#pypy IRC log for Tuesday, 2010-12-21

lizardo (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) joined #pypy.00:06
jbaker (~jimbaker@c-24-8-39-78.hsd1.co.comcast.net) left irc: Quit: jbaker00:13
dmalcolm (~david@nat/redhat/x-zexqprwtuompfkoa) left irc: Ping timeout: 265 seconds00:22
Trundle (~andy@python/site-packages/trundle) left irc: Remote host closed the connection00:25
bobbyz (~bobbyz@c-67-184-47-92.hsd1.il.comcast.net) left irc: Ping timeout: 240 seconds00:28
__name__ (~name@sburn/devel/name) left irc: Remote host closed the connection00:42
pybot joined #pypy.01:05
-ChanServ (ChanServ@services.)- [#stackless] Welcome to #stackless. This channel is for the discussion of matters relating to Stackless Python. Remember that others in this channel may not be in the same timezone as you, so stick around for an answer (even if only in your channel history) if you ask a question. The mailing list is also recommended as being a faster place to get help (http://www.stackless.com/mailman/listinfo/stackless).01:06
lizardo_ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) joined #pypy.01:11
lizardo (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) left irc: Ping timeout: 240 seconds01:13
fenrrir (~fenrrir@187.40.133.96) joined #pypy.01:30
lizardo__ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) joined #pypy.01:32
lizardo_ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) left irc: Ping timeout: 264 seconds01:34
lizardo_ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) joined #pypy.01:50
lizardo__ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) left irc: Ping timeout: 240 seconds01:53
fenrrir (~fenrrir@187.40.133.96) left irc: Quit: Saindo02:05
lizardo__ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) joined #pypy.02:11
lizardo_ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) left irc: Ping timeout: 265 seconds02:14
lizardo__ (~lizardo@201-75-114-5-ma.cpe.vivax.com.br) left irc: Quit: Leaving02:22
jbaker (~jimbaker@64.134.124.160) joined #pypy.02:34
ousado_ (~ousado@p5B3C14C1.dip0.t-ipconnect.de) joined #pypy.02:35
ousado (~ousado@unaffiliated/ousado) left irc: Read error: Operation timed out02:35
Burgundy (blabla214@188.26.168.118) left irc: Ping timeout: 272 seconds02:57
bobbyz (~bobbyz@c-67-184-47-92.hsd1.il.comcast.net) joined #pypy.03:01
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy.03:25
Nick change: headtone -> facetone03:27
jbaker (~jimbaker@64.134.124.160) left irc: Quit: jbaker03:32
nettok (~quassel@200.119.183.37) left irc: Ping timeout: 240 seconds03:50
stakkars (~tismer@77-21-85-35-dynip.superkabel.de) left irc: Quit: schnarch04:03
nettok (~quassel@200.119.183.37) joined #pypy.04:20
j__ (~j@188.123.230.128) left irc: Ping timeout: 240 seconds04:28
stakkars (~tismer@77-21-85-35-dynip.superkabel.de) joined #pypy.04:30
jcea_BT (jcea@jabber.hst.ru) left #pypy.04:39
bobbyz (~bobbyz@c-67-184-47-92.hsd1.il.comcast.net) left irc: Read error: Operation timed out04:42
bobbyz (~bobbyz@12.131.26.130) joined #pypy.04:55
nettok (~quassel@200.119.183.37) left irc: Read error: Connection reset by peer05:23
nettok (~quassel@200.119.183.37) joined #pypy.05:23
AdemanI wonder if this isn't finishing because untranslated I swap too much05:25
AdemanI'm "hung" at a PyString_AsString() operating on the same string as this05:25
AdemanPyString_Size(<* <C object Struct PyObject { c_ob_refcnt, c_ob_type } at 0x16f84758>>)->2097152005:25
cwillu_at_work (~cwillu@cwillu.com) left irc: Ping timeout: 240 seconds05:36
stakkars (~tismer@77-21-85-35-dynip.superkabel.de) left irc: Quit: schnarch05:47
stakkars (~tismer@77-21-85-35-dynip.superkabel.de) joined #pypy.05:50
fijal (~fijal@41-134-41-226.dsl.mweb.co.za) joined #pypy.06:10
verte (~verte@python/site-packages/verte) joined #pypy.06:24
Alex_Gaynorfijal: looks liek armin fixed the twised regression06:30
fijalyes06:30
fijalhave you seen what was that?06:30
Alex_Gaynorfijal: it was the letting **kwargs be traced iff `not kwargs`06:30
Alex_Gaynorthat code was removed because it broke (f**0)06:30
Alex_Gaynorerr f(**0)06:31
fijalyeah06:31
Alex_Gaynorbut now it does both correctly :)06:31
Alex_Gaynorfijal: did you see the sqlalchemy benchmark mike bayer made?06:31
fijalCreating cling, an interactive interpreter interface for clang06:31
fijalwow, that sounds cool06:31
fijalAlex_Gaynor: no, I did not06:31
Alex_Gaynorhttp://twitter.com/#!/zzzeek/status/1615776362109747206:32
Alex_Gaynorhttp://twitter.com/#!/zzzeek/status/1615811444946534406:32
Alex_Gaynorhttp://twitter.com/#!/zzzeek/status/1668103140698112006:32
Alex_Gaynorwith sqlite the issue is obviously ctypes, I want to look at some traces for pg800006:33
fijalisn't pg8000 also ctypes?06:33
Alex_Gaynorno, it uses raw sockets I think06:33
fijalah, ok06:33
fijalraw sockets are kind of slow06:33
fijalsomeone has to look into that06:33
Alex_Gaynoryes, but we're still faster :)06:33
Alex_Gaynorif it was ctypes and we were faster than pg8000 would need to have tons of nonsense to eliminate :)06:34
