatting] Subprocess appears to leak FD's if stdin, stdout or stderr are opened with14:07
CIA-55subprocess.PIPE (Popen("dsa", stdin=PIPE))14:07
CIA-55waf uses subp ... * 14https://bugs.pypy.org/issue82814:07
CIA-5503npforce 07roundup * 10#823/osx nightly build is not 'nightly': [resolved] I see that the build is up to date now. * 14https://bugs.pypy.org/issue82314:08
jaimechen (~jaime@ joined #pypy.14:12
itaThomasWaldmann: the leak *is* in subprocess :P14:13
tav (~tav@host-2-96-37-56.as13285.net) left irc: Ping timeout: 250 seconds14:13
itaDa_Blitz: i had finally made a testcase, but yours is much more compact14:13
Da_Blitzit happens14:13
arigatoThomasWaldmann: thanks for insisting with reporting the bug :-)  it looks like real bug(s)14:18
xorAxAx 14:18
jaimechenHi,  if I want to run a normal web application like django with  pypy-c-sandbox, how much work need to be done?  thanks14:20
ThomasWaldmannarigato: thanks for debugging :)14:20
arigatoif anything I'd blame ctypes for sqlite3 giving Warning: One and only one statement required: '\xf20\xed\xc4\xf04\xcb[\xf2\xb8\xd6\xe7\xf0\x80\n\\\xf2\x84\xed\xc4\xf0\xf8\xfd[\xf2h\xed\xc4\xf0'14:20
ThomasWaldmannita: ah, ok14:20
arigatojaimechen: what is your goal?14:21
jaimechensomething like app hosting14:21
arigatopypy-sandbox needs someone to really care for it14:21
jaimechenI'm afraid I have to write too many do_ll_xx_ ll_ functions14:21
arigatoyes, precisely14:22
Action: Da_Blitz has achived somthing simmilar with linux namespaces and a python module14:22
arigatojaimechen: I can't tell in advance, but indeed I fear that there are a lot of ll_xxx functions that must be emulated14:22
arigatobefore something like django works14:23
jaimechenIs there a list that I can reference? 14:23
arigatono: it's all functions doing external calls in the PyPy source14:23
arigatoit's generated automatically and exhaustively14:23
linq (~ident@CPE00259c448466-CM00195eda813e.cpe.net.cable.rogers.com) joined #pypy.14:24
arigatoyou have no chance to cover them all, but even enough to get django working is unclear14:24
arigato(in my opinion)14:25
jaimechenthey are generally c lib or  sys functions calls, right?14:25
arigatoyou can find most of them by grepping for "llexternal(" in PyPy14:25
Da_Blitzactually i should convert it to work with pypy as well14:26
jaimechenfor example, if call X  exists both in module m1, m2, do I need to write two do_ll functions like do_ll_m1_ll, do_ll_m2_ll, or just one? 14:27
jaimechenif the latter, it's a little bit easier to cover the whole list14:27
jaimechenno need to find in all python modules14:28
arigatojust one, but now I'm unsure under which name14:28
arigatothe name like "ll_os.ll_os_execv" is written explicitly in rpython/module/ll_os.py14:29
arigatobut that uses extdef() instead of llexternal()14:29
arigatoI'm unsure which name gets picked for llexternal(), which are the more modern way of declaring external functions14:30
arigatowell try I suppose14:30
jaimechenthanks for the info:)14:30
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/1709 [12arigo]14:39
jaimechen (jaime@ left #pypy.14:41
Da_Blitzi guess they dont use linux then14:42
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) joined #pypy.14:46
rokujyouhitoma (~rokujyouh@KD121107219133.ppp-bb.dion.ne.jp) joined #pypy.14:56
Nick change: DasIch_ -> DasIch14:56
rguillebert (~hardshoot@fec76-1-78-234-124-105.fbx.proxad.net) left irc: Quit: WeeChat 0.3.515:02
rguillebert (~hardshoot@fec76-1-78-234-124-105.fbx.proxad.net) joined #pypy.15:02
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) joined #pypy.15:11
arigatoso, even a simple string concatenation can occasionally get a segfault15:15
arigatolooks like yet another asmgcc bug15:16
arigatoI think that's the primary cause for the _sqlite3 error15:18
Nick change: Guest17597 -> jaredw15:19
arigatoit would build a string representing a sql statement, and get occasionally some nonsense string instead15:19
gutworth (~benjamin@64-131-13-109.usfamily.net) left irc: Quit: leaving15:27
CIA-5503ita 07roundup * 10#828/subprocess on nightly builds, linux 64: 15:31
CIA-55Search no more: the problem was aggravated by a real file descriptor leak, the15:31
CIA-55subprocess problem is not going to show up in practice * 14https://bugs.pypy.org/issue82815:31
