| bigdog (~scmikes@8.8.197.72) joined #pypy. | 00:02 | |
| mwhudson (~mwh@canonical/launchpad/mwhudson) joined #pypy. | 00:05 | |
| nopper (~nopper@host9-194-dynamic.3-87-r.retail.telecomitalia.it) left irc: Remote host closed the connection | 00:05 | |
| Hory (~hory@78.96.96.85) left irc: Read error: Connection reset by peer | 00:05 | |
| pedronis (~pedronis@46-180.61-188.cust.bluewin.ch) left irc: Ping timeout: 264 seconds | 00:17 | |
| Ademan (~dan@astound-64-85-244-219.ca.astound.net) joined #pypy. | 00:22 | |
| Hory (~hory@78.96.96.85) joined #pypy. | 00:23 | |
| Ademan | fijal: sorry, scrambling to get health insurance, my old coverage just ended. My existing patches are already in trunk actually | 00:23 |
|---|---|---|
| Ademan | my other modifications to unicode and such I didn't finish yet, working on it now, also migrating to git | 00:24 |
| bigdog (~scmikes@8.8.197.72) left irc: Quit: Leaving. | 00:40 | |
| Hory (~hory@78.96.96.85) left irc: Quit: Leaving. | 00:57 | |
| mwhudson (~mwh@canonical/launchpad/mwhudson) left irc: Ping timeout: 276 seconds | 01:07 | |
| Nick change: pem_ -> pem | 01:09 | |
| mwhudson (~mwh@canonical/launchpad/mwhudson) joined #pypy. | 01:11 | |
| Alex_Gaynor | gutworth: while getting distracted on wikipedia I ended up at a link to an old pypy blog post with a pretty hilarious comment from you (knowing how it all worked out :)) | 01:14 |
| Alex_Gaynor | http://morepypy.blogspot.com/2009/03/applying-tracing-jit-to-interpreter.html?showComment=1236049441286#c6213116918081541944 | 01:14 |
| AndrewBC | heh | 01:28 |
| gutworth | haha, those were the days | 01:30 |
| gutworth | when I was naive | 01:31 |
| Alex_Gaynor | gutworth: you sounded like a wide eyed kid in a candy store (which is really how pypy makes us all feel) | 01:31 |
| gutworth | except now I'm making the poison candy myself, so I don't feel so bad :) | 01:33 |
| AndrewBC | How do you sleep at night!? | 01:34 |
| mwhudson | :) | 01:35 |
| Alex_Gaynor | gutworth: wait the candy is poisoned? | 01:36 |
| AndrewBC | The truth will ou | 01:38 |
| AndrewBC | t. Gah, my t key doesn't register | 01:39 |
| DasIch (~DasIch@p5DC5F838.dip.t-dialin.net) left irc: Ping timeout: 248 seconds | 01:40 | |
| Ademan | haha Alex_Gaynor | 01:42 |
| DasIch (~DasIch@p5DC5FB05.dip.t-dialin.net) joined #pypy. | 01:53 | |
| Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Read error: Connection reset by peer | 02:07 | |
| Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy. | 02:07 | |
| DasIch (~DasIch@p5DC5FB05.dip.t-dialin.net) left irc: Ping timeout: 258 seconds | 02:18 | |
| DasIch (~DasIch@p5DC5F42A.dip.t-dialin.net) joined #pypy. | 02:18 | |
| jcreigh | hmm, Dalvik's JIT is tracing | 02:36 |
| Ademan | jcreigh: hehe I'm watching the talk referenced from that one | 02:36 |
| jcreigh | me too :) | 02:37 |
| Alex_Gaynor | I got bored watching the talk :/ | 02:37 |
| Ademan | Alex_Gaynor: GET OUUUTTTT :-p | 02:37 |
| Alex_Gaynor | Also, why did he claim tracing JITs were very popular in the '90s, Gal's paper on them was just a few years ago, did I miss something? | 02:37 |
| jcreigh | Alex_Gaynor: yeah, I don't know what he's referring to with that | 02:38 |
| Ademan | Alex_Gaynor: I'm constantly surprised by how old technology is that's just recently getting attention | 02:38 |
| jcreigh | everything interesting was done 30 years ago in Lisp or Smalltalk. :) | 02:38 |
| Alex_Gaynor | :D | 02:39 |
| Alex_Gaynor | now we just have to make it useful | 02:39 |
| AndrewBC | heh | 02:40 |
| jcreigh | wow, running benchmarks they have heap usage of their JIT as low as 100-200 kilobytes | 02:40 |
| jcreigh | PyPy uses....more. | 02:40 |
| Ademan | Alex_Gaynor: self is from 1989, I had some memory that they had a tracing jit | 02:40 |
| Alex_Gaynor | jcreigh: true, but you need a lot less profiling data with java | 02:41 |
| Ademan | jcreigh: fix it now :-p | 02:41 |
| Ademan | is there a way to explicitly exclude files from svn diff ? | 02:43 |
| Ademan | aside from making a ridiculously complex shell expansion for the paths | 02:43 |
| mwhudson | you might be able to do diff --diff-cmd 'diff -x file' maybe? | 02:44 |
| Ademan | mwhudson: good idea, ugly but it should work haha | 02:45 |
| Ademan | i'm gonna pop into #svn real quick though just in case | 02:46 |
| jcreigh | Ademan: maybe find <your criteria> | xargs svn diff? | 02:46 |
| Ademan | jcreigh: also not a bad idea | 02:47 |
| mwhudson | i think jcreigh's idea is more likely to lead to happiness | 02:47 |
| jcreigh | oh, interesting. They don't garbage collect their JITed code. When the allotted area fills up, they just nuke it and start over. | 02:50 |
| Alex_Gaynor | not even an LRU cache or something? | 02:50 |
| Ademan | what's their justification for that? surely that could cause alot of redundant jit compiling | 02:51 |
| Ademan | i'm with you Alex_Gaynor | 02:51 |
| bigdog (~scmikes@72-197-8-8-arpa.cust.cinci.current.net) joined #pypy. | 02:51 | |
| jcreigh | Ademan: He just kind of mentions it offhand without giving much justification | 02:52 |
| Alex_Gaynor | probably a "memory" issue is there justfication, though I guess that's a forest/trees issue | 02:52 |
| Alex_Gaynor | </backseat-driver> | 02:53 |
| jcreigh | he says "right now" they don't have GC on the code cache | 02:54 |
| jcreigh | might just be a 1.0 thing | 02:54 |
| Ademan | regarding the 2.2 release :-p | 02:54 |
| jcreigh | well, they're calling it the 1.0 release of the JIT | 02:54 |
| Ademan | ah, fair enough, my understanding is dalvik's had a JIT for some time now, just not enabled in release builds | 02:55 |
| Alex_Gaynor | well since 2.1 I think | 02:55 |
| Alex_Gaynor | it's not very old, even in beta | 02:55 |
| Ademan | I suppose that's not very old... | 02:56 |
| jcreigh | they don't trace across methods yet either | 03:04 |
| Alex_Gaynor | grumble, I didn't know that, maybe I should have watched the whole video. that's interesting since usually the win with tracing is "free" inlining | 03:04 |
| jcreigh | Alex_Gaynor: about halfway in the actual developer starts talking | 03:05 |
| jcreigh | actually, the first guy might be a developer, I just got a "project manager" vibe from him. | 03:05 |