fijalsqlalchemy is scary btw06:34
Alex_Gaynoryes, a little, but it's pretty well done06:34
fijalI looked into tracebacks06:34
Alex_GaynorI suppoes PyPy developers shouldn't talk about scary06:34
fijaland it's all __gets__ and stuff06:34
fijalwell...06:34
Alex_Gaynorsounds like good for inlining :)06:34
fijalin a way our code is mostly rectangular06:34
Alex_Gaynoreh, what does that mean?06:35
fijalexcept we have all this fastpaths *until* you do stuff like this06:35
fijalI mean we don't use metaclasses for the fun of it06:35
fijalwe sometimes do06:35
Alex_Gaynorpairtype?06:35
Alex_Gaynor__extends__06:35
fijalanyway, what sqlalchemy should do is to generate code by metaprogramming06:35
fijalit won't be fast with this level of dynamism06:35
fijallike use __setattr__ on object at object creation and not custom __getattr__06:36
fijalseriously06:36
fijalI can explain to zzzek if he's into listening06:36
Alex_GaynorHe probably would be, as long as it doesn't hurt on CPython06:36
Alex_Gaynorsince that's where most users are06:36
fijalyeah, it does not06:37
fijalit's usually a serious effort though06:38
Alex_GaynorI suppose suggest it, worst that happens is no one works on it.06:39
Alex_GaynorTODO.append I guess06:39
Alex_Gaynoreh, I love when people do semi-nonsense like `foo == None`06:40
fijalit depends what your goals are06:41
fijal"make sqlalchemy uber-fast on pypy"06:41
fijalor something else :)06:41
verteand real nonsense, like enjoy == True06:41
fijalverte: you can have a custom __eq__ on enjoy but not custom __nonzero__!!!06:41
Alex_Gaynorfijal: btw, what were you saying about numpy last night?06:42
fijalwhat what?06:42
vertefijal: well, there is no accounting for taste06:43
Alex_Gaynorsomething about "we can do better than that" I don't really remember, it was 4 in the morning06:43
fijalheh :)06:43
fijalI'm thinking about doing some stuff around micronumpy06:43
fijaland compile loops in real time06:43
fijalto have better locality06:43
Alex_Gaynorwhat does that mean "real time"?06:43
fijalso we can do better than providing just numpy.core06:44
fijalJIT06:44
fijallike a + b + c would become one loop and not 206:44
Alex_Gaynorand not allocate the intermediate array06:44
Alex_GaynorI suppose first we should just have numpy.core, because that's orthagonal to optimized a + b + c I think06:44
fijalyes and no06:45
fijalwell, I'll see06:45
Alex_Gaynoralready we sohuld be much faster if users write custom algorithms, since we can unbox06:45
kenaan03fijal 10r8008914 pypy/benchmarks 05A(/lib/mako/ .hgignore .hgtags CHANGES LICENSE MANIFEST.in README README.py3k distribute_setup.py /lib/mako/doc/ /lib/mako/doc/build/ Makefile /lib/mako/doc/build/builder/ __init__.py builders.py ...06:45
fijalI did not even start yet06:45
fijalyeah06:45
fijalbut users don't do that06:45
fijalanyway06:45
fijalI want out of line guards first06:45
Alex_Gaynordon't they? I don't do any numpy so I'm just guessing here06:45
kenaan03fijal 10r8009014 07M(pypy/benchmarks/benchmarks.py): Add mako benchmark ...06:46
fijalno, they're too slow06:46
vertewhich is why we have all these ufuncs06:46
CIA-903fijal 07roundup * 10#596/Adding Mako template engine benchmark to benchmarks.: [resolved] Added in 80090. Sorry it took so long... * 14https://codespeak.net/issue/pypy-dev/issue59606:47
Alex_Gaynorwell, with pypy it can be fast I guess06:47
fijalAlex_Gaynor: that does not matter06:47
Alex_Gaynortrue06:47
fijalbesides, a + b + c is better than writing a loop06:47
Alex_Gaynori should stop inventing new work and help with jittypes I guess06:47
fijalI wonder what are rules for GCC vectorization btw06:48
fijaloh wow06:48
fijalthere are some docs06:48
Alex_Gaynorprobably the rules are whatever the current code does06:48
Alex_Gaynoro_O wow06:48
fijalgcc is so bad at docs that I usually don't even bother looking06:48
vertereading the gcc source is never enlightening06:49
bobbyz (~bobbyz@12.131.26.130) left irc: Ping timeout: 260 seconds06:51
Alex_Gaynorhttp://paste.pocoo.org/show/307728/06:51
hruske (~Gasper@188.230.156.183) joined #pypy.06:53
Alex_Gaynorfijal: W_Imap.w_func can be marked _immutable_fields_ right?06:54
Alex_Gaynoractually so can iterator_w right?06:56
Alex_Gaynoriterators*06:56
Alex_GaynorI guess i should write tests06:57
fijalAlex_Gaynor: pfff, another case for optimizations06:57
fijalnote that removing guards is *always* good :)06:57
Alex_Gaynoryes, I'll write tests and commit later06:58
Alex_Gaynorpypy is the art of 3 second optimizations and then tests06:58
verte (~verte@python/site-packages/verte) left irc: Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org06:58
fijalAlex_Gaynor: gcc accumulated such stuff over years and years06:59
Alex_Gaynorfijal: tests or optimizations?06:59
fijaland we're not yet at the level of "this asm intructions would be cooler"06:59
fijaloptimizations06:59
Alex_Gaynoryes someday we'll have to write better instruction scheduling, for now we can focus on big picture things06:59