kkris1 (~kris@80-123-47-216.adsl.highway.telekom.at) joined #pypy.15:31
itagrrr, i have found a segfault with pypy-jit, but it does not show up when running under gdb (does not happen on pypy-nojit)15:33
kkris (~kris@93-82-45-149.adsl.highway.telekom.at) left irc: Ping timeout: 258 seconds15:33
timonatorcan you let it create a coredump and analyse that?15:39
asabil (~asabil@ left irc: Quit: Leaving15:39
Shanita (~John@osbk-4d08b018.pool.mediaWays.net) left irc: Read error: Connection reset by peer15:39
italooks like it does not generate any core file15:40
asabil (~asabil@ joined #pypy.15:42
Moku (~John@osbk-4d08b018.pool.mediaWays.net) joined #pypy.15:44
nopf (~froschi@ left irc: Read error: Connection reset by peer15:44
nopf (~froschi@ joined #pypy.15:44
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/581 [12hakanardo, jit-short_from_state]15:45
[mat^2] (~mathias@ joined #pypy.15:55
mat^2 (~mathias@ left irc: Ping timeout: 258 seconds15:56
arigatoita: I just fixed two of them today15:57
arigatoso maybe with tomorrow's nightly build it will work better for you too15:58
CIA-5503dablitz 07roundup * 10#828/subprocess on nightly builds, linux 64: 15:59
CIA-55Attached is a quick patch to add my close on wait idea, .communicate and some of15:59
CIA-55the other functions call self.wait and so this should handle tho ... * 14https://bugs.pypy.org/issue82815:59
Da_Blitz hmm, i am tempted to write a quick patch to turn Popen into a context manager to automatically clean up the fd's16:00
Da_Blitzi suppose i should send that upstream instead of to pypy16:00
Sho_ (~EHS1@kde/hein) joined #pypy.16:01
aat (~aat@cpe-72-225-174-173.nyc.res.rr.com) joined #pypy.16:07
antocuni (~antocuni@host11-57-dynamic.56-82-r.retail.telecomitalia.it) joined #pypy.16:09
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy.16:11
XJDHDR (c4d280b3@gateway/web/freenode/ip. left irc: Quit: Page closed16:19
tav (~tav@host-92-20-5-153.as13285.net) joined #pypy.16:19
fenrrir (~fenrrir@ joined #pypy.16:35
rokujyouhitoma (~rokujyouh@KD121107219133.ppp-bb.dion.ne.jp) left irc: Quit: rokujyouhitoma16:39
rokujyouhitoma (~rokujyouh@KD121107219133.ppp-bb.dion.ne.jp) joined #pypy.16:40
rokujyouhitoma (~rokujyouh@KD121107219133.ppp-bb.dion.ne.jp) left irc: Client Quit16:41
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/415 [12hakanardo, jit-short_from_state]16:45
danielnorberg (~daniel@c83-251-114-201.bredband.comhem.se) left irc: Ping timeout: 250 seconds16:51
arigatoI just added another bug by fixing the asmgcc but earlier in the day16:52
danielnorberg (~daniel@c83-251-114-201.bredband.comhem.se) joined #pypy.16:52
voidspace (~voidspace@python/psf/voidspace) joined #pypy.16:54
voidspace (~voidspace@python/psf/voidspace) left irc: Client Quit16:56
ronnyarigato: how much is missing for shadowstack?17:00
arigatoof performance17:01
ronnyhow hard will getting there be?17:01
arigatoI don't know17:01
ronnyand must it be as fast? given the other speedips like jit_from_state17:01
antocuniarigato: http://speed.pypy.org/comparison/?exe=1%2BL%2Bdefault%2C1%2BL%2Bshadowstack-perf&ben=1%2C27%2C2%2C25%2C3%2C4%2C5%2C22%2C6%2C7%2C8%2C23%2C24%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C28%2C30%2C32%2C29%2C33%2C31%2C17%2C18%2C19%2C20&env=1&hor=false&bas=1%2BL%2Bdefault&chart=normal+bars17:02
antocunithis is the comparison between pypy-c-jit-latest and the shadowstack-perf branch17:02
arigatook so not that clear, but there are bad cases17:03
antocunihowever, it's a bit hard to extract numbers from the graph17:03
antocuniI think that stackcheck is just very noisy17:03
arigatowe should just ignore it, it runs for ~10 seconds and I suppose that it can gets much slower or faster depending on whether during these 10 seconds there is a full collection or two, for example17:04
arigatoand maybe the algorithm is actually not deterministic and can take longer or shorter17:05
arigato(don't remember but wouldn't be surprized)17:05
antocunifor the other benchmarks, the difference seems to be <10%, apart for go, pyflate-fast and maybe telco17:05
guibouantocuni: what is the purpose of the branch shadowstack ?17:06
arigatono it's raytrace-simple I think, not pyflate-fast17:06
antocuniah yes, indeed17:07
antocuniwell, it's clear that the current UI is not enough :-/17:07
