#pypy IRC log for Saturday, 2011-12-10

jaredw (~jared@122-59-91-107.jetstream.xtra.co.nz) joined #pypy.00:00
mtigas (~mtigas@users.spokesman.com) left irc: Quit: mtigas00:03
kennethreitz (~kennethre@204.14.152.118) left irc: Ping timeout: 240 seconds00:03
bgola (~bgola@c9518082.virtua.com.br) left irc: Remote host closed the connection00:04
Alex_Gaynorjterrace: I'd say i'm surprised by the spelling error, but I'm not even close :)00:07
jterracehehe00:07
jterracequestion for you00:07
jterraceis there a way to check if a space object is a string?00:07
harrison (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) joined #pypy.00:07
jterracespace.isinstance_w(w_item, space.w_str) works but the compile tests fail00:08
jterracebecause FakeSpace doesnt have w_str00:08
Alex_Gaynoradd w_str to FakeSpace then00:08
jterraceoh i wanst sure if that was just a missing type or if it wasnt there for a reason00:08
Alex_Gaynorit's the minimal thing to make those tests work00:09
jaredw (~jared@122-59-91-107.jetstream.xtra.co.nz) left irc: Remote host closed the connection00:09
jterraceok00:09
Alex_Gaynorno, nothing about it is intentional in that sense00:09
jterracei have some more changes for exposing uint dtypes (with tests), and i added the rest of the functionality to fromstring function, which involved changing the dtype coerce methods to handle string00:09
Alex_Gaynordo they not?00:10
jterracethey dont00:10
jterracee.g., np.int32('30') did not work00:10
jterracebecause space.int(space.wrap('3')) does not work00:10
harrison_ (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) joined #pypy.00:11
harrison (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) left irc: Ping timeout: 252 seconds00:12
jterraceAlex_Gaynor: did you get a chance to review my updated pull request for tolist()? i have another pull request ready to submit if you accept it :)00:14
Alex_Gaynorjterrace: sigh, I fail, will get to it tommorrow, right now I'm more tired than I ever have been before00:14
jterraceheh np00:15
jterracethe rest are sitting on jterrace/pypy for now00:15
Alex_Gaynork00:15
kennethreitz (~kennethre@204.14.152.118) joined #pypy.00:15
bbot24Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/16500:26
Rotund (~joe@184-158-101-126.dyn.centurytel.net) joined #pypy.00:33
RotundHi guys. Do you have a port of pylibmc for pypy? Or do you suggest another memcache module?00:34
voidspace (~anonymous@python/psf/voidspace) joined #pypy.00:38
JaRoel|4d (~jaroel|4d@office.fourdigits.nl) left irc: Remote host closed the connection00:43
jterrace (~jterrace@aegis.CS.Princeton.EDU) left irc: Quit: Leaving.00:46
linq (~ident@24-246-25-39.cable.teksavvy.com) joined #pypy.00:50
pjenveyhiphop VM, huh?00:52
rokujyouhitoma (~rokujyouh@173.227.61.3) left irc: Remote host closed the connection01:03
Nick change: Gulaway -> Gulopine01:03
rokujyouhitoma (~rokujyouh@173.227.61.3) joined #pypy.01:03
overminder (overminder@2002:af9f:6c2e::af9f:6c2e) joined #pypy.01:17
nedbat (~nedbat@python/psf/nedbat) joined #pypy.01:18
bbot24Failure: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/73901:19
JaRoel|4d (~jaroel|4d@sink.jaroel.nl) joined #pypy.01:20
squiddy (~squiddy@f053084199.adsl.alicedsl.de) left irc: Quit: Leaving01:20
papercraneRotund: there's the python version, which should work, but it has some unfortunate problems under heavy load01:22
ericflo (~ericflo@75.103.8.110) left irc: Quit: ericflo01:27
bbot24Failure: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/97301:28
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 258 seconds01:31
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/188301:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/62101:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/63701:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/115201:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/82201:36
bbot2Started: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/74401:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/146401:36
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/30001:38
voidspace (~anonymous@python/psf/voidspace) left irc: Quit: voidspace01:41
Rotundpapercrane, thanks.01:56
Vorpal (~AnMaster@unaffiliated/anmaster) left irc: Ping timeout: 244 seconds01:57
nedbat (~nedbat@python/psf/nedbat) joined #pypy.02:00
rokujyouhitoma (~rokujyouh@173.227.61.3) left irc: Remote host closed the connection02:01
Vorpal (~AnMaster@unaffiliated/anmaster) joined #pypy.02:03
derdon (~derdon@pD9E1CE32.dip.t-dialin.net) left irc: Remote host closed the connection02:06
bbot213Exception: 15http://buildbot.pypy.org/builders/own-win-x86-64/builds/81 [12ctismer, win64-stage1]02:11
Ademan (~dan@adsl-71-141-224-79.dsl.snfc21.pacbell.net) joined #pypy.02:13
papercrane (~papercran@75.101.111.82) left irc: Ping timeout: 240 seconds02:22
harrison_ (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) left irc: Ping timeout: 252 seconds02:29
harrison (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) joined #pypy.02:33
gutworth (~benjamin@dhcp-128-36-84-91.central.yale.edu) joined #pypy.02:34
mikefcWould it be possible to write image format interpreters in rpython and compile them to a JIT?02:34
Action: mikefc is still hazy on everything under pypy's hood02:35
Action: gutworth apparently runs a grad student tutoring charity02:35
gutworthyes02:36
gutworthbut you could write them in python02:36
gutworthand get them jitted :)02:36
harrison_ (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) joined #pypy.02:37
harrison (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) left irc: Ping timeout: 268 seconds02:38
mikefcgutworth: yeah.  but if i went one level down and treated the image as the program, and wrote the interpreter directly in rptyhon?  02:39
Action: mikefc has too many stupid ideas and not enough time02:39
Kaskuka (~John@osbk-4d08b5d8.pool.mediaWays.net) joined #pypy.02:40
Guest95497 (~John@osbk-4db17084.pool.mediaWays.net) left irc: Ping timeout: 240 seconds02:40
gutworthyou could do it02:42
gutworthso I suggest you use your time efficiently and write it in python02:42
mikefcyeah :)02:42
linq (~ident@24-246-25-39.cable.teksavvy.com) left irc: Ping timeout: 252 seconds02:43
bbot24Failure: 15http://buildbot.pypy.org/builders/own-linux-x86-32/builds/188302:45
Vorpal (~AnMaster@unaffiliated/anmaster) left irc: Ping timeout: 248 seconds02:51
gtaylor (~gtaylor@108-196-160-120.lightspeed.tukrga.sbcglobal.net) left irc: Quit: Konversation terminated!02:58