| jcreigh | again, it's mentioned somewhat offhand. "We only need to copy the top stack frame, because we currently only trace within methods..." | 03:07 |
| DasIch (~DasIch@p5DC5F42A.dip.t-dialin.net) left irc: Ping timeout: 265 seconds | 03:18 | |
| DasIch (~DasIch@p5DC5FBC3.dip.t-dialin.net) joined #pypy. | 03:19 | |
| Ademan | that was something I thought a bit about, could our tracing JIT be used to produce a method JIT? I think it could (just don't specialize except on input args, don't follow just a single branch) I don't know enough about the JIT to know if that's viable, and chances are even if it was, it wouldn't be good huh? | 03:36 |
| mwhudson (~mwh@canonical/launchpad/mwhudson) left irc: Ping timeout: 276 seconds | 03:37 | |
| mwhudson (~mwh@canonical/launchpad/mwhudson) joined #pypy. | 03:40 | |
| jcreigh | Ademan: well....it's tricky. Remember that the JIT traces the interpretation *of* an interpreter. So as I understand it, it doesn't know anything at all about methods. | 03:40 |
| Ademan | jcreigh: right, but a "straight" tracing metainterp doesn't know about loops either, but we added the necessary hints to make it possible | 03:42 |
| jcreigh | ah | 03:42 |
| Ademan | I figure we could do similarly for methods, I dunno | 03:42 |
| jcreigh | well, I suppose anything is *possible* | 03:42 |
| Ademan | hehe | 03:42 |
| jcreigh | I am similarly not sure whether it would be a win or not. | 03:42 |
| Ademan | nor am I, just musing I guess | 03:43 |
| Ademan | haha, in the 2008 Google I/O talk "you don't need a JIT" | 03:45 |
| Ademan | well, true for some definitions of "need" i guess | 03:46 |
| jcreigh | yeah, and I bet you don't for most applications. | 03:46 |
| jcreigh | I wonder how much of the push for a JIT is just for buzzword compatibility. | 03:47 |
| Ademan | hah, I dunno, since I have a Droid, and I'm often quite unsatisfied with the "smoothness" of the animations, I think a JIT couldn't hurt, unless it's due to other problems than slow maths | 03:48 |
| jcreigh | I wonder how compiled Obj-C compares to interpreted Dalvik | 03:49 |
| Ademan | interesting, Dalvik does static optimization at install time | 03:50 |
| jcreigh | oh? how so? | 03:51 |
| Ademan | looks like mostly just inlining | 03:51 |
| Ademan | http://www.youtube.com/watch?v=ptjedOZEXPM 0:24 | 03:52 |
| jcreigh | Dalvik is register-based, too, so there's got to be a compilation step from stack-based Java bytecode to Dalvik. | 03:52 |
| jcreigh | I wonder how much opportunity there is to make that step more efficient. (In terms of generated Dalvik bytecode) | 03:53 |
| Ademan | that's part of the SDK, done statically before distrobution | 03:53 |
| jcreigh | right | 03:53 |
| jcreigh | but it would also be an area for potential optimization. | 03:53 |
| jcreigh | or maybe the stack -> register transformation is easier than I think. | 03:54 |
| Ademan | no idea, but around 27 he shows a simple example and compares jvm to dalvik bytecode | 03:57 |
| Ademan | "instruction dispatch is one of the killers of performance in interpreters" I thought we found just the opposite, interpreter overhead was pretty minimal... I suppose maybe that's relative to the overhead of maintaining the frame in python, so it's not really apples to apples here | 03:58 |
| jcreigh | If I understand correctly, they're just talking about getting from the code to interpret one opcode to the code that interprets the next one. | 04:00 |
| jcreigh | since you have to load up the next bytecode, and then do an indirect jump (or something) based on that | 04:00 |
| jcreigh | hurts branch prediction, etc. | 04:00 |
| jcreigh | which is why for VMs design to be interpreted, some are moving towards register-based. (Dalvik, Lua) | 04:01 |
| Ademan | sure, but I would assume that's part of interpreter overhead, maybe I'm wrong there though | 04:03 |
| Ademan | anyways, it's my anniversary and I need to pull a dinner reservation out of my butt... | 04:04 |
| Ademan | bbl | 04:04 |
| gutworth (~benjamin@64-131-1-181.usfamily.net) left irc: Ping timeout: 264 seconds | 04:06 | |
| pem (~pem@2001:cc0:201e:107:221:86ff:fe1a:e5aa) left irc: Read error: Operation timed out | 04:13 | |
| ly- (~ly-@unaffiliated/ly-) left irc: Quit: ly- | 04:14 | |
| pem (~pem@2001:cc0:201e:107:221:86ff:fe1a:e5aa) joined #pypy. | 04:14 | |
| e-max_ (~e-max@217.14.244.16) joined #pypy. | 04:23 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy. | 05:10 | |
| jcreigh (~jason@72-160-63-13.dyn.centurytel.net) left irc: Remote host closed the connection | 05:27 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 240 seconds | 05:29 | |
| eldragon | targeting stack-based VMs complicated a lot the implementation of new dynamical languages, compared to register-based. | 05:40 |
| eldragon | the new register-based VM should have #regs <= native machine's #regs, possible lesser than 8 or 16 registers. | 05:44 |
| eldragon | i686 has 8, x86-64 has 16, ARM has 16. | 05:44 |
| eldragon | it makes VM's instructions smaller | 05:45 |
| eldragon | my definitive decision is lesser than 16 regs because 8 regs is too small. | 05:51 |
| eldragon | and three-addresses (dst <- src1 op src2 ) instead of two-addresses ( srcdst <- srcdst op src2 ) | 05:56 |
| aleale (~aleale@0x57386838.honqu1.dynamic.dsl.tele.dk) joined #pypy. | 06:09 | |
| aleale (~aleale@0x57386838.honqu1.dynamic.dsl.tele.dk) left irc: Quit: aleale | 06:24 | |
| Ademan | eldragon: what new register based vm are you referring to? | 06:53 |
| ElectronicRU (~Electroni@5f44bbf7.dynamic.mv.ru) joined #pypy. | 06:56 | |
| Shanita (~John@f054163034.adsl.alicedsl.de) joined #pypy. | 06:58 | |
| Moku (~John@f049073109.adsl.alicedsl.de) left irc: Ping timeout: 260 seconds | 07:00 | |
| aukbc (~aukbc@122.165.25.137) joined #pypy. | 07:00 | |
| aukbc | hi, Normally we use the combination of HTML and Python. Can i use the combination of PHP and Python | 07:02 |
| aukbc (aukbc@122.165.25.137) left #pypy ("Leaving"). | 07:05 | |
| AndrewBC | I shudder to think of what that person is doing | 07:05 |
| Ademan | seriously, what? | 07:06 |
| AndrewBC | yeah | 07:07 |
| antocuni (~antocuni@host204-121-dynamic.11-79-r.retail.telecomitalia.it) joined #pypy. | 07:18 | |
| ronny | AndrewBC: this was just a good morning wtf, could have been worse | 07:24 |
| AndrewBC | Heh | 07:24 |