antocunibut having support for branches is already cool17:07
antocuniguibou: to improve the "shadowstack" gc root finder17:07
antocuniwhich is more portable and less buggy than the alternative, asmgcc17:08
antocunibut slightly slower17:08
kenaan12arigo default 11e650c095b599 15/pypy/rpython/lltypesystem/rstr.py: Baaaah.  Missing keepalive_until_here.17:09
kenaan12arigo default 119d91006bdc19 15/pypy/rpython/lltypesystem/rffi.py: Another keepalive_until_here(); this one is probably needed too in theory, but I don't have a particular test case ...17:09
kenaan12arigo release-1.6.x 11e79d6ab8a933 15/pypy/rpython/lltypesystem/rstr.py: Baaaah.  Missing keepalive_until_here.17:09
kenaan12arigo release-1.6.x 11c7329816d569 15/pypy/rpython/lltypesystem/rffi.py: Another keepalive_until_here(); this one is probably needed too in theory, but I don't have a particular test...17:09
kenaan12arigo default 11678a28dd058f 15/pypy/translator/c/gcc/: Test and fix.17:09
kenaan12arigo release-1.6.x 117ceb018da5a6 15/pypy/translator/c/gcc/: Test and fix.17:09
arigatook so I didn't solve anything except a bug added earlier this afternoon17:09
arigatonow Thomas's moin example still hits "cannot fint gc roots!"17:10
ronnyhow negelctable does the speed benefit turn in the face of breakages17:10
arigatobut it's apparently in a subprocess17:10
arigatowhat do I do know?17:11
arigatothe Python program starts a subprocess *somewhere*, it hits "cannot find gc roots!", the subprocess exits with an error, and then the parent process (which I'm gdb'ing) just exits silently17:12
arigatohow do I find out where and how the subprocess was created?17:13
Da_Blitzmonkeypatch the subprocess calls?17:13
arigatoah I can get a core dump, probably17:13
arigatohum no17:13
arigatonow I'm confused17:13
arigatoyes, I got a core dump.  just the date on this computer is strange17:14
arigatoah, I get it.  it looks like a real bug17:16
kennethreitz (~kennethre@c-24-127-96-129.hsd1.va.comcast.net) left irc: Quit: Computer has gone to sleep.17:17
ojii (~ojii@ left irc: Remote host closed the connection17:21
lucian (~lucian@78-86-217-168.zone2.bethere.co.uk) joined #pypy.17:25
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection17:28
arigatoaaaaah subtle enough17:28
arigatothere is a timing issue with multiple threads and with x86's CALL_RELEASE_GIL17:28
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) joined #pypy.17:30
rguillebert (~hardshoot@fec76-1-78-234-124-105.fbx.proxad.net) left irc: Quit: WeeChat 0.3.517:36
antocuni (~antocuni@host11-57-dynamic.56-82-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds17:38
Nick change: Gulaway -> Gulopine17:40
voidspace (~anonymous@python/psf/voidspace) joined #pypy.17:42
ordiclic (~ordiclic@leet.rootards.org) left irc: Read error: Operation timed out17:55
fenrrir (~fenrrir@ left irc: Quit: Saindo17:59
danielnorberg_ (~daniel@c83-251-114-201.bredband.comhem.se) joined #pypy.18:01
danielnorberg (~daniel@c83-251-114-201.bredband.comhem.se) left irc: Read error: Connection reset by peer18:01
Nick change: danielnorberg_ -> danielnorberg18:01
linq (ident@CPE00259c448466-CM00195eda813e.cpe.net.cable.rogers.com) left #pypy.18:09
voidspace (~anonymous@python/psf/voidspace) left irc: Quit: voidspace18:18
ojii (~ojii@62-13.3-85.cust.bluewin.ch) joined #pypy.18:20
zk (~is@whatit.is) left irc: Ping timeout: 246 seconds18:23
zk (~is@whatit.is) joined #pypy.18:24
whitelynx (~whitelynx@li117-47.members.linode.com) joined #pypy.18:27
itacpython 1m42, pypy 2m0018:36
lac  arigato: :-( :-(18:40
Action: lac and cfbolz off to dinner18:40
arigatowell, today is a much better day than just after the release 1.6 in order to find such issues18:41
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection18:41
arigatoI have to thank Thomas Waldmann, who submitted the buggy behavior again today (it may be the same as the one he already mentioned during EuroPython)18:42
danielnorberg (~daniel@c83-251-114-201.bredband.comhem.se) left irc: Quit: danielnorberg18:49
pjenveythat sounds like a fun one arigato18:54
arigatoI'm running tests, but I think it's fixed by a 10-lines change and a 10-lines comment18:55
berdario (~quassel@host46-87-dynamic.0-87-r.retail.telecomitalia.it) left irc: Ping timeout: 260 seconds19:02