Alex_Gaynorthings that I can fix with one line of code: http://paste.pocoo.org/show/307733/07:00
fijalhow you approach instruction scheduling at all?07:00
Alex_GaynorI don't know much, just that you use a DAG07:00
fijalfor some obscure reason chrome thinks that all stuff that has no other favicon is wikipedia07:00
fijalare itertools looked at?07:00
Alex_Gaynoryes, I have trace that shows it07:00
Alex_Gaynorhttp://paste.pocoo.org/show/307734/07:01
fijalok07:01
bobbyz (~bobbyz@c-67-184-47-92.hsd1.il.comcast.net) joined #pypy.07:07
fijalgood point on pypy-dev about pypy not showing up on "python performance" searches07:08
bobbyzfijal, just out of curiosity, what are out of line guards?  I've seen you mention that work a couple times now07:13
fijalbobbyz: guards that are not within your code07:14
fijalbut instead they invalidate assembler if you trip them07:14
lmoura_ (~lauromour@187.59.116.139) joined #pypy.07:15
lmoura (~lauromour@186.212.118.58) left irc: Read error: Connection reset by peer07:16
bobbyzhmm....trying to think that through.  Would that have to do more with freeing mem allocated for assembler that has been recognized as no longer useful then?07:17
fijalsort of07:18
fijalbut most gains come from the fact that you don't need all those annoying checks any more07:18
bobbyzahhhhh, ok, now I see07:19
bobbyzinteresting, I look forward to seeing that in action07:19
fijalwell, I'm kind of busy with my day job so it might take some time07:20
fijal(it's not hard, just annoying)07:20
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Remote host closed the connection07:21
bobbyzgotcha, well I'm in no particular rush...I'm just fascinated by the different approaches each of you are taking07:24
bobbyzhakan with his loop unrolling07:24
bobbyzyou with these out of line guards07:24
bobbyzthe jit'd ctypes module...all sorts of interesting things in development07:24
bobbyzI'm slowly learning a bit more about the pypy code base as well, but don't have enough time during the day to really dive in07:25
fijalheh :)07:26
fijalwellcome to the club I feart07:26
fijalfear07:26
bobbyzyeah, I envy those that are full-time supported to work on pypy07:26
bobbyzonly a handful though right?07:27
bobbyzArmin and a couple others?07:27
hruske (~Gasper@188.230.156.183) left irc: Quit: Leaving07:27
fijalnoone is full-time07:28
fijalat most half-time07:28
fijaland it's armin anto and carl07:28
fijal"supported"07:28
fijalwell, nothing stops you from doing a PhD and convincing your supervisor to work on PyPy07:28
fijal:]07:28
bobbyzlol, yeah right  :)  See you in 5 years07:29
bobbyzheh07:29
fijalcoffee07:30
bobbyzsleep for me...good night, enjoy the coffee!  Hopefully it's good coffee...probably better than anything here in the US07:32
fijalnot sure07:32
fijallocals are not very big on coffee07:32
fijalgood night07:33
bobbyz:)07:33
ErKa (keryell@m3.wifi.enstb.org) joined #pypy.08:06
CIA-903virhilo 07roundup * 10#596/Adding Mako template engine benchmark to benchmarks.: [chatting] thanks * 14https://codespeak.net/issue/pypy-dev/issue59608:17
bsod1 (~osa1@188.58.161.46) joined #pypy.08:33
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy.08:33
fijalsqlalchemy is sloooow08:41
ivanmake pypy generate entire ORM frameworks08:43
fijalheh08:43
fijalivan: it's about writing stuff in a sane style08:44
fijallike08:44
fijal"reduce level of dynamism"08:44
mvt (~mvantelli@87.213.45.85) joined #pypy.08:48
Alex_Gaynorfijal: sqlalchemy makes up for tons of python code with doing less SQL calls, I guess it's our job to make python fast :)09:01
fijalwell09:01
fijalyes and no :)09:01
Alex_Gaynorwell our job with cooperation09:01
Alex_Gaynorgrumble http://paste.pocoo.org/show/307752/ stupid cpython09:06
fijalyeah, I know09:07
fijalit takes 26s to pack 1M of data09:08
fijalthat's *really* slow09:08
Alex_GaynorI was considering switching Django templates to use StringIO, but I can't really if it's slower on CPython :/09:08
cwillu_at_work (~cwillu@cwillu.com) joined #pypy.09:13
jell26s for only 1M?09:13
jellomg09:13
jellfijal: try it on something younger thatn 486dx6609:14
jellthan*09:14
fijalcan I turn on somehow eXception on warning?09:17
fijaldone09:18
Alex_Gaynorfijal: -Werror09:21
fijalit's slooooow09:25
Burgundy (blabla214@188.26.10.250) joined #pypy.09:25
Ademanpy.py seems to hang on a cpyext call PyString_AsString() which at worst should be a memcpy, the string in question is about 2mb, is it plausible that ll2ctypes is struggling with this?09:28
fijalyeah09:30
fijalit does insane stuff09:30
Ademanthanks, I guess it's time to make sure my branch still translates anyways09:32
fijalso09:34
fijalthe dude asks on pypy-dev about missing attribute09:34
fijaland when I google what it does I get the very same post09:35
ivanAlex_Gaynor: how about an abstraction that uses += in CPython and cStringIO in pypy09:36