| ronny | AndrewBC: imagine he had a python interpreter written in php thats using ast walking instead of bytecode | 07:25 |
| Action: AndrewBC has a conniption' | 07:26 | |
| eldragon | Ademan, my new reg-based VM will be for smalltalk | 07:27 |
| Ademan | eldragon: ahh, based on PyPy though? | 07:30 |
| Ademan | I was excited for a moment, I always wondered why we slavishly copy CPython to the point of sharing bytecode, but I know there are practical reasons for that | 07:30 |
| bigdog (~scmikes@72-197-8-8-arpa.cust.cinci.current.net) left irc: Ping timeout: 260 seconds | 07:31 | |
| Electronic (~Pythonist@5f44bbf7.dynamic.mv.ru) joined #pypy. | 07:39 | |
| Electronic (~Pythonist@5f44bbf7.dynamic.mv.ru) left irc: Remote host closed the connection | 07:42 | |
| Electronic (~Pythonist@5f44bbf7.dynamic.mv.ru) joined #pypy. | 07:44 | |
| Electronic (~Pythonist@5f44bbf7.dynamic.mv.ru) left irc: Client Quit | 07:45 | |
| ElectronicRU (~Electroni@5f44bbf7.dynamic.mv.ru) left irc: Quit: Ex-Chat | 07:46 | |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) joined #pypy. | 07:46 | |
| antocuni (~antocuni@host204-121-dynamic.11-79-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds | 07:46 | |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) left irc: Read error: No route to host | 07:59 | |
| amaury_ (~5bc3480e@gateway/web/freenode/x-cdxmljbnwgkznjyq) joined #pypy. | 07:59 | |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) joined #pypy. | 08:02 | |
| amaury_ | bah, all tests in cpyext are broken | 08:05 |
| Ademan | amaury_: how so? | 08:10 |
| amaury_ | My recent changes broke all of them | 08:11 |
| amaury_ | http://buildbot.pypy.org/summary?branch=%3Ctrunk%3E | 08:11 |
| Ademan | :-) | 08:12 |
| hpk (~hpk@HSI-KBW-085-216-105-248.hsi.kabelbw.de) left irc: Ping timeout: 276 seconds | 08:13 | |
| amaury_ | ah, the issue is obvious when I set DEBUG_REFCOUNT=True | 08:17 |
| pedronis (~pedronis@46-180.61-188.cust.bluewin.ch) joined #pypy. | 08:18 | |
| hpk (~hpk@HSI-KBW-085-216-105-248.hsi.kabelbw.de) joined #pypy. | 08:26 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy. | 08:30 | |
| Kaskuka (~John@g228211221.adsl.alicedsl.de) joined #pypy. | 08:51 | |
| Shanita (~John@f054163034.adsl.alicedsl.de) left irc: Ping timeout: 240 seconds | 08:51 | |
| pedronis (~pedronis@46-180.61-188.cust.bluewin.ch) left irc: Ping timeout: 276 seconds | 08:53 | |
| arigato (~arigo@c-7649e353.022-54-67626719.cust.bredbandsbolaget.se) joined #pypy. | 08:56 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 240 seconds | 08:56 | |
| arigato | hi | 08:56 |
| Shinato (~John@g227164154.adsl.alicedsl.de) joined #pypy. | 08:58 | |
| Ademan | hey arigato | 09:00 |
| Kaskuka (~John@g228211221.adsl.alicedsl.de) left irc: Ping timeout: 276 seconds | 09:00 | |
| arigato | argh, the status of pypy is becoming worse and worse | 09:02 |
| arigato | we keep having various kinds of failures (and speed.pypy.org is no longer working) | 09:03 |
| Moku (~John@f048243017.adsl.alicedsl.de) joined #pypy. | 09:04 | |
| Shinato (~John@g227164154.adsl.alicedsl.de) left irc: Ping timeout: 276 seconds | 09:05 | |
| verte (~verte@120.156.135.178) joined #pypy. | 09:06 | |
| Ademan | what's wrong other than speed.pypy.org? | 09:06 |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy. | 09:07 | |
| arigato | http://codespeak.net:8099/summary | 09:07 |
| antocuni (~antocuni@host83-56-dynamic.14-87-r.retail.telecomitalia.it) joined #pypy. | 09:09 | |
| Ademan | arigato: oh I think amaury_ is fixing that | 09:09 |
| arigato | ah | 09:10 |
| arigato | is it really caused by r75092? | 09:10 |
| Ademan | he said it was his fault, no idea which rev that is | 09:10 |
| Ademan | it should address all of the cpyext failures though | 09:10 |
| Ademan | no idea about the other ones listed though | 09:10 |
| arigato | ah, ok, I'm talking actually about the app-level {trunk} failures | 09:11 |
| arigato | that was not clear however :-) | 09:11 |
| kenaan | 03arigo 10r7509814 07M(pypy/trunk/lib-python/conftest.py): Revert r75092. It causes massive failures of the nightly applevel tests. ... | 09:14 |
| arigato | hpk: ping | 09:14 |
| Hory (~hory@78.96.96.85) joined #pypy. | 09:19 | |
| hpk | arigato: hi | 09:20 |
| hpk | i see, the late night change caused some trouble - i'll look into it again | 09:21 |
| arigato | sorry, I saw that indeed it fixes the case of just running py.test test_foobar.py | 09:24 |
| arigato | but just doing "py.test --collect" in lib-python now finds duplicates (both 2.5.2 and modified-2.5.2) and eventually crashes | 09:25 |
| pedronis (~pedronis@46-180.61-188.cust.bluewin.ch) joined #pypy. | 09:27 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 260 seconds | 09:28 | |
| arigato | hpk: the current error is shown at the end of: | 09:35 |
| arigato | http://codespeak.net:8099/builders/jit-benchmark-linux-x86-32/builds/330/steps/shell_2/logs/stdio | 09:35 |
| hpk | arigato: fijal showed me this yesterday - still not sure i understand at all - is this running on tannit? | 09:36 |
| hpk | no, on bigdog it seems? | 09:36 |
| arigato | I think that to upload benchmark results, we contact "speed.pypy.org/result/add" | 09:37 |
| hpk | but is'a a DNS error, no? | 09:37 |
| arigato | this error is (as far as I know) reported by speed.pypy.org | 09:37 |
| hpk | doesn't look like it | 09:37 |
| hpk | speed doesn't have a /home/buildslave FWIW | 09:37 |
| arigato | my guess is that speed.pypy.org is trying to contact the database | 09:37 |
| hpk | i think someone is trying to contact speed.pypy.org, probably the bigdog host | 09:37 |
| arigato | yes, but then it just sends the error message back to the client (tannit) | 09:38 |
| hpk | i don't get it - what makes you think this has anything to do with speed.pypy.org? | 09:38 |
| hpk | to me it looks like: something runs on tannit, forces to run something on "bigdog", and probably there DNS fails | 09:39 |
| arigato | because e.g. "ping speed.pypy.org" works fine on tannit32 | 09:39 |
| arigato | I don't see how it has anything to do with bigdog | 09:40 |
| hpk | in the first line it reads: --force-host bigdog | 09:40 |
| hpk | of your paste | 09:40 |
| arigato | uuuh? | 09:40 |
| arigato | what is this option about? | 09:41 |