whyking (~quassel@pool-108-12-201-88.prvdri.fios.verizon.net) joined #pypy.19:08
rguillebert (~hardshoot@fec76-1-78-234-124-105.fbx.proxad.net) joined #pypy.19:11
justinpeelThe RTyper is complaining about using a non-constant index in a tuple. The tuple is prebuilt and the index I'm using is an attribute of a prebuilt class. Do I have to set the attribute as an immutable field?19:13
eddd (eddward@ joined #pypy.19:15
justinpeelI tried asserting that the attribute is within the length of the tuple but the rtyper wasn't satisfied with that19:15
edddhello, anyone tried to compile pypy on mac os 10.5 ?19:16
[mat^2] (~mathias@ left irc: Ping timeout: 258 seconds19:17
edddbuild's running about an hour, and it consumed 800 MB of ram.. its wired ;]19:18
exarkun800MB wired ram?19:18
lucians/wired/weird/ ?19:19
exarkun(wired memory on OS X is memory that cannot be paged to disk)19:20
JaRoel|4d (~jaroel|4d@2001:980:7133:1:388c:c003:45ff:f8ae) joined #pypy.19:20
Fluxidfirst time building?19:20
edddweird, typo, sorry <:19:20
exarkunBuilding PyPy always takes a lot of memory.  More than 800MB, I think.19:20
edddso, i have to wait a few hours ? 19:21
lucianeddd: possibly. very long, anyway19:22
eddddang it19:22
exarkunis there more free memory available?  or has the system started using swap?19:22
exarkunonce you go into swap, translation will probably not finish in a reasonable amount of time19:22
edddi have 600 MB inactive, so it should be ok.19:24
arigatojustinpeel: if I understood correctly, it is a class-level attribute?19:32
fijal (~fijal@ joined #pypy.19:32
arigatodo you have several classes with different value of the attribute?19:32
arigatofijal: hi19:32
arigatojustinpeel: if I remember correctly and if I have in mind the right tuple, then you should probably not use a tuple at all,19:34
arigatobut just attach to the class what you'd find from reading the tuple item19:34
arigatoyou can also use a list instead of a tuple, but then it's slightly less efficient code because reading the item will be done at runtime19:35
arigato(and it doesn't work if the list is not homogenous)19:35
arigato(of course by "attach to the class what you'd find from reading the tuple item", I mean write a loop that does it for all classes, not change everything)19:41
Shanita (~John@osbk-4db16392.pool.mediaWays.net) joined #pypy.19:42
arigatowell if I remember correctly, you even have a case that can be written with nested scopes:19:42
arigatodef make_case(case_number):19:42
arigato   class Foo(Bar):19:42
arigato      def method(self):19:42
arigato           use x here19:42
arigato   x = sometuple[case_number]19:42
Moku (~John@osbk-4d08b018.pool.mediaWays.net) left irc: Ping timeout: 250 seconds19:43
Arfrever (~Arfrever@gentoo/developer/Arfrever) joined #pypy.19:43
kenaan12arigo default 113be002cc751f 15/lib_pypy/_sqlite3.py: Improve the error message in _sqlite3.  Mostly useful for debugging PyPy itself...19:48
kenaan12arigo default 11dbecbb82ebbb 15/pypy/jit/backend/x86/assembler.py: Bug fix.  Hard to test.19:48
arigatoThomasWaldmann: now moin moin seems to work; at least what you told me works fine, even after 3000 requests19:49
arigato(fwiw it goes down to 13ms per request; I don't know what you'd get with CPython on this machine)19:49
fijalarigato: is this also going for 1.6?19:50
arigatoit's just a bug fix19:50
arigatoah, forgot to transplant it, thank you19:50
kenaan12arigo release-1.6.x 119482fc70908e 15/pypy/jit/backend/x86/assembler.py: Bug fix.  Hard to test.19:51
arigatowe still have the sqlite3 bug of "one and only one statement required"19:53
fijalone alex was talking about?19:53
ThomasWaldmannarigato: you found the gc problem?19:54
arigatoThomasWaldmann: actually it was a jit problem19:55
arigatoThomasWaldmann: caused by multiple threads and calls to C functions19:55
arigatonot a very well-tested combination, from our point of view19:55
arigatofijal: yes19:56
fijalarigato: we also have the threading bug19:56
fijalI looked at it, but nothing happened other than staring with puzzlement19:57
ThomasWaldmannah, good you found it :)19:57
ThomasWaldmannbtw, moin pypy timings are about 2x slower than on cpython19:58
fijalThomasWaldmann: doing what?19:59
arigatoThomasWaldmann: right, but you need to repeat these timings, because the JIT warm-up time might be large19:59