mikefcif it's his semester research project, it sure sounds like he hasn't done any work at all so far..03:03
harrison_ (~quassel@adsl-76-217-38-197.dsl.chcgil.sbcglobal.net) left irc: Ping timeout: 252 seconds03:11
asenchi (~asenchi@206.173.142.114) left irc: Ping timeout: 260 seconds03:16
kalenz (~kalenz@ping.kalenz.fr) left irc: Read error: Operation timed out03:29
kalenz (~kalenz@ping.kalenz.fr) joined #pypy.03:30
Nick change: kalenz -> Guest1068303:30
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.04:00
bbot24Finished ??: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/62104:15
bbot24Finished ??: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/63704:15
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/62204:15
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/63804:15
overminder (overminder@2002:af9f:6c2e::af9f:6c2e) left irc: Ping timeout: 252 seconds04:20
Rotund (~joe@184-158-101-126.dyn.centurytel.net) left irc: Ping timeout: 255 seconds04:21
Rotund (~joe@184-158-101-126.dyn.centurytel.net) joined #pypy.04:24
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) left irc: Quit: Computer has gone to sleep.04:28
Count_Niedar (gfghfghf@host-173-230-2-133.vanodod.clients.pavlovmedia.com) left irc: Quit: ( www.nnscript.com :: NoNameScript 4.22 :: www.esnation.com )04:31
Niedar (~dgdfg@host-173-230-2-190.vanodod.clients.pavlovmedia.com) joined #pypy.04:37
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/115204:37
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 255 seconds04:42
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-Ojit-no-jit-linux-x86-32/builds/82204:47
Arach (~arach@2.95.83.234) joined #pypy.04:55
Niedar (~dgdfg@host-173-230-2-190.vanodod.clients.pavlovmedia.com) left irc: Quit: Leaving05:10
gutworth (~benjamin@dhcp-128-36-84-91.central.yale.edu) left irc: Ping timeout: 240 seconds05:17
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.05:23
media-498 (~media-498@unaffiliated/firippu) joined #pypy.05:23
media-498 (~media-498@unaffiliated/firippu) left irc: Read error: Connection reset by peer05:58
Rotund (~joe@184-158-101-126.dyn.centurytel.net) left irc: Ping timeout: 252 seconds05:59
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) left irc: Quit: Computer has gone to sleep.06:01
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-64/builds/63806:20
fprimex (~brent@fl-67-233-144-239.dhcp.embarqhsd.net) left irc: Quit: stuff!06:21
ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) joined #pypy.06:28
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/62206:42
bbot23Success: 15http://buildbot.pypy.org/builders/own-linux-x86-64/builds/74406:42
apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Remote host closed the connection06:56
apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) joined #pypy.06:56
apanda (~apanda@dhcp-44-187.EECS.Berkeley.EDU) left irc: Ping timeout: 255 seconds07:01
apanda (~apanda@c-24-7-101-62.hsd1.ca.comcast.net) joined #pypy.07:16
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.07:18
rokujyouhitoma (~rokujyouh@12.51.221.130) joined #pypy.07:22
Nick change: Gulopine -> Gulaway07:35
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) left irc: Quit: Computer has gone to sleep.07:39
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-app-level-linux-x86-32/builds/146407:45
asmeurer__ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) left irc: Quit: asmeurer__07:59
fijal (~fijal@196.209.233.36) joined #pypy.08:12
fijal (~fijal@196.209.233.36) left irc: Read error: Connection reset by peer08:15
`fox` (~fox@host91-171-dynamic.245-95-r.retail.telecomitalia.it) joined #pypy.08:21
fijal (~fijal@196.209.233.36) joined #pypy.08:36
fijal (~fijal@196.209.233.36) left irc: Read error: Connection reset by peer08:39
stakkars_ (~tismer@p5DDB72FD.dip.t-dialin.net) joined #pypy.08:45
stakkars (~tismer@p5DDB7E68.dip.t-dialin.net) left irc: Ping timeout: 252 seconds08:49
Nick change: stakkars_ -> stakkars08:49
fijal (~fijal@196.209.233.36) joined #pypy.08:56
squiddy (~squiddy@f053082034.adsl.alicedsl.de) joined #pypy.09:02
arigato (~arigo@89.204.154.165) joined #pypy.09:06
arigatooups.  everything is broken09:06
fijal:/09:07
fijalarigato: good morning armin09:07
fijal:-)09:07
arigatohi :-)09:07
arigatoI'm annoyed that we are not getting any benchmark since 2 days, but checking in at least 3 different things whose speed influence we'd like to know09:08
arigatoah wrong, we got the automatic results yesterday09:09
arigatobut all the manual {benchmark-run} I started failed09:09
fijal"good"09:10
fijalnot so good09:10
antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) joined #pypy.09:16
kenaan12arigo temp2 112db53dd9c46e 15/: close temporary branch09:18
kenaan12arigo default 11b8e86a5f522b 15/pypy/module/imp/importing.py: Baaaaah.  Don't rely on "is" being true for two independently wrapped strings.  It fails on py.py, for example.09:18
kenaan12arigo default 11998ae4038c2f 15/pypy/translator/goal/: fixes for test_app_main.09:23
arigato"oh well"09:26
arigatothe point is that I believe that we have now reached the "5x" goal again09:26
arigatobut I'm annoyed because I cannot see it :-)09:26
fijal (~fijal@196.209.233.36) left irc: Ping timeout: 240 seconds09:30
rokujyouhitoma (~rokujyouh@12.51.221.130) left irc: Remote host closed the connection09:35
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-macosx-x86-64/builds/30009:42
pedronis_ (~pedronis@73-53.195-178.cust.bluewin.ch) joined #pypy.10:08
pedronis (~pedronis@73-53.195-178.cust.bluewin.ch) left irc: Read error: Connection reset by peer10:08
setmeaway2 (oosool3@119.201.52.190) left irc: Quit: Leaving10:12
setmeaway (~setmeaway@119.201.52.190) joined #pypy.10:12
durin42 (~durin@adium/durin42) left irc: Ping timeout: 240 seconds10:18
durin42 (~durin@adium/durin42) joined #pypy.10:20
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.10:24
`fox` (~fox@host91-171-dynamic.245-95-r.retail.telecomitalia.it) left irc: Ping timeout: 244 seconds10:26
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Ping timeout: 244 seconds10:29
danchr (~danchr@cl-848.chi-02.us.sixxs.net) left irc: Ping timeout: 252 seconds10:31
kennethreitz (~kennethre@204.14.152.118) left irc: Quit: Computer has gone to sleep.10:32
danchr (~danchr@cl-848.chi-02.us.sixxs.net) joined #pypy.10:35
xiaochen (~chobits@123.125.146.24) joined #pypy.10:47
kenaan12arigo default 11386488dbe96d 15/pypy/module/imp/test/test_import.py: Skip these tests in py.test -A.10:49