fijalAlex_Gaynor: honestly, looking at django templates, speed of string creation is by far not the most important part09:37
Ademanfor that matter, WHY is cStringIO too slow in CPython?09:37
fijalhow about making sure you dont' do a billion dict lookups?09:37
ivanAdeman: turning it into a str copies everything09:39
Alex_Gaynorfijal: I have a proposal somewhere for compiling django templates into nice python09:39
Ademanfijal: in numpy itemsize refers to the c value of an item, numpy.array([1,2,3], dtype=int).itemsize == 409:40
Ademaner I guess that probably doesn't apply to the long type09:40
kursor (~kursor@p57BB055A.dip0.t-ipconnect.de) joined #pypy.09:41
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Read error: Operation timed out09:42
fijalthat's __itemsize__09:43
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy.10:00
Ademangrrr, debug_refcount somewhat needs kwargs10:04
Ademanactually, that stack frame nastiness can't be rpython anyways...10:05
hruske (~Gasper@89-212-113-197.static.t-2.net) joined #pypy.10:09
voidspace (~voidspace@python/psf/voidspace) joined #pypy.10:15
nettok_ (~quassel@200.119.183.37) joined #pypy.10:27
mvt` (~mvantelli@87.213.45.85) joined #pypy.10:32
mvt (~mvantelli@87.213.45.85) got netsplit.10:33
nettok (~quassel@200.119.183.37) got netsplit.10:33
sav (~lsd@jagat.xored.org) got netsplit.10:33
lizardo (~lizardo@189.2.128.130) joined #pypy.10:34
nettok_ (~quassel@200.119.183.37) left irc: Ping timeout: 260 seconds10:38
Ademan"[version:ERROR] No subversion revision number available!" Hehe, looks like I might need to sync with trunk :-)10:39
voidspace (~voidspace@python/psf/voidspace) left irc: Ping timeout: 272 seconds10:40
voidspace (~voidspace@python/psf/voidspace) joined #pypy.10:43
sav (~lsd@jagat.xored.org) got lost in the net-split.10:44
nettok (~quassel@200.119.183.37) got lost in the net-split.10:44
mvt (~mvantelli@87.213.45.85) got lost in the net-split.10:44
pedronis (~pedronis@253-109.104-92.cust.bluewin.ch) left irc: Ping timeout: 260 seconds10:44
pedronis (~pedronis@253-109.104-92.cust.bluewin.ch) joined #pypy.10:49
fijalhow do I merge .hgignore from default to branch?10:53
ronnyfijal: you cant merge single files10:55
fijalthat kind of sucks10:55
fijalok, how do I copy .hgignore?10:55
ronnyi think you can use hg revert for a single file or hg cat10:56
fijal"hg revert"?10:56
fijalI have .hgignore on default10:56
fijaland no .hgignore on branch10:56
WonTu (~WonTu@p57B56179.dip.t-dialin.net) joined #pypy.11:00
WonTu (WonTu@p57B56179.dip.t-dialin.net) left #pypy.11:00
sav (~lsd@jagat.xored.org) joined #pypy.11:04
arigato (~arigo@fwstups.cs.uni-duesseldorf.de) joined #pypy.11:16
fijalarigato: mornign11:17
arigatohi11:17
antocuni (~antocuni@host158-121-dynamic.10-79-r.retail.telecomitalia.it) joined #pypy.11:17
antocunihi11:19
ErKa (keryell@m3.wifi.enstb.org) left irc: Ping timeout: 260 seconds11:22
fijalPyPy does not support '?' in structs11:40
fijalwhich is apparently C type _Bool11:41
arigatodetails++ please11:44
fijalstruct.pack('?', 3) does nto work11:45
fijalit does work in CPython11:45
fijaland is documented11:45
arigatonot in Python 2.511:45
fijaluh11:45
fijalok11:45
fijalindeed11:46
bsod1 (~osa1@188.58.161.46) left irc: Ping timeout: 260 seconds11:46
CIA-903arigo * 6ae8db82e09f r40152 10bitbucket/.hgignore: 11:47
CIA-9Also ignore a top-level "bin" directory, which may be11:47
CIA-9created by "pypy setup.py install". * 14https://bitbucket.org/pypy/pypy/changeset/6ae8db82e09f/11:47
CIA-903arigo * 83ebeb857ddf r40153 10bitbucket/pypy/module/ (2 files in 2 dirs): Bump the version to 1.4.1. * 14https://bitbucket.org/pypy/pypy/changeset/83ebeb857ddf/11:47
CIA-903arigo * e503e483e9ac r40154 10bitbucket/: Making the release branch. * 14https://bitbucket.org/pypy/pypy/changeset/e503e483e9ac/11:50
bsod1 (~osa1@188.58.252.112) joined #pypy.11:54
Shinato (~John@osbk-4d08b23a.pool.mediaWays.net) left irc: Read error: Connection reset by peer11:58
arigatowould be convenient if amaury was here :-(11:59
Moku (~John@osbk-4d08b23a.pool.mediaWays.net) joined #pypy.12:02
Ademanhurgh hg push is slow12:04
CIA-903ademan * 2f54d43d54d6 r40155 10bitbucket/pypy/module/cpyext/test/test_floatobject.py: I must have been out of it when I wrote this test, PyFloat_FromString() takes a PyObject * * 14https://bitbucket.org/pypy/pypy/changeset/2f54d43d54d6/12:07
CIA-903ademan * 72d3c7738241 r40156 10bitbucket/pypy/module/cpyext/api.py: Changed logging cpyext calls to cpyext-call and moved function __name__ lookup to import time. * 14https://bitbucket.org/pypy/pypy/changeset/72d3c7738241/12:07
CIA-903ademan * 8c5587a07805 r40157 10bitbucket/pypy/module/cpyext/floatobject.py: 12:07
CIA-9Implementation of PyFloat_FromString() is much closer to being correct, but the docs don't describe whether or not to set an error when it can't parse12:07
CIA-9the string. * 14https://bitbucket.org/pypy/pypy/changeset/8c5587a07805/12:07