fijal(and does it use a DB)19:59
arigatoso if you just measured the short time before it crashed, then it's likely that it's better in long runs19:59
ThomasWaldmannfijal: i just looked at moin log output. about all timings are slower, about 2x.20:00
ThomasWaldmannand i looked at timings after 1000 or 159 requests (when it died)20:01
fijalI'm not sure about moin, but twisted takes a bit more than that20:01
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy.20:01
fijal(couple seconds)20:02
fijalof consecutive hitting with requests20:02
arigatoah, I found the cause of "one and only one statement required"20:02
justinpeelarigato: 20:03
justinpeelarigato: I'm not sure that I understand20:03
ThomasWaldmannarigato: as your pypy works now, fire 10000 requests at it :)20:03
ThomasWaldmannand compare with cpython20:03
arigatowell I'm busy debugging the next issue :-)20:04
Action: ThomasWaldmann would rather wait for next nightly than torture my little machine with translating20:04
arigatojustinpeel: I don't know what your exact problem is20:05
linq3 (~ident@CPE00259c448466-CM00195eda813e.cpe.net.cable.rogers.com) joined #pypy.20:05
justinpeelI create specialized classes for each type (int32, float, etc)20:05
justinpeeland I have a tuple of those20:05
arigatouh.  how can it work even on top of CPython?20:05
arigato(sorry, talking about "one and only one statement required" again)20:06
edddlucian: compilation took all my ram i processor usage fell to 4%... any chances to finish this to the  morning ? <:20:06
justinpeelI use the dtype.num attribute to choose which class to use20:06
lucianeddd: if you're not swapping yet, it should finish at some point20:06
arigatoit looks like it's using deallocated memory, but in a way that kind of works on CPython just because20:06
arigatojustinpeel: ah, ok.  well, don't do that20:07
arigatocollect the classes in a list instead of in a tuple20:07
arigatoand it should just work20:07
justinpeelwhy a list rather than a tuple?20:07
justinpeelI thought that it was complaining earlier when I used a list like this..20:08
arigatobecause it's homogenous20:08
justinpeelokay, I'll give it a go20:08
arigatoit's possible to have a variable of type "one of these classes" in RPython20:08
edddlucian: im sure its swapping...20:08
lucianeddd: it might not translate in reasonable time, then20:09
donri (~donri@unaffiliated/dagodenhall) joined #pypy.20:09
edddwhat do you mean by reasonable?20:09
arigatoeddd: see http://pypy.org/download.html#building-from-source20:09
arigatofor infos about how much RAM you need20:09
edddi have 2gb, and os take more or less 500mb, so im screwd :<20:11
lucianeddd: less than a few days20:11
fijaleddd: 32 or 64 bit?20:11
fijaland why you  can't use nightly?20:12
lucianeddd: have you tried translating with pypy?20:12
fijal2gb should be just enough for translation on 32bit20:12
arigatofijal: it's an OS/X translation20:12
arigatowhich means the system eats tons of RAM too20:12
fijalah ok20:12
fijalyes I see20:12
fijalos x is worse20:12
lucianit's quite bad with memory for some reason, yes20:13
arigatoOS/X >> Ubuntu >> my laptop, as far as memory consumed by the OS goes :-)20:13
arigato(but I don't dare put WIndows in this list)20:13
edddeh, screw it, ill wait until i come back home and 8GB ddr3 with ubuntu.20:15
lucianeddd: that is a wise choice20:15
fijaleddd: good idea20:15
edddi spend a lot of time on mac, so its sad, that i cant test pypy here.20:16
fenrrir (~fenrrir@ joined #pypy.20:17
fijalcan't you try the nightly build?20:18
stakkars (~tismer@p5DCE36EE.dip.t-dialin.net) joined #pypy.20:18
kenaan12arigo default 11e75a624ff472 15/lib_pypy/_sqlite3.py: Fix a bug in _sqlite3.py: memory is freed too early20:18
kenaan12arigo release-1.6.x 11140dbb95dac4 15/lib_pypy/_sqlite3.py: Fix a bug in _sqlite3.py: memory is freed too early20:18
edddsomeone could make nightly build for mac os x8620:19
stakkars (~tismer@p5DCE36EE.dip.t-dialin.net) left irc: Client Quit20:19
fijalwe have for x86_6420:19
fijalthat's wrong?20:19
arigato32-bit-only macs are getting old nowadays, but there are still some20:20
fijalit's kind of hard to find a good buildbot for them though20:21
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/16320:21
arigatoyes, because you need an old mac with (comparatively) tons of ram20:21