kenaan12hakanardo jit-targets 1151117c91db41 15/pypy/module/pypyjit/test_pypy_c/test_call.py: fix test10:56
kenaan12hakanardo jit-targets 11b63f055c491d 15/pypy/module/pypyjit/test_pypy_c/test_00_model.py: at bit more logging10:56
kenaan12hakanardo jit-targets 1138465568087c 15/: hg merge 8de6f245c95910:57
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/623 [12hakanardo, jit-targets]10:58
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 268 seconds11:03
fijal (~fijal@41.51.194.164) joined #pypy.11:08
fijal (~fijal@41.51.194.164) left irc: Read error: Connection reset by peer11:11
derdon (~derdon@p5DE89DB3.dip.t-dialin.net) joined #pypy.11:11
arigato (~arigo@89.204.154.165) left irc: Ping timeout: 252 seconds11:12
fijal (~fijal@41-133-220-93.dsl.mweb.co.za) joined #pypy.11:12
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.11:14
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 244 seconds11:20
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/623 [12hakanardo, jit-targets]11:21
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.11:22
lucian (~lucian@cpc1-newc15-2-0-cust84.gate.cable.virginmedia.com) joined #pypy.11:29
kenaan12hakanardo jit-targets 118474a7603bbe 15/pypy/: hg merge default11:29
kenaan12hakanardo jit-targets 11b08cf0eb7990 15/pypy/jit/backend/x86/regalloc.py: fix11:31
voidspace (~voidspace@python/psf/voidspace) joined #pypy.11:38
kkris (~kris@80-123-38-195.adsl.highway.telekom.at) joined #pypy.11:42
nedbat (~nedbat@python/psf/nedbat) joined #pypy.11:45
unbit (~unbit@93-58-162-185.ip159.fastwebnet.it) left irc: Read error: Connection reset by peer11:51
unbit (~unbit@93-58-162-185.ip159.fastwebnet.it) joined #pypy.11:51
DasIch_ (~dasich@p4FFDCD7B.dip.t-dialin.net) left irc: Read error: Connection reset by peer11:52
ojii (~ojii@22-14.62-81.cust.bluewin.ch) joined #pypy.11:53
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.12:09
lvh (~lvh@python/psf/lvh) joined #pypy.12:16
lvhHey :)12:17
lvhSo I found something out recently and I'm asking some of my old colleagues nicely if I can write a report on this12:17
ronnyyo lvh 12:18
lvhMy old day job did some heavy GIS number crunching; everything was in Python, but there were large parts that had to be rewritten in some god-awful mix of Fortran, MATLAB, ASM and C because everything else was too slow12:18
lvhExcept that now it's not, and they're throwing out crazy amounts of my old Fortran code12:18
lvhso I'd like to say 1) yay pypy, 2) yay TDD and crazy extensive test suites12:18
ronnysounds good, when snakes eat fortran there is win12:19
lvhronny: Unfortunately they're not ASM free yet12:21
ronnylvh: one thing after another, right?12:21
lvhronny: But that's because we have this ancient packet radio standard to support, and the hardware that supports it is something that fits in an SGI server from the nineties12:22
lvhAnd the driver for that is ASM.12:22
lvhat least it's MIPS ASM, I guess :)12:22
ronnylol12:22
lvhhey, it works!12:22
ronnymips asm is fun, writing a os for a mips thing isnt12:22
lvhThat thing still runs the latest version of IRIX it supported, so its' not that bad12:23
lvhwhatever, we whistle bytes into it over serial port and bytes come back12:23
lvhafter that it's basically a black box coffee table to me12:23
lvh(the fun part is that it actually literally is a coffee table now)12:23
ronnyheh12:23
norox (~norox@2a00:801::2d0:b7ff:fea0:c753) joined #pypy.12:24
ronnylvh: whats a good place to discuss the async pep?12:27
lvhronny: no idea12:28
lvhpython-dev maybe?12:28
lvhit's on hiatus, I don't have time to work on it, sorry12:28
lvhstartup == too much work12:28
ronnylvh: i see12:29
lvhronny: I also had to cancel the talk because I have to be in Chicago. During that day of Pycon.12:31
ronnyok12:32
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.12:33
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer12:33
lvhI'm not entirely happy with that but unfortunately I don't have a lot of options. I guess the github issue tracker is a good place to vent grievances12:33
lvhif I had time to monitor and reply to everything I'd say python-dev12:33
khs (~khs@2001:700:300:2120:725a:b6ff:fee5:a44) joined #pypy.12:33
ronnylvh: i think i'll handle the stuff in detail after my thesis12:35
lvhronny: when's that12:35
fijalo/12:44
fijallvh: yay!12:44
lvhfijal: Less fortran!12:45
lvhfijal: Apparently version+1 is going to be 100% fortran free; they're even throwing the compiler out from the build machines12:45
lvhdefinitely a good day :)12:45
fijalcool!!!!!12:45
lvhC with embedded ASM is probably going to stay for a bit longer though12:46
lvh(change is dangerous)12:46
fijalright12:47
fijaland it's hard to test as well12:47
lvhwell, we have functional tests12:47
fijallvh: btw, you can always hire me as a consultant to make stuff faster12:47
lvhthe problem is that they take 2 days to run12:47
lvhfijal: I would; but I don't work there anymore :)12:48
fijalhaha :)12:48
fijalwhat do you do these days?12:48
lvhfijal: startup12:50
lvhfijal: so let's go with "working until my eyes pop out of their sockets, pop them back in, work some more"12:51
fijalokey12:51
Action: fijal does not do that12:51
lvhfijal: hence why we're not hiring you yet ;)12:52
fijalyeah, makes sense12:54
lvhfijal: Maybe we will eventually; we do some big data some more and man I don't want to write java :(12:54
voidspace (~voidspace@python/psf/voidspace) left irc: Read error: Connection reset by peer12:55
voidspace (~voidspace@python/psf/voidspace) joined #pypy.12:55
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer12:58
khs (~khs@2001:700:300:2120:725a:b6ff:fee5:a44) left irc: Read error: No route to host13:03
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.13:04
kenaan12hakanardo jit-targets 11f78f16899afb 15/pypy/jit/metainterp/: fall back on jumping to preamble if retracing fails13:04
fijallvh: haha13:09
lvh(for all the nasty things I can say about java, *man* the recent versions of the JVM are fast.)13:11
fijalnote that Java has not much to do with JVM13:12
fijalI would personally probably just emit JVM bytecode13:12
fijalit's more dynamic13:12
lvhwell, I meant that recent versions of the JVM can run java pretty damn fast13:13
lvhmoreso it seems than other jvm-based languages13:14
fijalyeah13:14
fijalbut other jvm-based languages are not that great at compiling to jvm13:14
fijalanyway13:14
fijalit kinda proves that the jvm-based approach kinda fails13:15
fijallvh: we can compete though :)13:15
lvhdefinitely13:15
lvhmostly because your language isn't as bad13:15