CIA-903ademan * dcafe3478e6e r40158 10bitbucket/pypy/module/cpyext/pyobject.py: Fixed translation errors coming from using kwargs. debug_refcount is slightly less nice, but when used with cpyext-call it should be nicer anyways. * 14https://bitbucket.org/pypy/pypy/changeset/dcafe3478e6e/12:07
arigatoI guess we should really move cpyext development to its own branch12:08
arigatoinstead of having it done in 3 different places12:09
antocuniprobably it's the proper "mercurial way" to do things12:09
antocuniyou can then easily merge it into both trunk and fast-forward12:09
arigatowell it looks very wrong whatever vcs you use :-)12:09
antocuniarigato: no, I mean having a separate branch just for cpyext12:09
antocuni(basically, I agree with you :-))12:10
arigato:-)12:10
fijalheh12:12
fijalI bet there are 7 people who would have their own opinions12:12
fijaleh :/12:14
antocuniso, it looks like emails are working fine now12:14
fijalI have multiple required versions of py.test12:14
Action: antocuni disables the bitbucket ones12:14
antocuniif you notice that you do a commit but don't receive the mail, please tell me12:15
arigatopom pom I hate windows but I'm past hating it *actively* nowadays12:16
fijalarigato: :)12:17
Shanita (~John@osbk-4db169a8.pool.mediaWays.net) joined #pypy.12:22
Moku (~John@osbk-4d08b23a.pool.mediaWays.net) left irc: Ping timeout: 265 seconds12:22
CIA-903fijal * 1677c895cf0b r40160 10bitbucket/ (3 files in 3 dirs): 12:25
CIA-9Reproduce the problem caught by test_ztranslation here. Invalidation of single12:25
CIA-9bridge * 14https://bitbucket.org/pypy/pypy/changeset/1677c895cf0b/12:25
CIA-903fijal * a3422281e05e r40159 10bitbucket/pypy/jit/metainterp/ (test/test_basic.py test_outofline.py): Move out of line tests out of test_basic * 14https://bitbucket.org/pypy/pypy/changeset/a3422281e05e/12:25
antocuniuhm indeed, the payload sent by bitbucket contains the commit unsorted12:27
fijalgreat? :(12:29
antocuniwell, it's not to hard to sort them by revision12:30
antocunis/to/too12:30
arigatopom pom "hg clone" takes forever on this windows box12:32
ambv (~ambv@host-128.192.109.111.psi.de) joined #pypy.12:38
Kaskuka (~John@osbk-4db17852.pool.mediaWays.net) joined #pypy.12:47
Shanita (~John@osbk-4db169a8.pool.mediaWays.net) left irc: Ping timeout: 245 seconds12:47
fenrrir (~fenrrir@187.40.152.160) joined #pypy.12:50
CIA-903antocuni * eafd7a336e51 r396 10bitbucket/bitbucket_hook/ (hook.py test/test_hook.py): make sure to send the emails sorted by revision number, because the bitbucket payload might be unsorted * 14https://bitbucket.org/pypy/buildbot/changeset/eafd7a336e51/13:02
ArneBab (~quassel@HSI-KBW-109-192-087-031.hsi6.kabel-badenwuerttemberg.de) left irc: Ping timeout: 276 seconds13:16
Nick change: jae_ -> jae13:18
romao (~romao@unaffiliated/errado) joined #pypy.13:32
ArneBab (~quassel@HSI-KBW-109-192-087-031.hsi6.kabel-badenwuerttemberg.de) joined #pypy.13:33
voidspace_ (~voidspace@python/psf/voidspace) joined #pypy.13:35
Edmund_Ogban (~quassel@static-217-133-80-37.clienti.tiscali.it) joined #pypy.13:36
WonTu (~WonTu@p57B56179.dip.t-dialin.net) joined #pypy.13:39
arigatohttp://pypy.org/download/13:40
WonTu (WonTu@p57B56179.dip.t-dialin.net) left #pypy.13:40
arigatothe 1.4.1 release files are there13:40
arigatofor linux13:40
arigatoif anyone feels like downloading them and checking them :-)13:40
arigato(also, side challenge: why are they 2MB larger than pypy-1.4?)13:40
Action: antocuni downloads13:41
antocuniarigato: different compression level?13:41
arigatothe mac32 is in "make", so it should follow shortly13:41
arigatoantocuni: uh?13:41
antocunito explain the different size, maybe one was built with bz2 -9 and the other wasn't13:42
arigatobut -9 is the default13:42
antocunithen I don't know :-)13:42
arigatono, it expands to something really bigger13:44
arigatostrip?13:44
arigatoyes :-(13:44
arigatopff13:45
antocunipypy 32 bit seems to work13:46
antocuniah, and sys._mercurial returns ('PyPy', 'release-1.4.1', 'e503e483e9ac')13:46
antocuniwhich I suppose it's nice13:47
arigato:-)13:47
arigatofijal: if you have a comment about strip, now is the time, otherwise I'm just going to hack it again13:47
arigato"use package.py and not make_release.py" is fine as an answer13:49
arigatoin that case I would love to see make_release.py disappear :-)13:49
arigatoah13:51
arigatook, I see13:52
arigatohttps://codespeak.net/issue/pypy-dev/issue58713:52
arigatocauses "strip -x" to be executed, even on linux, which still keeps 10MB of symbols in the executable13:52
arigatobut on Linux, cpyext seems to work fine even if the executable was fully strip'ed13:54
jelltry http://www.muppetlabs.com/~breadbox/software/elfkickers.html sstrip13:54
antocuniit's a bit unfortunate that the tarballs extract to a directory which is different than their filename13:54
ErKa (keryell@m3.wifi.enstb.org) joined #pypy.13:54
antocunie.g., you cannot extract *linux and *linux64 in the same dir13:54
arigatouh?13:54
arigatoah13:54
Action: arigato fixes that13:55