arigatoah well, I suppose it's possible to just use a recent mac in 32-bit mode20:22
fijalwhere is it supposed to be?20:22
arigatofijal: do we want a better approach to debugging this than fixing one file at a time per day?20:23
fijalwell, if you have a build you can run this script on windows and try it20:23
fijalI don't have windows though20:24
arigatobah, pyexpat was disabled during translation20:24
arigatolike bz2, _ssl and zlib20:24
fijalit would be faster if lib-python tests didn't take 8h20:24
arigatoso we need more efforts to get a correct windows build20:25
amaury_fijal: they are not slow, but there sefault with a popup20:25
fijalamaury_: disabling dr watson does not help?20:25
amaury_I cannot, need root access20:25
fijalanto should have one20:26
fijalif not anto (who's on holiday) then iko20:26
fijalcan you mail iko?20:26
Action: fijal feels like debugging windows through a very thick glass20:26
Action: amaury_ feels the same20:26
arigatothere are even Permission denied OSErrors near the start of the translation output20:27
fijalamaury_: mine seems thicker ;-)20:28
gorakhargosh (~user@ left irc: Quit: ERC Version 5.3 (IRC client for Emacs)20:31
CIA-5503arigo 07roundup * 10#821/__debug__ is always True: 20:32
CIA-55[chatting] pypy ignores the -O option. We decided it was ok a long time ago. Maybe we20:32
CIA-55should reconsider now? (It seems that -O has the effect ... * 14https://bugs.pypy.org/issue82120:32
fijalarigato: we have a duplicate of that one20:34
fijalI didn't mark it because duplicates show up20:34
stakkars (~tismer@p5DCE36EE.dip.t-dialin.net) joined #pypy.20:35
amaury_arigato: I would like to set a specific variable for the win32 buidbot20:41
amaury_is it possible to do it in the description of builds?20:42
amaury_something like PYPY_EXTERNAL_LIBRARIES=c:\buildslave\support20:42
fijalmaybe it should be always set?20:43
arigatofijal: duplicates should not show up any more, starting from today20:44
fijalok cool20:44
fijalthen we can close another one :)20:44
justinpeelarigato: I think that fixed it. Thanks!20:45
fijalwe're down to 115 open issues20:45
arigatoalso, maybe (after the release) we can also do something more about -O (which means just "kill asserts")20:45
arigatosomething which doesn't involve a .pyc/.pyo mess20:46
arigatoyou just set the tracer to ignore asserts20:46
arigatoso you still have asserts "for a while", and then they go away when the code is jitted20:46
arigato(only with -O)20:46
arigato(and I'm fine if you think this is a horrible idea)20:46
Action: fijal thinks it's a horrible idea20:47
fijalI think this and threading issue are the only candidates for release blockers20:48
fijaland running tests with 2.720:49
CIA-5503fijal 07roundup * 10#585/Missing euc-kr encoding: [resolved] Fixed I think. * 14https://bugs.pypy.org/issue58520:50
fijalarigato: https://bugs.pypy.org/issue52020:50
fijalwhat's that?20:50
justinpeelfijal: I think that I'm actually making headway with the dtypes stuff. I know that Alex is working on an alternate way of doing it, but I'm making pretty good headway20:51
fijaljustinpeel: good :) discuss with alex though20:51
fijalI think20:51
fijalI mean nothing is stopping anyone from doing both but well20:51
fijalalex sometimes likes redoing stuff in a nicer way20:52
justinpeelI think that his method will be the only way to go for arbitrary dtypes20:52
fijalas in not only floats/ints?20:52
fijalI did not really look20:52
justinpeelbut I think that the way that I'm doing it will be faster for the core types20:52
justinpeelyes, as in struct sorts of types..20:52
CIA-5503arigo 07roundup * 10#819/Arithmetic is slower than CPython in extreme cases: 20:53
CIA-55[wontfix] It's known; it's within the "1.5x the time of CPython" factor of general20:53
CIA-55slowness of PyPy. Of course the JIT cannot help this at all. ... * 14https://bugs.pypy.org/issue81920:53
justinpeelyou can have a combination of an int and a float for instance20:53
fijalarigato: any good reason why CPython is faster?20:54
fijaljust C-level hacks?20:54
justinpeelalso, short floats and long floats won't be working for now20:54
justinpeelbecause the annotator doesn't like them20:54
justinpeelfijal: well, the faster method of multiplication is disabled right now for one thing20:55