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer13:26
fijalnedbat: ping?13:29
nedbatfijal: hey13:31
fijalnedbat: can you reply on an issue?13:31
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.13:31
fijalyou'll know13:31
nedbatsure, where?13:31
fijalhttps://bugs.pypy.org/issue94913:31
nedbatsigh: invalid cert, why does firefox refuse to do anything?13:32
gsneddersBecause most people don't understand what an invalid cert is, and it's better to fail-secure than fail-insecure13:33
khs (~khs@2001:700:300:2120:725a:b6ff:fee5:a44) joined #pypy.13:33
Action: fijal is not convinced13:33
fijalgsnedders: hi13:33
fijalnedbat: just accept it13:33
voidspacebut is it better to succeed insecure - for a resource that doesn't *need* to be secure - or fail secure13:33
fijalgsnedders: it's not like I can easily say "this cert is fine"13:33
nedbatI know: chrome lets me do that, firefox provides no mechanism.13:33
fijalhow do you make it work in chrome?13:34
nedbatchrome gives me a dialog, with a button, Proceed Anyway.13:35
fijalright13:35
nedbatfijal: it doesn't matter, i'm reading it in Chrome.  You're right, I know all about this.13:35
fijalbut it does not give you a way to say "I'm fine with this cert"13:36
fijalI won't pay XX$/year for the certificate which is equally valid, just browsers are happier13:36
gsneddersvoidspace: Trusting the user to make the right decision often fails. And I'm pretty sure Fx does have a UI for it...13:36
fijalI'm fine if anyone else feels like sponsoring it13:36
gsneddersUnder the, "I Understand the Risks" of the Fx page you can add an exception.13:37
nedbatgsnedders: I don't get a page in Fx, maybe an option I've turned off?13:38
CIA-1303fijal 07roundup * 10#827/import xml.etree.cElementTree fails: [resolved] I'm closing this issue, feel free to reopen if it still causes issues * 14https://bugs.pypy.org/issue82713:39
nedbatfijal: my branch has fixes for a few of the problems this bug reports. Do you want to merge it?  There are tests. :)13:39
fijalnedbat: I won't merge it now, but yeah, I can13:39
fijalnedbat: like evening/tomorrow (no promises though)13:39
gsneddersnedbat: This is in Fx11, been a while since I've used anything older.13:40
fijalnedbat: feel free to answer to him "yes, it's broken on 1.7, we have some fixes, not yet merged"13:40
nedbat11? I have 8...13:40
gsneddersI get it in Fx8 too.13:40
nedbatfijal: ok, great. 13:40
xiaochen (~chobits@123.125.146.24) left irc: Quit: leaving13:40
gsneddersnedbat: i.e., what is currently in mozilla-central, effectively trunk13:41
nedbatgsnedders: what happened to 9+10?  I don't understand version numbers any more.13:41
nedbatfijal: should I send a pull request?13:41
fijalnedbat: version numbers are so yesterday13:41
fijalnedbat: you don't have to, it's on nedbat-sandbox right?13:41
nedbatfijal: correct13:41
fijalyeah, no point13:42
gsneddersnedbat: 9 is currently in beta13:42
gsneddersnedbat: 10 is currently in "aurora" (practically alpha)13:42
gsneddersnedbat: and then 11 is the current development tree13:42
Action: fijal thinks it's heavy nonsense13:42
gsneddersBasically you have four branches (release + those) and every six weeks there's a new release, so the current beta becomes release, aurora becomes beta, and a new aurora is branched off from the main development tree.13:43
CIA-1303nedbat 07roundup * 10#949/1.7 Sandbox problems: 13:55
CIA-13There are a few separate problems mentioned here. I have a branch (nedbat-sandbox) with13:55
CIA-13fixes for a few of them:13:55
CIA-13- The gcc invocation has to ... * 14https://bugs.pypy.org/issue94913:55
CIA-1303nedbat 07roundup * 10#949/1.7 Sandbox problems: 13:59
CIA-13I forgot to mention: fijal has said he will merge the nedbat-sandbox branch13:59
CIA-13soonish... * 14https://bugs.pypy.org/issue94913:59
kenaan12fijal default 1122d20cc8666b 15/lib_pypy/distributed/socklayer.py: Fail slightly bettter than a SyntaxError14:02
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer14:06
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.14:06
derdon (~derdon@p5DE89DB3.dip.t-dialin.net) left irc: Remote host closed the connection14:07
hakanardois test_pypy_c currently broken: http://paste.pocoo.org/show/518812/14:10
fijalhakanardo: armin broke stuff14:13
hakanardook14:14
arigato (~arigo@fwstups.cs.uni-duesseldorf.de) joined #pypy.14:14
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 248 seconds14:15
arigatohakanardo__: sorry, checkins from yesterday and today are breaking in the jit-targets branch14:19
arigatoI can have a look14:19
hakanardoarigato: thanx14:20
arigatoI doubt that just changing the assert is enough14:20
abki_hi14:20
arigatohi14:20
abki_I have troubles running pypy on my machine14:20
abki_it says ./bin/pypy: error while loading shared libraries: libbz2.so.1.0: cannot open shared object file: No such file or directory14:20
hakanardook, is default currently working again?14:20
arigatoI think so, yes14:20
abki_I assume the sole way to solve this is installed this version for libbz2 ?14:21
arigatoabki_: yes, or hack and put a symlink and hope, if you have a different version of libbz2.so.1.x14:21
Nick change: pedronis_ -> pedronis14:21
hakanardoarigato: cool, I want to run benchmarks to compare default and jit-targets14:22
hakanardoif they turn out ok I think it could be merged...14:22
arigatocool14:22
arigatoindeed, x86/test/test_runner.py explodes on the line following the assert14:23
arigatothe x86 backend in jit-targets really needs clean-up14:24
arigatoe.g. we cannot have a JUMP going to a loop any more, so a few attributes like looptoken._x86_arglocs should be killed14:24
abki_I only have libbz2.so and symlinking doesn't work14:24
arigatoabki_: you need to put the symlink in the same /usr/lib/ directory14:25
abki_arigato: that's what I've done, I got a "invalid elf header"14:25
arigato(usual mess of binary distributions on linux)14:25
arigatono clue then, try installing libbz2 from source14:25
abki_already tried that it conflicts with some packages I don't want to break my system14:26
abki_I build from source14:26
hakanardoarigato: I think the logic in pyjitpl and be simplified and clean up quite a bit too once multiple intermediate labels are supported14:27
hakanardobut it will alter it's behaviour  a bit too so its for a separate branch14:28
fijal (~fijal@41-133-220-93.dsl.mweb.co.za) left irc: Ping timeout: 255 seconds14:30
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer14:31
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.14:31