Action: arigato also fixes: let's call "strip -x" on mac and "strip" everywhere else13:55
arigatono clue what the exact rule is, however13:56
CIA-903arigo 07roundup * 10#587/[PATCH] Plain 'strip' in package.py breaks cpyext on Mac OS X: [chatting] 'strip -x' leaves behind a lot of symbols, making the Linux executables 10MB bigger than they need to be --- it seems that a full 'stri ... * 14https://codespeak.net/issue/pypy-dev/issue58713:57
CIA-903arigo * 42ea9c611870 r40161 10bitbucket/pypy/tool/release/package.py: 'strip' fun. * 14https://bitbucket.org/pypy/pypy/changeset/42ea9c611870/14:00
exarkun (~exarkun@wordeology.com) left irc: Remote host closed the connection14:04
exarkun (~exarkun@wordeology.com) joined #pypy.14:05
CIA-903arigo * 6078d6bed05a r40162 10bitbucket/pypy/tool/release/make_release.py: Updates. * 14https://bitbucket.org/pypy/pypy/changeset/6078d6bed05a/14:08
j__ (~j@188.123.230.128) joined #pypy.14:09
arigatoantocuni: ok, updated with really-strip'ed versions14:14
arigatoand added the macosx 32 version14:15
Rhy0lite (~dje@nat/ibm/x-vjaqxqvnovekjdak) joined #pypy.14:25
kenaan03arigo 10r8009114 pypy/extradoc/pypy.org/source 07M(compat.txt download.txt features.txt index.txt): Release 1.4.1. ...14:28
kenaan03arigo 10r8009214 pypy/extradoc/pypy.org 07M(compat.html download.html features.html index.html): Regenerate html. ...14:29
fenrrir (~fenrrir@187.40.152.160) left irc: Quit: Saindo14:35
bobbyz (~bobbyz@c-67-184-47-92.hsd1.il.comcast.net) left irc: Read error: Operation timed out14:43
dju (dju@fsf/member/dju) left irc: Read error: Connection reset by peer14:47
CIA-903arigo * 008b5cf5174c r40163 10bitbucket/pypy/doc/release-1.4.1.txt: Add the announcement for 1.4.1. * 14https://bitbucket.org/pypy/pypy/changeset/008b5cf5174c/14:48
kenaan03arigo 10r8009314 06D(pypy/extradoc/planning/1.4.1.txt): Kill this file. ...14:48
kenaan03arigo 10r8009414 07M(pypy/extradoc/pypy.org/js/detect.js): Update. ...14:51
voidspace_ (~voidspace@python/psf/voidspace) left irc: Quit: voidspace_14:51
Action: arigato reads speed.pypy.org -- we seem to have gained an overall 14% over the last two months, not bad14:55
arigato28% over 100 days14:55
elmom_ (~elmom@a88-114-241-230.elisa-laajakaista.fi) joined #pypy.14:55
antocuniarigato: is the "Antocuni" in the release announcement supposed to be my nickname or my actual name?14:57
arigato:-)14:57
arigatofeel free to expand it to your actual name14:57
Action: antocuni does, also for ronny14:57
arigatothanks14:58
antocuniactually, for ronny I have to copy&paste it because I never managed to get it right :-)14:58
arigato:-/14:58
CIA-903antocuni * 79e4194c71d1 r40164 10bitbucket/pypy/doc/release-1.4.1.txt: put ronny first, and use the actual name because the capitalized Antocuni is really ugly :-) * 14https://bitbucket.org/pypy/pypy/changeset/79e4194c71d1/15:00
arigatosomething is still wrong with "os.openpty" & friends not showing up in translated pypys15:00
antocunidid you translate them with pypy?15:01
arigatono15:01
arigatoI even added asserts that they were really in interpleveldefs, and the asserts pass15:01
arigatobut in the end, they are just not here15:02
arigatogetloadavg() is there, for example, but not lchown() -- two recent additions15:03
antocunidoes this also mean that these are completely untested at applevel?15:04
arigatoabsolutely15:04
arigatoor else we have modified the tests so long ago that we forgot them15:05
antocuniah, maybe they are tested but preceded by "if hasattr(os, 'openpty')..."15:05
arigatoyes, that's also likely15:05
antocuniarigato: what about 1) putting the sys.recusionlimit point higher in the list, because it's anyway a difference with cpython and it's better to highlight it rather than to hide and 2) saying that our mechanism is better because you cannot really segfault even if you try?15:08
arigato2) well, it's not completely clear that it's better, because you cannot go past the limit of 768KB (even on Linux where you can have 8MB of stack without problem)15:09
antocuniah, didn't know15:09
antocuniwhy?15:09
arigato(but it was already the case in all previous versions of pypy)15:09
arigatojust because we wrote "768KB" in src/stack.h15:10
arigatoI think on Windows the limit is closer to 1MB than 815:10
exarkundoes that mean it will segfault if you run pypy with a smaller stack limit than 768kB?15:10
arigatoyes15:11
exarkun:(15:11
exarkunpthread_attr_getstack pls15:11
antocuniok, so let's not say that it's better15:11
arigato"it's different"15:11
antocuniyes15:11
arigatoexarkun: we welcome contributions, as you know :-)  right now the issue is that writing "768*1024" is more portable than "pthread_attr_getstack()"15:12
exarkunI can't check out pypy anymore :(15:12
antocuniexarkun: why?15:13
exarkunbecause my hg is too old, I guess15:13
exarkunit doesn't support subthings15:13
antocuniah yes, if it's too old you have troubles with svn externals I think15:13
dmalcolm (~david@c-24-61-12-82.hsd1.ma.comcast.net) joined #pypy.15:15