| hpk | no clue, sounds like forcing a run on a specific host to me :) | 09:42 |
| arigato | no, it seems to be just choosing a specific name to report the results for | 09:42 |
| arigato | that's still really broken | 09:42 |
| hpk | i'd look for the DNS problem on tannit however, however | 09:44 |
| hpk | something in the server logs? | 09:44 |
| arigato | why would its DNS fail to see http://speed.pypy.org/ during test runs but work right now? | 09:45 |
| hpk | arigato: don't know - are the benchmarks maybe exercising DNS and there is some throttling or so? | 09:45 |
| arigato | I doubt it | 09:46 |
| arigato | it has failed that way very consistently | 09:46 |
| Kami_ (~kami@kamislo.broker.freenet6.net) left irc: Read error: Operation timed out | 09:46 | |
| hpk | arigato: in any case, the last apache2-log entry related to a result-url on speed is this: | 09:46 |
| hpk | 85.55.96.130 - - [04/Jun/2010:02:39:21 -0500] "POST /result/add/ HTTP/1.1" 200 29 | 09:46 |
| hpk | otherwise, no attempt was made i'd think | 09:47 |
| hpk | arigato: i can't really look into anything on tannit, maybe you can give me sudo rights | 09:47 |
| arigato | ah, then tannit really managed to connect to speed.pypy.org, as I said | 09:47 |
| arigato | it does not look like the issue is on tannit at all | 09:47 |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy. | 09:48 | |
| kenaan | 03arigo 10r7509914 07M(pypy/build/bot2/pypybuildbot/builds.py): Remove the --force-host option, which does not make sense ... | 09:48 |
| hpk | arigato: huh? this was from 02:39 - your log says 05:42 | 09:48 |
| hpk | and times between tannit and speed are within a 5 minute deviation | 09:49 |
| Kami_ (~kami@kamislo.broker.freenet6.net) joined #pypy. | 09:49 | |
| hpk | arigato: i really fail to follow your logic here | 09:49 |
| arigato | no | 09:49 |
| arigato | argh, the time on tannit32 is wrong | 09:49 |
| arigato | well it's set to UTC | 09:50 |
| hpk | even then | 09:51 |
| arigato | hpk: let me ask this then: who other than tannit's benchmark runner would access /result/add on speed? | 09:51 |
| hpk | arigato: 85.55.96.130 as the apache-log says? | 09:52 |
| arigato | no :-( | 09:53 |
| hpk | no? | 09:54 |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) left irc: Remote host closed the connection | 09:54 | |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) joined #pypy. | 09:54 | |
| arigato | well, no (didn't you already do the reverse lookup?) | 09:54 |
| hpk | sorry, i am lost - 85.55.96.130 maps to 130.pool85-55-96.dynamic.orange.es for me - do you mean that? | 09:56 |
| Arfrever (~Arfrever@gentoo/developer/Arfrever) joined #pypy. | 09:57 | |
| arigato | yes | 09:57 |
| arigato | but in the same logs, the previous entry is 83.140.78.199 | 09:57 |
| arigato | which *is* tannit | 09:57 |
| hpk | and you claim this is tannit? | 09:57 |
| hpk | ah, oh? an address in spain? | 09:57 |
| arigato | no, sorry for the confusion, I thought *you* claimed this was tannit | 09:57 |
| arigato | anyway, tannit successfully contacted /result/add a bit earlier | 09:58 |
| hpk | so what? | 09:59 |
| hpk | i never claimed that there never was a connection | 10:00 |
| arigato | ...maybe we should each investigate his own lead instead of trying to convince the other that his lead is more promizing? | 10:00 |
| hpk | sure, i cannot do much on tannit, but i gave you root on speed - feel free to proove it's speed's fault, i happily retract from this discussion :) | 10:01 |
| ErKa (keryell@keryell.pck.nerim.net) joined #pypy. | 10:01 | |
| hpk | arigato: this is what you think, right, that there is a configuration problem on speed - or did i misunderstand you? | 10:02 |
| arigato | yes | 10:02 |
| hpk | k, i indeed see no evidence supporting this, but maybe i am missing something | 10:02 |
| verte (~verte@120.156.135.178) left irc: Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org | 10:05 | |
| hpk | arigato: not sure it helps or relates, but on tannit dmesg shows some "nf_conntrack: table full, dropping packet" messages - maybe the logs tell something more | 10:16 |
| arigato | bah | 10:20 |
| arigato | I repeated manually the procedure on tannit | 10:20 |
| arigato | and it worked | 10:20 |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Remote host closed the connection | 10:21 | |
| arigato | it does not make sense, now on http://speed.pypy.org/ all intermediate revisions showed up too | 10:22 |
| arigato | but I uploaded only the latest one right now | 10:22 |
| hpk | odd | 10:28 |
| kenaan | 03arigo 10r7510014 07M(pypy/benchmarks/saveresults.py): To be sure: check explicitly if we *can* contact speed.pypy.org ... | 10:35 |
| arigato | for all I know it's some obscure firewalling going on on tannit32 in the environment in which buildbot is started | 10:37 |
| arigato | hpk: yes, you are right, there are messages at the right time in the logs | 10:40 |
| arigato | nf_conntrack: table full, dropping packet. | 10:40 |
| hpk | maybe http://paulroberts69.spaces.live.com/blog/cns!665BC38F152E1206!1645.entry or something like this can help | 10:42 |
| hpk | (there is a number of hits on google) | 10:43 |
| hpk | still would be interesting to know if the benchmarks cause many connections or if it is some external factor | 10:43 |
| arigato | is that because we just ran the twisted_web benchmark? | 10:43 |
| hpk | that was my suspicion because otherwise tannit dns-resolves fine | 10:44 |
| arigato | yes, I know that some of the twisted benchmarks cause maaaaany connections | 10:44 |
| hpk | well, then it's pretty clear, isn't it? | 10:44 |
| arigato | yes :-( | 10:44 |
| hpk | i guess you can increase the limit through a sysctl setting or decrease the number of connections in the benchmark | 10:45 |
| arigato | I guess I will just have it wait a bit if it fails to resolve the name | 10:46 |
| arigato | hum | 10:46 |
| nopper (~nopper@host199-190-dynamic.57-82-r.retail.telecomitalia.it) joined #pypy. | 10:47 | |
| hpk | arigato: well, DNS-badness might cause other issues and one can get again puzzled at some seemingly unrelated issue | 10:47 |
| arigato | or just retry uploading a few times, waiting one minute inbetween? | 10:47 |
| arigato | yes :-( | 10:48 |
| hpk | really, i'd aim to avoid the problem, sysctl is not that bad - i know you dislike changing/interfering with a system at this level, but well | 10:48 |