harrison (~quassel@adsl-69-209-194-17.dsl.chcgil.sbcglobal.net) joined #pypy.14:32
abki_the translation process seems to have discovered every bits it needs14:33
harrison_ (~quassel@adsl-69-209-194-17.dsl.chcgil.ameritech.net) joined #pypy.14:36
overminder (~overminde@143.89.167.3) joined #pypy.14:36
harrison (~quassel@adsl-69-209-194-17.dsl.chcgil.sbcglobal.net) left irc: Ping timeout: 240 seconds14:36
arigatohakanardo: yes14:38
arigatosame for the backend14:38
kenaan12arigo jit-targets 119addc6b0c977 15/pypy/jit/backend/: Fix.14:39
mfoord (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy.14:42
mfoord (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host14:42
mfoord (~anonymous@python/psf/voidspace) joined #pypy.14:42
arigatohakanardo: also, I think it would be relatively simple after jit-targets to change the way CALL_ASSEMBLER is emitted14:44
arigatoto pass only 4 or 5 arguments instead of all the stuff from the virtualizable14:44
arigatoand have the called loop start by reading things out of the virtualizable14:44
arigatobefore the first LABEL14:45
Alex_Gaynorarigato: will we still need to allocate the frame object?14:45
Vorpal (~AnMaster@unaffiliated/anmaster) joined #pypy.14:45
hakanardoarigato: cool14:46
arigatoAlex_Gaynor: yes, that's independent14:46
Alex_Gaynorarigato: PyPy: the art of keeping messes independent14:46
arigatohakanardo: then we can kill execute_token() and set_future_variable() and friends14:47
arigatoand from the interpreter we can just call the generated assembler like CALL_ASSEMBLER does14:47
arigatobecause it will then have a known fixed number of arguments14:47
hakanardonice14:47
arigato(I started in jit-simplify-backendintf long ago already, but was stopped by the fact that I'd then need to generate an extra loop just to read out of the virtualizable and jump to the real loop)14:49
hakanardoand without labels that would generate a lot of mov's14:50
arigatoyes14:50
mfoord (~anonymous@python/psf/voidspace) left irc: Quit: mfoord14:50
Action: arigato looks at failures in jit-targets/.../x86/test14:50
mfoord (~anonymous@87-194-212-65.bethere.co.uk) joined #pypy.14:52
mfoord (~anonymous@87-194-212-65.bethere.co.uk) left irc: Changing host14:52
mfoord (~anonymous@python/psf/voidspace) joined #pypy.14:52
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.14:55
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer14:55
kenaan12arigo jit-targets 1195089d049fe7 15/pypy/jit/backend/x86/: Reset the targettoken's _x86_loop_code attribute between tests.14:59
arigatodone, the rest I think is also failing in metainterp/test/14:59
arigatobut test_ztranslation works, so pypy has chances to translate15:00
Alex_Gaynorarigato: the comment above TargetToken._x86_loop_code = 0 should be updated I think15:00
arigatothanks15:01
arigatoif you had an idea to avoid the hack altogether it would be even better :-)15:01
Alex_GaynorI haven't even investigated what the hack is :)15:01
kenaan12arigo jit-targets 116c2822188033 15/pypy/jit/backend/x86/regalloc.py: Update comment.15:01
arigatowe use this default value of 0 to know if the JUMP at the end of the current loop goes back to a LABEL in the same loop (so not compiled yet) or not15:02
Alex_Gaynorthis determines whether we need to clean up the stack or not?15:03
arigatono15:03
arigatothis is for "hint_frame_positions": to know when we can initialize this dict15:03
arigatoif we JUMP somewhere else, it can be initialized before assembling the loop's operations15:04
arigatobut if we JUMP back to the middle of the same loop, it can only be initialized when we assemble the corresponding LABEL15:04
Alex_GaynorI haven't had enough coffee or tea this morning I think :)15:05
arigato:-)15:05
arigatosorry for giving a bit obscure explanations15:05
Alex_Gaynorno worries, I need to catch a train anyways15:05
Alex_Gaynorbye15:05
arigatosee you15:05
kenaan12alex_gaynor default 116c4b9c3bc0d6 15/pypy/module/micronumpy/test/test_zjit.py: some of these generate getfield_gc_pures now15:06
hakanardoarigato: I think metainterp/test/ passes except for test_ztranslation15:07
hakanardobut I'll rerrun to make sure...15:08
kenaan12arigo jit-targets 11959af14f07a4 15/pypy/jit/backend/x86/regalloc.py: Comments.15:08
arigatohakanardo: I think the merge added a few extra tests using the old interface15:08
hakanardoright, I fixed them15:09
hakanardo(but did not push)15:09
arigatoah .-)15:09
arigatook15:09
hakanardoand now I had to merge twice...15:09
kenaan12hakanardo jit-targets 11a1c0dc4d1391 15/pypy/jit/metainterp/test/test_ajit.py: fix+15:10
kenaan12hakanardo jit-targets 11f95dbc2efd61 15/pypy/jit/metainterp/test/: port tests15:10
kenaan12hakanardo jit-targets 117c782d5ad65b 15/pypy/jit/backend/: hg merge15:10
kenaan12hakanardo jit-targets 118164cef19074 15/pypy/jit/backend/x86/regalloc.py: hg merge15:10
arigatosorry :-)15:10
hakanardonp :)15:10
hakanardowould you like to look at test_ztranslation in metainterpreter aswell?15:11
hakanardoI have no clue about whats going on there...15:11
arigatook15:11
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer15:11
hakanardosomething blows up in llinterp15:12
arigatowhich test?15:12
hakanardoboth of them...15:12
arigatook15:12
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.15:12
hakanardothanx15:13
hakanardonote that translating pypy using translate,py -Ojit does work15:13
arigatogood :-)15:14
hakanardo(or atleast it did prior to merging default today)15:14
cheater (~cheater@ip-80-226-24-11.vodafone-net.de) joined #pypy.15:16
cheaterhi15:16
cheateri understand pypy has a special version of etree, different from the one available via pip install.. is that correct?15:16
cheateri mean that it also supports different (more?) syntax15:16
kenaan12arigo jit-targets 110da687824062 15/pypy/jit/backend/x86/test/test_recompilation.py: Fix: read 'frame_depth' before assembling the bridge.15:16
cheateri am looking to extend etree to support a little more xpath syntax, however i don't know where i would go with that15:18
arigatocheater: no15:18
arigatowe don't have the version that is written in C in CPython,15:19
arigatoand instead we hacked so that if you try to import it,15:19
arigatoyou end up with the Python version15:19
cheaterarigato: there's a C version of it? i thought it was all Python15:19
arigatocheater: no, there is a C version in CPython15:19
cheaterwhere is the .so?15:19
cheaterjust out of curiosit]15:19
cheatery15:19
arigato_elementtree.so15:20
Shanita (~John@osbk-4db06a58.pool.mediaWays.net) joined #pypy.15:20