arigatook, seems that pypy-1.4.1 really has posix.lchown() and others15:17
arigatoit's only missing openpty()15:19
arigatoI can only imagine by now some obscure issue, like posix.openpty() being undefined in the non-pty context of a buildslave translation15:19
exarkunthat'd be pretty weird15:19
exarkunnot exactly the same, but15:19
exarkun>>> os.popen('python -c "import os; print os.openpty"').read()15:19
exarkun'<built-in function openpty>\n'15:19
arigatoeeeeh.  I found it out15:19
arigato/usr/bin/python on tannit has no os.openpty15:20
vborja_ (~quassel@189.143.30.112) joined #pypy.15:20
vborja_ (~quassel@189.143.30.112) left irc: Remote host closed the connection15:20
arigatowhy not??15:20
vborja (~quassel@189.143.113.53) left irc: Read error: Operation timed out15:20
CIA-903arigo * 4feaa37309b0 r40165 10bitbucket/pypy/doc/release-1.4.1.txt: 15:25
CIA-9* Reformat to a suitable e-mail width.15:25
CIA-9* Move sys.setrecursionlimit() earlier. * 14https://bitbucket.org/pypy/pypy/changeset/4feaa37309b0/15:25
arigatobah15:30
arigatoI'm trying a small custom extension module and it just segfaults at import-time on Mac OS X15:30
arigatobut it works fine on Linux15:31
arigatoit also works on CPython on Mac OS X15:34
antocuniuhm, is there a way to have a function call constfolded by backendopts?15:36
antocunisomething like @jit.purefunction, but at translation time15:36
arigato@memo?15:45
hruske (~Gasper@89-212-113-197.static.t-2.net) left irc: Quit: Leaving15:48
timonator (~timonator@redshirt.wakelift.de) left irc: Read error: Operation timed out15:48
arigatocpyext does not work at all on Mac OS X15:49
ronnysup15:50
arigatohi15:50
jbaker (~jimbaker@c-24-8-39-78.hsd1.co.comcast.net) joined #pypy.16:05
antocuniuhm, now I have a jit ffi test with long longs that mostly passes if I comment out the dump_storage() at the end of resume.finish16:09
antocunielse, I get an obscure "ValueError: Not a ctypes allocated structure" when trying to printing some box16:09
antocuniwhich seems to contain a dead array of chars (probably a void pointer?)16:10
arigatomust be more or less the first time we try to run untranslated metainterp tests with things like raw pointers16:12
antocuniah, ok16:12
antocuniI suppose it's when trying to jit this functions: http://paste.pocoo.org/show/307919/16:13
antocuni(btw, can you think of a better way to do it without allocating things?)16:13
antocuniI suppose I can add dont_look_inside16:13
antocuniuhm, no16:14
arigatono, it should already not look inside16:14
arigato(yes: by writing it in C directly, see pypy/rlib/rstack.py)16:15
antocuniyes, but it doesn't help, I still get the error16:15
arigato(sorry, it's in pypy/rlib/rmd5.py)16:15
arigatowell of course, adding dont_look_inside doesn't change anything16:15
arigatomoreover, written that way, it doesn't give standard-compliant C code16:17
arigatoa C compiler may decide to reorder the read before the write16:18
arigatoit's all a bit of a mess16:18
antocuniare you sure? I thought C compilers need to be pessimistic about aliasing16:18
antocunibut I agree that writing a bit of custom C code is probably a better idea16:19
antocuni(eheh, then it won't work on CLI :-))16:19
Action: antocuni off16:23
antocunisee you16:23
arigatosee you16:23
antocuniah, I committed but forgot to push :-)16:23
CIA-903antocuni * c4da81b326ca r40168 10bitbucket/pypy/ (5 files in 5 dirs): 16:23
CIA-9make test_slonglong_args "mostly" passing with the jit: it fails when trying16:23
CIA-9to print some debug info, it will probably be fixed by rewriting16:23
CIA-9longlong2float and float2longlong in C (which is a good idea anyway) * 14https://bitbucket.org/pypy/pypy/changeset/c4da81b326ca/16:23
CIA-903antocuni * 4a34e0e14528 r40166 10bitbucket/pypy/rlib/test/test_libffi.py: make the condition a bit clearer and more explicit * 14https://bitbucket.org/pypy/pypy/changeset/4a34e0e14528/16:23
CIA-903antocuni * 1651aae99e79 r40167 10bitbucket/pypy/rlib/ (libffi.py test/test_libffi.py): support for ulonglong * 14https://bitbucket.org/pypy/pypy/changeset/1651aae99e79/16:23
antocuni (~antocuni@host158-121-dynamic.10-79-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds16:28
lmoura_ (~lauromour@187.59.116.139) left irc: Read error: No route to host16:31
CIA-903arigo * 9dc9d95b8be0 r40169 10bitbucket/pypy/doc/release-1.4.1.txt: 16:36
CIA-9cpyext does not work on Mac OS X. Document it as such and do the16:36
CIA-9release anyway. * 14https://bitbucket.org/pypy/pypy/changeset/9dc9d95b8be0/16:36
CIA-903arigo * c407c9dc5382 r40170 10bitbucket/pypy/doc/release-1.4.1.txt: Mention it here too. * 14https://bitbucket.org/pypy/pypy/changeset/c407c9dc5382/16:39
lmoura (~lauromour@187.59.116.139) joined #pypy.16:41
mvt` (~mvantelli@87.213.45.85) left irc: Quit: Leaving16:41
ErKa (keryell@m3.wifi.enstb.org) left irc: Ping timeout: 264 seconds16:42
DasIch (~DasIch@p5DC5F0DA.dip.t-dialin.net) left irc: Ping timeout: 265 seconds16:44
DasIch (~DasIch@p5DC5FD4D.dip.t-dialin.net) joined #pypy.16:56