| arigato | and then (again) we'll have the same digging the next time we move hosts until we remember? | 10:49 |
| arigato | I'd actually rather add just a time.sleep() just after each benchmark run | 10:50 |
| hpk | well, what about modifiying the benchmark to suggest/recap what we found? | 10:50 |
| arigato | by modifying it, you mean doing more changes than just time.sleep() afterwards? | 10:51 |
| hpk | printing a message on failure, i mean | 10:52 |
| kenaan | 03arigo 10r7510114 07M(pypy/benchmarks/saveresults.py): Revert r75100. Found out the issue, will be documented. ... | 10:52 |
| arigato | ah, in saveresults.py -- ok | 10:52 |
| hpk | like "DNS resolution failed, maybe the benchmarks saturated the kernel networking table? waiting 1 minute to retry" | 10:53 |
| hpk | or so - this way you can still avoid the problem with a sysctl setting, but be relatively sure the next time these benchmarks are run somewhere, somebody has a clue | 10:53 |
| arigato | well what I'm thinking of doing now is adding a sleep after each twisted benchmark run | 10:53 |
| hpk | my point is: i don't think it's a great idea to run a benchmark program that regularly causes connection trouble | 10:54 |
| arigato | in that way, after we run the benchmark, everything works fine again | 10:54 |
| arigato | just because running the benchmark includes sleeping at the ned | 10:54 |
| arigato | ..at the end | 10:54 |
| ElectronicRU (~Pythonist@5f44bbf7.dynamic.mv.ru) left irc: Ping timeout: 240 seconds | 11:08 | |
| kenaan | 03arigo 10r7510214 07M(pypy/benchmarks/own/twisted/benchlib.py): Add a sleep of 90 seconds at the end of any twisted benchmark ... | 11:09 |
| ElectronicRU (~Pythonist@5f44a0e4.dynamic.mv.ru) joined #pypy. | 11:21 | |
| arigato (~arigo@c-7649e353.022-54-67626719.cust.bredbandsbolaget.se) left irc: Quit: See you | 11:25 | |
| fiftysixquarters (~pillin@tshlabs.org) joined #pypy. | 12:05 | |
| arigato (~arigo@ratthing-216.openend.se) joined #pypy. | 12:36 | |
| ElectronicRU (~Pythonist@5f44a0e4.dynamic.mv.ru) left irc: Quit: #E>6C O >B 20A (xchat 2.4.5 8;8 AB0@H5) | 12:38 | |
| hpk | would be curious if the twistd/buildmaster on codespeak would run with less than 312 MB RSS when running on pypy-c | 12:50 |
| arigato | :-) | 12:51 |
| arigato | hpk: good, the latest run of jit benchmarks no longer has the issue | 12:52 |
| hpk | great | 12:52 |
| arigato | it's not finished but already that step passes | 12:52 |
| hpk | wonder what happens if you run two benchmarking processes at the same time ;) | 12:52 |
| arigato | :-) | 12:53 |
| hpk | are you waiting for results on your blackhole branch now? | 12:53 |
| arigato | well, I'm first waiting for the run on trunk to finish | 12:53 |
| arigato | then I'll start a run on the blackhole branch | 12:53 |
| kenaan | 03arigo 10r7510314 07M(pypy/benchmarks/own/twisted/benchlib.py): Be "future-proof" and wait 2 minutes 5'' instead of 1 minute 30''. ... | 12:54 |
| hpk | haha | 12:54 |
| arigato | :-/ | 12:54 |
| arigato | random hacking, really | 12:55 |
| hpk | and here i am, always have armin's "don't use magic numbers" in my mind :) | 12:55 |
| arigato | well, I thought about calling "netstat" and counting the lines, but it looks messy too :-) | 12:55 |
| hpk | indeed :) | 12:55 |
| hpk | can speed render a comparison between different branches? | 12:55 |
| arigato | I don't know | 12:55 |
| arigato | I somehow guess that the blackhole results will just go in the single place too | 12:56 |
| hpk | maybe you could fake it as being a different host? | 12:56 |
| arigato | ha ah argh | 12:56 |
| hpk | although they cannot be compared either | 12:56 |
| arigato | there is an argument "branch" to the save() function | 12:56 |
| arigato | but it is ignored | 12:57 |
| Action: hpk looks into the lib-python failures | 12:57 | |
| Action: hpk stops himself and first goes on his balcony with some tea, appreciating his new plants a bit | 12:57 | |
| arigato | :-) | 12:58 |
| kenaan | 03arigo 10r7510414 07M(pypy/benchmarks/own/twisted/benchlib.py): A different hack: call 'netstat' and stop when ... | 13:14 |
| arigato | (just playing around really) | 13:14 |
| hpk | arigato: when i run "python pypy/test_all.py --pypy=pypy/translator/goal/pypy-c lib-python/" without your r75102 revert-commit, it works fine. wonder what the actual invocation is that caused the failures tonight | 13:16 |
| arigato | uh | 13:16 |
| arigato | when I run "cd lib-python; py.test --collect" I see that it's collecting far too much | 13:16 |
| kenaan | 03hpk 10r7510514 05A(pypy/branch/lib-python/): trying out lib-python tests ... | 13:17 |
| ErKa (keryell@keryell.pck.nerim.net) left irc: Read error: Operation timed out | 13:17 | |
| hpk | arigato: ah right | 13:18 |
| arigato | http://codespeak.net:8099/builders/pypy-c-app-level-linux-x86-32/builds/618/steps/shell_2/logs/stdio | 13:19 |
| arigato | there, you see e.g. that test_cgi was run 4 (??) times | 13:19 |
| arigato | only the first time passed | 13:19 |
| hpk | ok, now i get it | 13:19 |
| hpk | thanks | 13:19 |
| exarkun | ehm | 13:22 |
| exarkun | http://pastebin.com/32Ng7ky7 | 13:23 |
| Alex_Gaynor | heh | 13:25 |
| hpk | seems like there is some special casing of a None value missing | 13:25 |
| arigato | >>> sys.modules['foo'] = 42 | 13:25 |
| arigato | >>> import foo | 13:25 |
| arigato | >>> foo | 13:25 |
| arigato | 42 | 13:25 |
| arigato | on CPython at least | 13:25 |
| Alex_Gaynor | arigato: on cpython setting to None means the cache is ignored | 13:27 |
| hpk | it's for short-cutting lookups of stdlib/global modules within packages | 13:27 |
| arigato | well it's different if you use dotted names | 13:28 |
| exarkun | It's the same everywhere, I'd say. | 13:28 |
| exarkun | It means there is no such module. | 13:28 |
| arigato | argh | 13:28 |
| hpk | arigato: what, looking at the import logic? :) | 13:29 |
| exarkun | If the module you were looking for happens to be foo.urllib2, then the relative import machinery goes off and looks for another module (urllib2) when it finds that foo.urllib2 doesn't exist. | 13:29 |
| arigato | it's nonsense anyway, even in CPython: | 13:29 |
| arigato | http://paste.pocoo.org/show/221866/ | 13:29 |