Kaskuka (~John@osbk-4d08b5d8.pool.mediaWays.net) left irc: Ping timeout: 240 seconds15:20
exarkuncheater: Are you going to maintain a fork of elementtree?15:20
cheaterexarkun: i was hoping to fix a bug (which according to warnings emitted by etree is in the pipeline somehow) and possibly add a selector15:21
cheaterexarkun: i think those could easily go back into the main tree.15:21
cheaterbut if not i'll have a fork.15:22
exarkunIs it a divergence between PyPy and CPython?15:22
cheaterfrom what i see, and i might be doing something stupid, but the pypy one is version 1.7 with vastly different source, whereas the one my cpython has is 1.3 i believe15:22
fijal (~fijal@197.175.204.67) joined #pypy.15:22
Action: fijal is helping to build pypy debian packages15:23
cheateroh no, this one is 1.2.615:23
cheaterthe cpython one that is.15:23
exarkuncheater: The main reason I'm asking is the first paragraph of http://effbot.org/zone/element-xpath.htm15:23
cheaterit's not a divergence, it's just an extension of the xpath support15:24
exarkun"a full XPath engine is outside the scope of the core library."15:24
cheateri'm not sure what you are getting at15:25
cheaterare you saying that because someone who first wrote the library never wanted to bother adding full xpath support, i shouldn't add support for one minor piece of syntax, and etree should forever remain without this support?15:26
exarkunI just thought you might not have known that upstream doesn't seem very interested in xpath.  Maybe you knew already, and maybe upstream would be happy to take this patch.  I don't know.15:26
exarkuncheater: No, I just wanted to save you some time.15:26
cheateri've read that one already15:26
cheaterhowever, etree is the only thing i can find that has xpath support at all15:26
exarkunI pointed out what I wanted to point out, so I don't see any reason for the conversation to continue any further.15:26
cheaterunless you can suggest something else?15:27
exarkunlxml has very good xpath support15:27
DasIch (~dasich@p4FFDCD7B.dip.t-dialin.net) joined #pypy.15:27
cheateri was not aware lxml works under pypy at all?15:27
cheaterdoes it?15:27
jterrace (~Adium@dynamic-oit-vapornet-b-25.Princeton.EDU) joined #pypy.15:27
exarkunI don't think it does.15:27
cheaterso we're back to the situation of me needing to extend etree somehow.15:27
exarkunYou could work on getting lxml to work with PyPy, instead.15:28
jterracewhat are you trying to add to etree?15:28
exarkunThat would be a much better outcome, in my opinion. :)15:28
cheaterlxml is completely based on libxml2.15:28
cheaterwhich is a wrapper around a C lib.15:28
cheateri understand pypy doesn't support c libs at all.15:28
arigato?15:28
exarkunPyPy supports C libraries very well.15:28
cheatergetting libxml in its whole to work seems much more difficult than getting etree to work15:29
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer15:29
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.15:29
exarkunThat may be true.15:29
cheaterwhere the second "work" just means adding or fixing some details15:29
Nick change: Guest10683 -> kalenz15:29
exarkunThe extra work comes with extra payoff, though.15:29
arigatobut we're unlikely to include a custom enhanced version of etree in pypy, though15:29
fijalnice15:29
fijala cpython test15:29
jterraceis it possible to append an object-space object to an object-space list?15:30
cheateri'm not going to be able to rationalize that extra work for the people paying for it.15:30
fijalhttp://paste.pocoo.org/show/518851/15:30
fijala question - what *actually* happens when you fail to import resource?15:30
exarkuncheater: Oh well.15:31
cheaterexarkun: to be honest, i like etree much more than lxml15:31
cheaterand even if i don't succeed in fixing its xpath up to where i can use it, i can ducttape over it15:31
exarkunIt's always very nice to walk around on duct tape.15:32
jterracewhat's wrong with etree xpath?15:32
cheaterthere are only a few use cases here that need special attention.15:32
cheaterexarkun: yes.15:32
exarkunjterrace: It's minimally functional.15:32
cheaterjterrace: it also has a bug15:32
fijalarigato, exarkun: any idea what the code above does if you fail to import resource?15:32
cheaterjterrace: the extra functionality i need is doing x/y/@property15:32
fijal:-)15:32
exarkunfijal: if `import resource´ raises ImportError, then I would hope that execution proceeds to the except suite, which does nothing.15:32
fijalcheater: obviously you're free to do anything you want in your own program in case it was not clear15:32
exarkunfijal: But that's too obvious, so I suppose it does something else.15:33
fijalexarkun: no :)15:33
gutworth (~benjamin@dhcp-128-36-84-91.central.yale.edu) joined #pypy.15:33
fijalexarkun: it fails with NameError on resource.error15:33
cheaterjterrace: the bug is that if you have a document like <bar><baz/></bar> then to access baz you do ./baz or /baz. libxml2 does /bar/baz which is the correct thing to do.15:33
cheaterfijal: :)15:33
exarkunfijal: Aha, of course.  Nice.15:33
arigato:-)15:33
fijalthat's cpython test suite15:33
fijalgutworth: feel like fixing?15:33
exarkunThe correct spelling is `except (ImportError, ValueError): pass\nexcept resource.error: pass\n´ of course.15:33
fijalexarkun: right15:33
cheaterbut again.. i am wonhdering, why does the pypy version of etree have a different version, and different source? i can't find a place to download that version for cpython15:34
fijalexarkun: or two try: except: blocks15:34
cheaterwondering :)15:34
fijalbecause if import resource raise something else15:34
fijalyou don't want NameError15:34
arigatocheater: it does not15:34
arigatocheater: pypy has the pure Python version from CPython15:34
arigatocheater: and not the C version from CPython15:34
arigatofrom xml.etree import ElementTree15:35
Action: fijal ponders if he should file a bug on bugs.python.org15:35
kenaan12hakanardo jit-targets 113c35f57a7f67 15/pypy/jit/metainterp/: cleaner to use jump_args15:35
kenaan12hakanardo jit-targets 116c6cf4df068d 15/pypy/jit/backend/x86/test/test_recompilation.py: hg merge15:35
exarkunfijal: That's in a unit test?  I'd also say that the test should be split up into several, so that there is one test method for each expected code path, and then the methods that require resource to be available be skipped on platforms where it is not.15:37
fijal (~fijal@197.175.204.67) left irc: Read error: Connection reset by peer15:37
cheaterhmmmm.15:38