justinpeelin regard to that issue20:55
arigatofijal: I don't exactly know, but the fact is that pypy *is* slower than cpython20:56
arigatofor almost everything (as long as the jit is off)20:56
fijalwell, it's maybe fixable at some point20:56
fijalbut I guess I'm fine with 50% hit from using RPython20:56
fijalfor the extra expresivness20:56
arigatoit seems that way, yes20:56
fijalin theory we should be able to get that 50%20:57
fijalbut in practice I don't care20:57
fijalespecially because it's about emitting C code that makes gcc happy probably20:58
arigato(and I meant above that the slowness of long multiplications is probably caused for this general issue, rather than a precise performance bug in the code)20:58
arigatofijal: that's not really clear20:59
arigatoI think that gcc does a good job20:59
arigatoe.g. with a completely different optimizer like clang, we get almost the same result20:59
itafinally, i have a core file (stupid crash only occurs once in a while)21:06
itadamn, the nightly builds do not have any debug info?21:08
fijalita: they're stripped21:08
itameh ! :-/21:08
fijalmaybe we should have both21:08
itawhere are the non-stripped ones? i am pretty certain you will be interested in a random crash with threading21:09
fenrrir (~fenrrir@ left irc: Quit: Saindo21:10
fijalthey're nowhere21:11
fijalyop, we'll21:11
fijalyou need to have your own build though as of now21:11
itabah, i'll wait for the next nightly builds...21:12
whitelynx (~whitelynx@li117-47.members.linode.com) left irc: Quit: Ex-Chat21:12
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/702 [12alex]21:13
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1329 [12alex]21:13
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/940 [12alex]21:13
masquerade (~rdeaton@galactica.lug.udel.edu) left irc: Remote host closed the connection21:14
hakanardoany objections to merging jit-short_from_state?21:14
Alex_GaynorAll tests pass, no major speed regressions?21:15
ronnyarigato: 21:15
stakkars (~tismer@p5DCE36EE.dip.t-dialin.net) left irc: Quit: schnarch21:16
hakanardoAlex_Gaynor: there are the same failures as on defult21:16
Alex_Gaynorhakanardo: sure, that's what I meant :)21:16
Alex_Gaynoroh, does the jit viewer still work with it?21:17
hakanardoactuly, after merging turnk objspace.std.test.test_methodcache fails too21:18
Alex_Gaynorthat test randomly fails all the time though21:18
hakanardoso that's not worrying then?21:19
hakanardowe do have branch support on speed now, right?21:20
caiges (~caiges@ left irc: Ping timeout: 250 seconds21:20
hakanardocould I do a benchmark run on jit-short_from_state?21:20
hakanardousing the buildbot?21:20
fijal (~fijal@ left irc: Ping timeout: 258 seconds21:20
Alex_Gaynorhakanardo: yes, that's possible now (good idea), I have no idea how to do it.  it may just mean starting the builders on the buildbot with the branch21:21
hakanardoI'll try that21:21
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/29 [12hakanardo, jit-short_from_state]21:22
caiges (~caiges@ joined #pypy.21:23
Alex_Gaynorperhaps run the 32 bit version as well21:23
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/825 [12hakanardo, jit-short_from_state]21:23
lucian (~lucian@78-86-217-168.zone2.bethere.co.uk) left irc: Ping timeout: 246 seconds21:26
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Read error: Operation timed out21:27
Nick change: Husio_ -> Husio21:28
litb (~litb@p4FF7D78F.dip.t-dialin.net) joined #pypy.21:30
litbhello all21:30
litbis pypy 100% compatible with cpython?21:30
Alex_Gaynordepends what you mean by 100%21:30
Alex_Gaynorwe implement the Python language faithfully, we don't have all the same per-implmentation characteristics21:30
Alex_Gaynore.g. CPython has deterministic collection, we don't21:30
litbso that I can replace my /usr/bin/python  with a symlink to /usr/bin/pypy  and still have my KDE desktop applets run 21:31
litband all the other stuff that relies on python21:31
exarkunprobably not21:31
litblike, gtk and qt python bindinds21:31
arigato (~arigo@ left irc: Quit: See you21:33
litbexarkun: I wonder what the reason is that they probably will not work?21:36
exarkunlitb: gtk and qt won't work21:36
exarkunsince they're not Python, they use the CPython C extension APIs to hook the native libraries up to the Python interpreter21:37