| exarkun | The way packages work in Python is a little bit ridiculous. But being able to stick None into sys.modules is a pretty useful feature. | 13:30 |
| arigato | but you want to stick None to 'foo.urllib2', right? | 13:31 |
| arigato | this works as far as I know in pypy too | 13:31 |
| exarkun | Mostly I want to do it for top-level modules or packages | 13:31 |
| exarkun | Like 'profile' or 'OpenSSL' | 13:31 |
| exarkun | So I can have unit tests for the code that handles those modules being missing, even if they're not | 13:31 |
| exarkun | Some Twisted unit tests are failing on pypy-c now like this: http://pastebin.com/M9T5gBzC | 13:32 |
| exarkun | (It seems to be a new issue, I don't remember seeing these before, but who knows) | 13:32 |
| gutworth (~benjamin@64-131-1-181.usfamily.net) joined #pypy. | 13:34 | |
| arigato | exarkun: I guess it's from r71972: "Merge import-fiddle branch." | 13:37 |
| ErKa (keryell@dhcp3.info.enstb.org) joined #pypy. | 13:39 | |
| nopper (~nopper@host199-190-dynamic.57-82-r.retail.telecomitalia.it) left irc: Ping timeout: 265 seconds | 13:39 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) joined #pypy. | 13:42 | |
| arigato | yes | 13:44 |
| kenaan | 03arigo 10r7510614 pypy/trunk/pypy/module/imp 07M(importing.py test_import.py): Test and fix. ... | 13:49 |
| arigato | hpk: I'm going to reboot tannit | 13:49 |
| arigato | is that ok for you? | 13:49 |
| arigato | exarkun: according to the new tests, the issue you reported was introduced in r71972 and fixed just now | 13:52 |
| hpk | arigato: yes | 13:53 |
| exarkun | arigato: hooray :) | 13:56 |
| jcreigh (~jason@72-160-63-13.dyn.centurytel.net) joined #pypy. | 13:58 | |
| arigato | humpf, of course the last run of jit-benchmarks still had "--force-host=bigdog" | 13:59 |
| ly- (~ly-@239.149.67-86.rev.gaoland.net) joined #pypy. | 14:03 | |
| ly- (~ly-@239.149.67-86.rev.gaoland.net) left irc: Changing host | 14:03 | |
| ly- (~ly-@unaffiliated/ly-) joined #pypy. | 14:03 | |
| kenaan | 03arigo 10r7510714 07M(pypy/branch/blackhole-improvement/pypy/jit/tl/spli/interpreter.py): Reorder green vars. ... | 14:05 |
| Alex_Gaynor | arigato: is there anything left to do for blackhole-improvement, besides ensure all tests pass? | 14:06 |
| arigato | no | 14:07 |
| arigato | making sure that there is no performance drop | 14:08 |
| kenaan | 03arigo 10r7510814 pypy/branch/blackhole-improvement/pypy/jit 07M(jitprof.py jitoutput.py test_jitoutput.py): Fix for test_jitoutput. ... | 14:11 |
| arigato (~arigo@ratthing-216.openend.se) left irc: Quit: See you | 14:20 | |
| Moku (~John@f048243017.adsl.alicedsl.de) left irc: Ping timeout: 265 seconds | 14:29 | |
| Moku (~John@f054180163.adsl.alicedsl.de) joined #pypy. | 14:29 | |
| gonsor (~quassel@kel30.kel.stud.uni-goettingen.de) left irc: Ping timeout: 276 seconds | 14:37 | |
| Shanita (~John@f048103022.adsl.alicedsl.de) joined #pypy. | 14:40 | |
| Moku (~John@f054180163.adsl.alicedsl.de) left irc: Ping timeout: 260 seconds | 14:42 | |
| kenaan | 03hpk 10r7510914 07M(pypy/branch/lib-python/lib-python/conftest.py): new try to sanitize lib-python testing ... | 14:43 |
| copelco (~copelco@h-66-166-135-210.atlngahp.static.covad.net) joined #pypy. | 14:44 | |
| hpk | exarkun: i started a builder on the lib-python branch, got this: http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/620/steps/svn/logs/stdio - do you have any advise? | 14:45 |
| Shanita (~John@f048103022.adsl.alicedsl.de) left irc: Ping timeout: 240 seconds | 14:45 | |
| exarkun | That's new to me, sorry :/ | 14:46 |
| Shanita (~John@f049052223.adsl.alicedsl.de) joined #pypy. | 14:46 | |
| hpk | happened to me before when starting a builder for a branch, but i see no pattern | 14:47 |
| hpk | in the waterfall it is proceeding | 14:47 |
| hpk | can also not stop it | 14:47 |
| ambroff (~ambroff@65-113-99-90.dia.static.qwest.net) left irc: Quit: Ex-Chat | 14:57 | |
| ErKa (keryell@dhcp3.info.enstb.org) left irc: Ping timeout: 276 seconds | 14:58 | |
| e-max_ (~e-max@217.14.244.16) left irc: Ping timeout: 258 seconds | 15:06 | |
| kenaan | 03afa 10r7511014 07M(pypy/trunk/pypy/module/cpyext/api.py): Fix test failures. ... | 15:17 |
| kenaan | 03afa 10r7511114 pypy/trunk/pypy/module/cpyext 07M(pyobject.py typeobject.py): Little cleanup ... | 15:18 |
| arigato (~arigo@c-7649e353.022-54-67626719.cust.bredbandsbolaget.se) joined #pypy. | 15:19 | |
| panni_ (hannes@ip-95-222-52-93.unitymediagroup.de) joined #pypy. | 15:20 | |
| arigato | hpk: it's known that we do a full checkout instead of "svn switch" | 15:20 |
| arigato | I don't remember why -- there might be old reasons that no longer apply | 15:21 |
| amaury_ | svn switch does not always work | 15:22 |
| arigato | I think it does if we clean out all non-versionned files first | 15:23 |
| arigato | or are there more obscure cases? | 15:23 |
| hpk | arigato: the problem was a buildbot traceback, not knowing the build number - by now the traceback is gone so i guess it just takes a while | 15:24 |
| arigato | the traceback disappeared from a ../logs/stdio page? | 15:24 |
| hpk | yes | 15:25 |
| hpk | no | 15:25 |
| arigato | ??? | 15:25 |
| hpk | the url resulted in a traceback | 15:25 |
| Alex_Gaynor | arigato: getting a test fail in pypy/jit/tl/tla/test_tla.py, looks like it's an issue with ordering the greens | 15:25 |
| hpk | and now it doesn't | 15:25 |
| arigato | ah | 15:25 |
| arigato | ok | 15:25 |
| arigato | Alex_Gaynor: yes, fixed this afternoon :-) | 15:25 |
| Alex_Gaynor | arigato: hrm, I'm on HEAD, I wonder if it's a pyc issue | 15:26 |
| arigato | ah | 15:26 |
| arigato | maybe I missed that failure actually | 15:26 |
| arigato | there were more in pypy/jit/tl/spli/ | 15:26 |
| arigato | sorry | 15:27 |
| kenaan | 03arigo 10r7511214 07M(pypy/branch/blackhole-improvement/pypy/jit/tl/tla/tla.py): Reorder the green vars. ... | 15:27 |
| Alex_Gaynor | yay | 15:28 |
| arigato | well, then we have failures in rpython/memory/gc/ | 15:28 |
| Shanita (~John@f049052223.adsl.alicedsl.de) left irc: Remote host closed the connection | 15:33 | |
| Moku (~John@f049052223.adsl.alicedsl.de) joined #pypy. | 15:36 | |
| kenaan | 03hpk 10r7511314 07M(pypy/trunk/lib-python/conftest.py): try to redo the fix related to testing in lib-python. ... | 15:46 |