kursor (~kursor@p57BB055A.dip0.t-ipconnect.de) left irc: Quit: kursor16:56
Trundle (~andy@python/site-packages/trundle) joined #pypy.17:01
nettok (~quassel@200.119.156.58) joined #pypy.17:01
fenrrir (~fenrrir@201.45.145.134) joined #pypy.17:04
romao (~romao@unaffiliated/errado) left irc: Ping timeout: 240 seconds17:07
arigato (~arigo@fwstups.cs.uni-duesseldorf.de) left irc: Quit: See you17:08
romao (~romao@unaffiliated/errado) joined #pypy.17:22
woulf (~quassel@p5796F39A.dip.t-dialin.net) joined #pypy.17:23
Arfrever (~Arfrever@gentoo/developer/Arfrever) joined #pypy.17:37
timonator (~timonator@redshirt.wakelift.de) joined #pypy.17:47
ambv (~ambv@host-128.192.109.111.psi.de) left irc: Ping timeout: 240 seconds17:52
WonTu (~WonTu@p57B55F67.dip.t-dialin.net) joined #pypy.18:03
WonTu (WonTu@p57B55F67.dip.t-dialin.net) left #pypy.18:03
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Quit: Ex-Chat18:27
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy.18:27
Burgundy_ (blabla214@188.26.168.208) joined #pypy.18:29
Burgundy (blabla214@188.26.10.250) left irc: Ping timeout: 255 seconds18:30
__name__ (~name@sburn/devel/name) joined #pypy.18:37
voidspace (~voidspace@python/psf/voidspace) left irc: Quit: Uhm... gotta go18:39
Burgundy_ (blabla214@188.26.168.208) left irc: 19:03
Burgundy (blabla214@188.26.168.208) joined #pypy.19:03
gutworth (~benjamin@64-131-13-109.usfamily.net) left irc: Read error: Operation timed out19:05
gutworth (~benjamin@64-131-13-109.usfamily.net) joined #pypy.19:08
ErKa (~keryell@keryell.pck.nerim.net) joined #pypy.19:18
elmom_ (~elmom@a88-114-241-230.elisa-laajakaista.fi) left irc: Read error: Operation timed out19:45
romao (~romao@unaffiliated/errado) left irc: Quit: FSCK!20:07
WonTu (~WonTu@p57B55ED6.dip.t-dialin.net) joined #pypy.20:40
WonTu (WonTu@p57B55ED6.dip.t-dialin.net) left #pypy.20:40
vborja (~quassel@189.143.30.112) joined #pypy.20:43
ErKa (~keryell@keryell.pck.nerim.net) left irc: Ping timeout: 265 seconds20:51
hakanardo (~hakan@h59ec009e.seluldx.dyn.perspektivbredband.net) joined #pypy.20:52
Rhy0lite (~dje@nat/ibm/x-vjaqxqvnovekjdak) left irc: Quit: Leaving20:53
[Benjamin_!Benjamin@plb95-4-82-243-70-171.fbx.proxad.net] test 20:54
lizardo (~lizardo@189.2.128.130) left irc: Quit: Leaving21:00
hruske (~Gasper@188.230.156.183) joined #pypy.21:04
jbaker (~jimbaker@c-24-8-39-78.hsd1.co.comcast.net) left irc: Quit: jbaker21:07
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Remote host closed the connection21:14
vborja (~quassel@189.143.30.112) left irc: Read error: Connection reset by peer21:18
fenrrir (~fenrrir@201.45.145.134) left irc: Quit: Saindo21:35
tav (~tav@92.7.114.67) left irc: Ping timeout: 240 seconds21:49
tav (~tav@92.24.231.145) joined #pypy.21:55
Burgundy (blabla214@188.26.168.208) left irc: Ping timeout: 276 seconds21:57
CIA-903dmalcolm 07roundup * 10#612/1.4.1 source tarball embeds .svn directories: [new] FWIW, the 1.4.1 source tarball contains numerous ".svn" metadata directories. $ md5sum ../../SOURCES/pypy-1.4.1-src.tar.bz2 ebbbb156b1eb842 ... * 14https://codespeak.net/issue/pypy-dev/issue61221:57
Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy.21:58
jbaker (~jimbaker@64.134.24.133) joined #pypy.22:01
dju (dju@at.dmz.me) joined #pypy.22:12
dju (dju@at.dmz.me) left irc: Changing host22:12
dju (dju@fsf/member/dju) joined #pypy.22:12
bsod1 (~osa1@188.58.252.112) left irc: Ping timeout: 276 seconds22:19
vborja (~quassel@189.143.30.112) joined #pypy.22:31
vborja (~quassel@189.143.30.112) left irc: Remote host closed the connection22:46
Trundle (~andy@python/site-packages/trundle) left irc: Remote host closed the connection22:51
woulf (~quassel@p5796F39A.dip.t-dialin.net) left irc: Remote host closed the connection23:07
Arfrever (~Arfrever@gentoo/developer/Arfrever) left irc: Quit: Ex+re (KVIrc 4)23:16
antocuni (~antocuni@host185-123-dynamic.2-87-r.retail.telecomitalia.it) joined #pypy.23:24
pedronis_ (~pedronis@253-109.104-92.cust.bluewin.ch) joined #pypy.23:30
pedronis (~pedronis@253-109.104-92.cust.bluewin.ch) left irc: Ping timeout: 260 seconds23:30
Nick change: pedronis_ -> pedronis23:30
mat^2 (~mathias@95.209.212.42.bredband.oister.dk) joined #pypy.23:31
__name__ (~name@sburn/devel/name) left irc: Remote host closed the connection23:37
antocuni (~antocuni@host185-123-dynamic.2-87-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds23:52
Alex_Gaynorcool, all the fastforward test results are nonsense ATM :/23:53
jbaker (~jimbaker@64.134.24.133) left irc: Quit: jbaker23:54
Alex_Gaynorfijal: ATM there are a much of failures in the _io lib (on fast-forward) like http://buildbot.pypy.org/summary/longrepr?testname=AppTestBufferedRWPair%28%29.test_constructor_with_not_writable&builder=own-linux-x86-32&build=1300&mod=test.test_bufferedio23:56
Alex_Gaynoris tehre a way to fix this besides adding __del__?23:56
--- Wed Dec 22 201000:00

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