exarkunPyPy has some support for the CPython C extension API, but not complete support21:37
litbaahh I understand now21:37
Alex_Gaynorand it's slow21:37
litbits unfair that they use the CPython API and not a vendor neutral API21:38
Alex_Gaynorthat's life :)21:38
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) left irc: Read error: Connection reset by peer21:44
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) joined #pypy.21:44
intchanter (~intchante@c-76-27-100-38.hsd1.ut.comcast.net) left irc: Ping timeout: 240 seconds21:46
intchanter (~intchante@c-76-27-100-38.hsd1.ut.comcast.net) joined #pypy.21:48
lucian (~lucian@78-86-217-168.zone2.bethere.co.uk) joined #pypy.21:49
LoganLK (~LoganLK@ joined #pypy.21:51
bob489 (~Bob@ool-43539e48.dyn.optonline.net) joined #pypy.21:55
Circlefusion (~cf@74-138-197-146.dhcp.insightbb.com) left irc: Ping timeout: 240 seconds21:57
bbot2Started: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/61822:03
voidspace (~voidspace@python/psf/voidspace) joined #pypy.22:10
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/1329 [12alex]22:11
voidspace (~voidspace@python/psf/voidspace) left irc: Client Quit22:12
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/702 [12alex]22:15
JaRoel|4d (~jaroel|4d@2001:980:7133:1:388c:c003:45ff:f8ae) left irc: Remote host closed the connection22:16
JaRoel|4d (~jaroel|4d@sink.jaroel.nl) joined #pypy.22:17
ojii (~ojii@62-13.3-85.cust.bluewin.ch) left irc: Ping timeout: 258 seconds22:18
voidspace (~voidspace@python/psf/voidspace) joined #pypy.22:19
voidspace (~voidspace@python/psf/voidspace) left irc: Client Quit22:19
voidspace (~voidspace@python/psf/voidspace) joined #pypy.22:22
JaRoel (~jaroel|4d@2001:980:7133:1:8981:213b:bbed:eb5e) joined #pypy.22:25
JaRoel|4d (~jaroel|4d@sink.jaroel.nl) left irc: Read error: Connection reset by peer22:26
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/940 [12alex]22:29
linq3 (~ident@CPE00259c448466-CM00195eda813e.cpe.net.cable.rogers.com) left irc: Ping timeout: 258 seconds22:35
kkris1 (~kris@80-123-47-216.adsl.highway.telekom.at) left irc: Quit: Leaving.22:49
ericflo (~ericflo@c-98-234-217-1.hsd1.ca.comcast.net) joined #pypy.22:50
oal (~olav@5.79-160-122.customer.lyse.net) left irc: Remote host closed the connection22:54
Arfrever (~Arfrever@gentoo/developer/Arfrever) left irc: Quit: Ex+re22:54
iratsu (~alang@host109-150-241-124.range109-150.btcentralplus.com) joined #pypy.22:59
justinpeelare bools wrapped in rpython?23:06
justinpeelwell, maybe that's a dumb question..23:07
donri (~donri@unaffiliated/dagodenhall) left irc: Remote host closed the connection23:07
justinpeelfor ints with numpy, I use space.float_w(space.float(val)) for an inputted value that will be set in a float array.. what would I use a bool array?23:07
justinpeelI don't think that there is a bool_w23:08
justinpeelor maybe there is.. I should have just tried it23:08
Alex_Gaynorjustinpeel: space.is_true23:09
Alex_Gaynorthat returns an unwrapped bool23:09
Alex_Gaynor(which is really just an int who's lowest bit is only ever set, AFAIK, same as C)23:09
justinpeelokay, that's what I was trying and I got an error..23:13
justinpeelspace.is_true(space.bool(val)) was what I was using23:13
Alex_Gaynorjust, space.is_true(w_val)23:14
bbot24Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/29 [12hakanardo, jit-short_from_state]23:27
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/3023:27
bbot24Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/825 [12hakanardo, jit-short_from_state]23:27
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/82623:27
iratsu (~alang@host109-150-241-124.range109-150.btcentralplus.com) left irc: Ping timeout: 250 seconds23:37
mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) left irc: Remote host closed the connection23:38
mcdonc (~mcdonc@ip72-209-213-54.dc.dc.cox.net) joined #pypy.23:39
bbot24Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/61823:44
JaRoel (~jaroel|4d@2001:980:7133:1:8981:213b:bbed:eb5e) left irc: Remote host closed the connection23:48
JaRoel|4d (~jaroel|4d@sink.jaroel.nl) joined #pypy.23:49
justinpeel (~nepo@c-76-23-44-72.hsd1.ut.comcast.net) left irc: Ping timeout: 250 seconds23:53
--- Sun Aug 14 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!