stakkars_ (~tismer@p5DC4680D.dip0.t-ipconnect.de) joined #pypy.15:45
jterraceshould space.call_method(somelist_w, 'append', item_w) work?15:46
stakkars_hi all! I got windows problems15:47
Alex_Gaynorjterrace: no, it works on wrapped lists15:47
Alex_Gaynorsomelist_w.append works for unwrapped lists of (anything)15:47
jterracethat's what i want to do15:48
jterracei want to append an object-space object to an object-space list15:48
jterraceis that not kosher?15:48
cheaterarigato: sorry. the etree version in the pypy downloadable executable is 1.3 but if i check it from cpython it's 1.2.6. but if i check under python2.7 (my system's default is 2.6.5) then it's 1.3 too. you're right, thanks.15:48
stakkars_in running tests, I pretty got "access denied" errors when certain files are closed, but something goes wrong.15:48
Alex_Gaynorno, that's fine, if you have a wrapped list you use call_method15:48
stakkars_is that effekt known?15:48
jterraceok, so that above is legal rpython?15:48
Alex_Gaynorif you have a wrapped list it should be named w_somelist though :)15:48
Alex_Gaynorjterrace: yes15:48
jterraceok15:48
jterracenice catch on forgetting the asserts in those scalar tests15:49
jterracesuch a noob mistake15:49
stakkars_is there a known issue with tempfiles maybe? temp directories are possibly15:49
stakkars_suffering a racing condition in creating a new folder.15:50
Alex_GaynorThe wisdom of fresh eyes15:50
stakkars_I also have the effect that after building pypy, when trying a build right after that, I get access denied errors.15:50
stakkars_trying the second time then works. :-(15:51
stakkars_so there is a lot still crappy about windoze&.15:51
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer15:51
cheaterso how do i get the "fastest" pypy on my server?15:51
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.15:51
jterracethere is an ubuntu PPA15:51
cheaterdo i use the binary distribution, do i get the source and build it, do i translate it to C?15:51
cheateri'll look at the PPA thanks.15:52
cheaterif i wanted to do it manually though, what is the build process?15:52
jterracehttp://pypy.org/download.html#building-from-source15:52
stakkars_cheater: build it if you have over 2 GB memory15:52
cheateri do15:53
stakkars_cheater: go to pypy/translator/goal15:53
stakkars_and run:15:53
stakkars_python translate.py -Ojit targetpypystandalone15:54
cheaterwhat is targetpypystandalone?15:54
kenaan12arigo jit-targets 11195866ee2672 15/pypy/jit/backend/llgraph/llimpl.py: Fixes for metainterp/test/test_ztranslation.  Unsure if it's really worth the burden to maintain this mess just...15:54
stakkars_that is the target that you want to build. The standalone pypy-c15:54
arigatoit's all explained at http://pypy.org/download.html#building-from-source15:54
jterrace^15:54
Action: arigato just copy-and-pasted jterrace's url :-)15:55
stakkars_well, but sometimes just giving a hand motivates pretty much15:55
arigatohakanardo: fixed15:55
arigatostakkars_: sure, sorry15:56
stakkars_but right, I should always include the pointer to pypy.org as well15:56
overminder (~overminde@143.89.167.3) left irc: Ping timeout: 252 seconds15:57
stakkars_btw., I think to add some shortcut for pypystandalone, because you have to type this crap,15:58
stakkars_just in order to divide general options and backend options ;-)15:58
arigatotab-completion to the rescue15:58
arigato(you can also type "targetpypystandalone.py", it works equally)15:59
stakkars_yes, but on windows, tab-completion kills the rest of the line, where my options are sitting15:59
arigatoah well, then you need a better terminal15:59
bbot2Started: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/624 [12hakanardo, jit-targets]15:59
stakkars_(arigato: yes, that was added by me, because I was so annoyed of windows)16:00
arigatothere are tons of them for windows, I am told16:00
jterraceAlex_Gaynor: how does micronumpy/test/test_compile.py pass when i'm using invalid rpython? shouldnt they fail?16:00
stakkars_arigato: very nice hint, thank you! will look into it.16:00
Alex_Gaynorjterrace: That doesn't test RPython compilation, test_ztranslation does though16:01
jterracethat passes too16:01
arigatoah, then I'm interested :-)16:01
Alex_Gaynorpaste your diff?16:01
jterraceit's the same as the pull request right now16:01
jterrace215 passed, 5 skipped16:01
stakkars_btw, the make option is linux specific. The windows build just fails because of "-j".16:03
stakkars_again something to fix ;-)16:03
rokujyouhitoma (~rokujyouh@12.51.221.130) joined #pypy.16:04
arigatojterrace: ah bah16:04
arigatosorry16:04
Alex_Gaynorhow?!?16:04
jterracei dont know!16:04
jterracehehe16:04
arigatohum no, "unwrap()" is marked NOT_RPYTHON16:04
arigatouh no16:05
Alex_Gaynorbugs :)16:05
arigatotest_ztranslation.py fails for me16:05
jterrace./pytest.py pypy/module/micronumpy/test/*.py16:05
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer16:05
jterraceis that the right way to run it?16:05
Alex_Gaynorjterrace: You may need to merge in default16:05
jterracei did yesterday16:06
stakkars_jterrace: ou can omit the /*.py16:06
arigatoI'm running it at c904560ff1b316:06
arigatoand test_ztranslation explodes as expected16:06
jterraceme too16:06
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.16:06
jterraceit passes for me16:06
jterracehuh16:06
Alex_Gaynor ./pytest.py pypy/module/micronumpy is enough, but yeah16:06
rokujyouhitomahi, guys. my colleague written the article.16:07
rokujyouhitoma"The fastest Python Implementation x5 faster than PyPy"16:07
rokujyouhitomahttp://blog.shibu.jp/article/51727515.html?132351573816:07
stakkars_hae? faster than pypy???16:08
jterraceack i didnt run hg merge16:08
Action: jterrace curses in anger at learning mercurial16:08
arigatorokujyouhitoma: if he means "pypy is 5x faster than cpython", then isn't the title of this blog post bogus?16:09
arigatojterrace: :-/16:09
stakkars_jterrace: great, because you then can greatly diff against the working revision and fix bugs16:09
mfoord (~anonymous@python/psf/voidspace) left irc: Quit: mfoord16:10
arigatorokujyouhitoma: ah, sorry, I see what is meant.16:11
rokujyouhitomaarigato: "The fastest Python Implementation x5 faster than PyPy"16:11
arigatoit's about RPython16:11
arigatoyes16:11
rokujyouhitomayes :)16:11
Alex_GaynorRPython isn't an implemntation of Python :)16:13
stakkars_uff, misleading article, and basically arguing with RPython, which belongs to PyPy...16:14
cheaterheh.16:14
cheaterdoes twisted work under rpython?16:14