| kenaan | 03hpk 10r7511414 06D(pypy/branch/lib-python/): remove lib-python branch, seems buildbot jobs do fine ... | 15:46 |
| ErKa (keryell@keryell.pck.nerim.net) joined #pypy. | 15:55 | |
| kenaan | 03arigo 10r7511514 pypy/branch/blackhole-improvement/pypy 07M(builtin.py llinterp.py llmemory.py rffi.py rbuiltin.py test_rptr.py address.h): Refactor cast_adr_to_int(), step 1. ... | 16:00 |
| dmalcolm (~david@nat/redhat/x-qvcugsfcyzllxkkn) joined #pypy. | 16:02 | |
| kenaan | 03arigo 10r7511614 07M(pypy/branch/blackhole-improvement/pypy/rpython/lltypesystem/llarena.py): Fix. ... | 16:11 |
| dmalcolm (~david@nat/redhat/x-qvcugsfcyzllxkkn) left irc: Ping timeout: 240 seconds | 16:22 | |
| lucian (~lucian@94-193-66-109.zone7.bethere.co.uk) left irc: Ping timeout: 264 seconds | 16:33 | |
| lucian (~lucian@94-193-66-109.zone7.bethere.co.uk) joined #pypy. | 16:43 | |
| pem (~pem@2001:cc0:201e:107:221:86ff:fe1a:e5aa) left irc: Remote host closed the connection | 16:51 | |
| jcreigh | wait. Is the only thing that causes an exit of JIT code a guard failure? | 16:55 |
| gutworth | yep | 16:56 |
| Alex_Gaynor | indeed | 16:56 |
| Alex_Gaynor | oh and a segfault | 16:57 |
| jcreigh | you guys are clever. :) | 16:57 |
| jcreigh | haha | 16:57 |
| arigato | not completely | 16:57 |
| jcreigh | yeah, we'll just pretend that's a feature. (We could call it the "fast path") | 16:57 |
| gutworth | Alex_Gaynor: machine's refuse to segfault our jit | 16:57 |
| jcreigh | arigato: oh? | 16:57 |
| arigato | some paths are not loops (ending with JUMP) but just end in a FINISH | 16:57 |
| arigato | but indeed the common case is to have infinite loops ending in a guard failure | 16:58 |
| jcreigh | if they aren't loops, how are they ever JIT compiled in the first place? | 16:59 |
| arigato | it's for the case where the JIT has already compiled a loop, | 16:59 |
| arigato | then notices "oh this guard fails quite often, I'm going to compile from it" | 16:59 |
| arigato | then this compilation might not end in a loop at all | 16:59 |
| arigato | typical case when not only the run loops often, but is also often entered and left | 17:00 |
| arigato | I mean the loop is also often entered and left | 17:01 |
| jcreigh | hmm. | 17:02 |
| jcreigh | so what happens if the JIT guesses it's time to compile a loop at exactly the wrong time? eg, it decides to start tracing on the very last iteration of the loop. | 17:02 |
| jcreigh | does it just keep tracing until it hits some limit and gives up? | 17:02 |
| arigato | yes | 17:03 |
| gutworth | yes, there's a limit | 17:03 |
| gutworth | the function could also return | 17:03 |
| ambroff (~ambroff@38.99.63.41) joined #pypy. | 17:03 | |
| arigato | or it could also enter another loop, which the jit detects (then it throws away the start of the trace and keeps only the loop that it has really seen) | 17:03 |
| ErKa (keryell@keryell.pck.nerim.net) left irc: Ping timeout: 260 seconds | 17:07 | |
| bigdog (~scmikes@72-197-8-8-arpa.cust.cinci.current.net) joined #pypy. | 17:09 | |
| kenaan | 03arigo 10r7511714 pypy/branch/blackhole-improvement/pypy 07M(llimpl.py runner.py test_llgraph.py gc.py llmodel.py model.py runner_test.py test_ll_random.py assembler.py regalloc.py ri386.py assembler.py heaptracker.py jitcode.py jt ... | 17:25 |
| Alex_Gaynor | jesus that's a lot of whacking | 17:26 |
| bigdog1 (~scmikes@72-197-8-8-arpa.cust.cinci.current.net) joined #pypy. | 17:34 | |
| gutworth (~benjamin@64-131-1-181.usfamily.net) left irc: Ping timeout: 276 seconds | 17:34 | |
| bigdog (~scmikes@72-197-8-8-arpa.cust.cinci.current.net) left irc: Ping timeout: 240 seconds | 17:35 | |
| arigato | :-( | 17:37 |
| amaury_ | arigato: ? | 17:40 |
| arigato | just answering to Alex_Gaynor | 17:41 |
| arigato | now I'd love to understand how r74920 can cause test_finalizer in test_transformed_gc to fail | 17:42 |
| gutworth (~benjamin@64-131-1-181.usfamily.net) joined #pypy. | 17:46 | |
| amaury_ | yeah, the Boost.Python test suite starts to pass | 17:57 |
| amaury_ (~5bc3480e@gateway/web/freenode/x-cdxmljbnwgkznjyq) left irc: Quit: Page closed | 18:00 | |
| Trundle (~andy@p5B14D4E7.dip.t-dialin.net) joined #pypy. | 18:03 | |
| arigato | good :-) | 18:04 |
| arigato (~arigo@c-7649e353.022-54-67626719.cust.bredbandsbolaget.se) left irc: Quit: See you | 18:08 | |
| jcreigh | well, gdb's "reverse debugging" is all-but-useless on 64-bit, as libc uses some instructions that it does not support. | 18:40 |
| amaury_ (~5645c527@gateway/web/freenode/x-ophxivnzkuolduid) joined #pypy. | 18:49 | |
| kenaan | 03afa 10r7511814 07M(pypy/trunk/pypy/module/cpyext/test/test_longobject.py): Fix test on 64bit platforms ... | 18:50 |
| antocuni (~antocuni@host83-56-dynamic.14-87-r.retail.telecomitalia.it) left irc: Ping timeout: 276 seconds | 18:51 | |
| Alex_Gaynor | amaury_: could that be written assert module.from_unsignedlong() == sys.maxint - 1 ? | 18:51 |
| amaury_ | huh, probably | 18:52 |
| Alex_Gaynor | was just curious, I was afraid there might be somethign with C types and 128 bit systems or something :) | 18:52 |
| amaury_ | also, I expected the longlong tests to fail | 18:53 |
| amaury_ | is longlong 64bit as well on 64bit platforms? | 18:53 |
| Alex_Gaynor | I assume so, but I'm not sure | 18:53 |
| amaury_ | is longlong 64bit as well on 64bit platforms? | 18:55 |
| amaury_ | huh sorry | 18:55 |
| amaury_ | actually ulong(-1) == sys.maxint * 2 + 1 | 18:55 |
| kenaan | 03afa 10r7511914 07M(pypy/trunk/pypy/module/cpyext/test/test_longobject.py): Simplify test, thanks Alex ... | 18:56 |
| fiftysixquarters (~pillin@tshlabs.org) left irc: Ping timeout: 276 seconds | 19:46 | |
| kenaan | 03afa 10r7512014 pypy/trunk/pypy/module/cpyext 07M(sliceobject.py test_sliceobject.py): PySlice_New() allows NULLs. Test and fix. ... | 20:13 |
| dmalcolm (~david@nat/redhat/x-wephctpahpczjyot) joined #pypy. | 20:18 | |
| kenaan | 03afa 10r7512114 pypy/trunk/pypy/module/cpyext 07M(slotdefs.py test_typeobject.py): Implement the tp_str slot ... | 20:44 |
| pedronis (~pedronis@46-180.61-188.cust.bluewin.ch) left irc: Ping timeout: 265 seconds | 20:50 | |
| xorAxAx | fijal: btw, oanda seems to rock | 20:51 |
| fiftysixquarters (~pillin@tshlabs.org) joined #pypy. | 20:53 | |
| fiftysixquarters (~pillin@tshlabs.org) left irc: Client Quit | 20:53 | |
| DasIch (~DasIch@p5DC5FBC3.dip.t-dialin.net) left irc: Ping timeout: 252 seconds | 20:59 | |
| DasIch (~DasIch@p5DC5F174.dip.t-dialin.net) joined #pypy. | 20:59 | |
| Alex_Gaynor | fijal: oh I finally came up with a reason benchmark of django templates | 20:59 |
| Hory (~hory@78.96.96.85) left irc: Read error: Connection reset by peer | 20:59 | |
| Alex_Gaynor | basically US benchmark notwithstanding, both pypy and US are slower than Cpython on this | 21:00 |
| amaury_ | Boost.Python test suite: 47 passed, 36 failed | 21:07 |
| Alex_Gaynor | hey hey, not bad | 21:07 |
| Alex_Gaynor | can you paste the fails? | 21:07 |
| amaury_ | http://paste.pocoo.org/show/222042/ | 21:09 |
| ronny | Alex_Gaynor: how does django compare to jinja2 these days? | 21:13 |
| Alex_Gaynor | ronny: speedwise? god awful | 21:13 |
| ronny | heh | 21:14 |
| ronny | i wonder if django ever gets rid of that rotten batteries mentality | 21:14 |
| Hory (~hory@78.96.96.85) joined #pypy. | 21:17 | |
| Alex_Gaynor | ronny: not going to happen, ever. It's strongly viewed as being very advantageous in helping new uesrs get started | 21:17 |
| ronny | Alex_Gaynor: it also stringly helped me getting rid of django cause it couldnt get complex things done in sensible ways | 21:19 |
| Alex_Gaynor | ronny: :/ I've been thinking about a design for a new ORM, something to do with all the free time I have :) | 21:21 |
| ronny | Alex_Gaynor: there is sqlalchemy, it works | 21:21 |
| Alex_Gaynor | true, I want a hybrid of the two though :) | 21:21 |
| ronny | Alex_Gaynor: and all you really need to do is add django style query objects | 21:21 |
| ronny | im pretty sure you can do most of it by extending declarative a bit | 21:22 |
| Alex_Gaynor | ronny: that's not the parts I really want, some day I'll write up what I'm interested in seeing | 21:22 |
| Alex_Gaynor | for now I'll use django for some things and sqlalchemy for others :) | 21:22 |
| ronny | Alex_Gaynor: the thing is, sqlalchemy can do everything djangos orm cna do + a lot of awesome things | 21:22 |
| Alex_Gaynor | I don't disagree, but the API isn't as friendly for all things | 21:23 |
| Alex_Gaynor_ (~Alex_Gayn@76-206-247-179.lightspeed.cicril.sbcglobal.net) joined #pypy. | 21:23 | |
| ronny | Alex_Gaynor: so whats the issue with just adding some friendly api and being done? | 21:24 |
| ronny | no need to reimplement the really hard parts | 21:24 |
| Alex_Gaynor_ | writing things from scratch is fun :) | 21:24 |
| Alex_Gaynor_ | I wouldn't really expect anyone to use it, but it'd be fun | 21:24 |
| ronny | i see | 21:25 |
| ronny | hmm | 21:26 |
| kenaan | 03afa 10r7512214 pypy/trunk/pypy/module/cpyext 07M(slotdefs.py foo.c test_typeobject.py): implement wrap_descr_get, wrap_descr_set, wrap_descr_delete ... | 21:27 |
| ronny | i think i figured how to make a checkout api for anyvc | 21:27 |
| Alex_Gaynor (~alex@76-206-247-179.lightspeed.cicril.sbcglobal.net) left irc: Ping timeout: 265 seconds | 21:28 | |
| Nick change: Alex_Gaynor_ -> Alex_Gaynor | 21:28 | |
| ousado_ (~ousado@p5B3C0C06.dip0.t-ipconnect.de) joined #pypy. | 21:34 | |
| inhahe (inhahe@adsl-074-186-083-219.sip.mia.bellsouth.net) joined #pypy. | 21:35 | |
| gutworth | is it allowed to compare a unicode string to a str constant? | 21:36 |
| exarkun | allowed by whom? | 21:37 |
| Alex_Gaynor | exarkun: Rpython I suppose | 21:37 |
| gutworth | apparently it is | 21:37 |
| ousado (~ousado@p5B3C0B5C.dip0.t-ipconnect.de) left irc: Ping timeout: 252 seconds | 21:38 | |
| kenaan | 03getxsick 10r7512314 pypy/branch/fast-ctypes/pypy/rlib 07M(jitffi.py test_jitffi.py): support for non args (void) functions ... | 21:38 |
| Ademan | ugh, git svn feels so dirty... | 21:40 |
| gutworth | by defintion, it must; it has git | 21:40 |
| Alex_Gaynor | -1 gutworth | 21:40 |
| Ademan | gutworth: heh, i'm pretty sure the svn part is the problem :-p | 21:40 |
| Alex_Gaynor | svn is always the problem, the solution is to nuke it from orbit | 21:41 |
| exarkun | I'm sure there's enough blame to share. | 21:41 |
| Arfrever (~Arfrever@gentoo/developer/Arfrever) left irc: Quit: Ex+re (KVIrc 4) | 21:46 | |
| jcreigh | grr, Hisenbugs | 21:46 |
| kenaan | 03benjamin 10r7512414 07M(pypy/trunk/pypy/objspace/std/formatting.py): remove unused function ... | 21:48 |
| kenaan | 03getxsick 10r7512514 pypy/branch/fast-ctypes/pypy/rlib 07M(jitffi.py test_jitffi.py): support for functions which return void ... | 21:57 |
| Ademan | jcreigh: that's a great term haha | 22:00 |
| AndrewBC | grrr, trying to get pypy 1.2 to honor piping out or --output=FILE | 22:02 |
| AndrewBC | well, translate.py to be specific | 22:02 |
| Moku (~John@f049052223.adsl.alicedsl.de) left irc: Read error: No route to host | 22:05 | |
| Moku (~John@f049052223.adsl.alicedsl.de) joined #pypy. | 22:06 | |
| kenaan | 03getxsick 10r7512614 07M(pypy/branch/fast-ctypes/pypy/rlib/test/test_jitffi.py): more tests ... | 22:06 |
| gonsor (~quassel@p548A9179.dip0.t-ipconnect.de) joined #pypy. | 22:07 | |
| lucian (~lucian@94-193-66-109.zone7.bethere.co.uk) left irc: Quit: Leaving... | 22:11 | |
| kenaan | 03afa 10r7512714 pypy/trunk/pypy/module/cpyext 07M(test_typeobject.py typeobject.py): Small support for Py_TYPE(x)->tp_dict: ... | 22:27 |
| copelco (~copelco@h-66-166-135-210.atlngahp.static.covad.net) left irc: Ping timeout: 245 seconds | 22:33 | |
| amaury_ (~5645c527@gateway/web/freenode/x-ophxivnzkuolduid) left irc: Quit: Page closed | 22:38 | |
| gonsor (~quassel@p548A9179.dip0.t-ipconnect.de) left irc: Remote host closed the connection | 22:43 | |
| Moku (~John@f049052223.adsl.alicedsl.de) left irc: Ping timeout: 258 seconds | 23:12 | |
| Shanita (~John@f048167190.adsl.alicedsl.de) joined #pypy. | 23:12 | |
| Alex_Gaynor | I don't suppose anyone has ever tried to implement a JVM on pypy | 23:32 |
| Ademan | Alex_Gaynor: not yet :-) | 23:35 |
| AndrewBC (MGM@97.93.232.247) left irc: Quit: Bye! | 23:38 | |
| Shanita (~John@f048167190.adsl.alicedsl.de) left irc: Ping timeout: 248 seconds | 23:39 | |
| AndrewBC (~AndrewBC@97.93.232.247) joined #pypy. | 23:41 | |
| ambroff (~ambroff@38.99.63.41) left irc: Read error: Operation timed out | 23:53 | |
| --- Sat Jun 5 2010 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!