stakkars_I would say "very unlikely". Too much Python16:15
Alex_GaynorNot a chance16:16
Alex_Gaynorit runs fine under pypy though16:16
cheaterthat i know :)16:16
arigatonothing just happens to work in RPython16:16
cheater:)16:16
stakkars_look at simple things, try for instance to compile decimal.py under RPython.16:17
stakkars_then you learn very quickly what works and what not, and how to split a module16:18
myfhi pypyers, two days ago i asked a question about the slow speed beautifulsoup on pypy compared to python. does anyone has an idea now?16:18
myfhave16:18
cheatermaybe old style classes and new style classes in mixed inheritance?16:19
hakanardoarigato: thanx!16:19
arigatohakanardo: :-)16:20
arigatocheater: that's a random comment without even looking at the source code16:20
cheaterarigato: yes!16:21
cheaterarigato: it is.16:21
Rotund (~joe@184-158-101-126.dyn.centurytel.net) joined #pypy.16:22
arigatomyf: yes:16:26
arigatohttp://paste.pocoo.org/show/518872/16:26
arigatothis script shows that BeautifulSoup just requires tons of warm-up time16:26
arigatobut it ends up much faster than CPython16:26
arigatoafter 3-4 seconds16:26
arigatoreplace the range(50) with range(500) for more impressive results :-)16:27
myfthat's the warming up bit i suppose?16:28
myfexciting, let me test it out 16:28
arigatoI'm getting 12.4 seconds on pypy, and 52 seconds on python2.516:28
overminder (~overminde@wf097-252.ust.hk) joined #pypy.16:28
arigatoso it's not too far from our 5x-speed-up informal goal16:28
arigato(...that I just decided on the spot for the present purposes)16:29
myfsuppose you process a batch of files on beautifulsoup, pypy is going to be much faster. it's just taking a long time for the first one. is it?16:31
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer16:31
arigatoyes16:32
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.16:32
arigatoor even for the first 10 ones16:32
stakkars_arigato: always a good way to boost speed: compare against python 2.5 :-)16:32
arigatowe can see in this script, which prints one "." per processed file, that it starts slowly and gets faster and faster16:32
myfright right16:33
arigatostakkars_: ok ok :-)16:33
myfohh i see the result now. fascinating16:33
Rotundmyf, do you understand the basic idea of how pypy gets the speed ups over CPython16:33
arigatoalso, if the processed pages are all different instead of equal, it might increase the warm-up time even more16:33
myfon my machine pypy does it 3.4 sec and cpython 7.2416:34
myfRotund, would you enlighten me?16:34
arigatoyes, roughly the same here16:34
arigatofor 5016:34
myfright16:34
RotundIt is basically profiling your code as you run it16:35
myfthe jit bit?16:35
RotundIt then optimizes based on the profiling16:35
RotundSo, if you see that variable X is always an int, you can speed that up.16:35
nedbat (~nedbat@python/psf/nedbat) joined #pypy.16:35
arigatoit's not as fancy as this implies16:35
myfsomeone has said that beautifulsoup has tons of regex in it so during the warm up time it is taking sweet time to compile16:36
arigatoit "profiles" in the sense of "oh, this loop has been looping 1000 times, let's compile it"16:36
myfhmm16:36
arigatoand then to compile, it just traces one iteration16:36
RotundYou need to make sure that X is an int, but if it is, you get a big boost as you don't have to JIT and then keep running it and you may notice other speedups16:36
Rotundarigato, my understanding was that it would also use variable types as hints16:37
arigatonot directly: it's a bit more indirect, but it gets a similar result16:38
arigatobasically it is a tracing JIT, so it compiles only the path that was followed during the traced iteration16:38
RotundOkay. I mainly listened into the pypy training before the sprints at PyCon16:38
arigatoif this path contain 5 type checks that fail and a 6th one that succeed on a variable X,16:38
arigatothen it's optimized into a single type check (the one that succeeds) in the machine code16:39
arigatoand moreover if the variable X doesn't change during the loop, the type check is then moved outside the loop16:39
nedbatcan anyone take a look at test_sandlib.py or another sandbox test file and help me understand how to test an issue?16:40
arigatoyes?16:40
RotundOne way or another, you need to "prime the pump" to get the big speedups out of pypy.16:40
nedbatCurrently the sandbox fails to "import site".  I wanted to add a test that also fails to import site, but I couldn't make it fail with a small case.16:40
arigatoRotund: yes, and this warm-up time is quite long so far (it could be optimized too)16:41
nedbatarigato: I tried this http://paste.pocoo.org/show/518885/  at the bottom of test_sandlib.py, but it doesn't fail.16:41
stakkars_ (~tismer@p5DC4680D.dip0.t-ipconnect.de) left irc: Quit: schnarch16:42
arigatonedbat: I think that generally speaking, side.py should not be executed in a sandboxed process16:42
arigatonedbat: because it can do too much random things16:42
arigatosite.py sorry16:42
nedbatarigato: that may be.  Currently, it is imported by default in the sandboxed proc, and fails, due to calling os.getuid and os.geteuid.16:42
arigatonedbat: your example doesn't make sense I think16:42
nedbatarigato: can you explain how?  I'm sure i'm confusing layers.16:43
arigatocompile() compiles an RPython program16:43
arigatonedbat: just fix it by calling the sandboxed pypy with the "-S" option16:43
nedbatarigato: i'm not trying to fix my code (I've already done precisely that), I'm trying to fix pypy.16:43
arigatocompile() compiles an RPython program, which means these tests are about testing the sandboxing feature of the translation tool-chain, but not the sandboxed pypy at all16:44
nedbatarigato: today if you run the sandbox example, it printer "couldn't import site"16:44
nedbat*prints16:44
arigatothe fix is in pypy_interact.py16:44
nedbatarigato: you mean to add the -S ?16:45
arigatoe.g. line 41, add "-S" to the arguments, yes16:45
arigatoyou could also argue that a sandboxed pypy should by default not run site.py, but changing that is a bit more involved,16:46
arigatobecause the code is the same between non-sandbox and sandbox pypy's16:46
nedbatarigato: could you point me to a test that tests the "sandboxed pypy"?  I would like very much to have a test that covers this issue.16:46
arigatothere is no test :-(16:47
arigatobecause it requires a complete translation of a sandboxed pypy first16:47
arigatoso if sandboxing were a feature more actively maintained, we would have a buildbot setup that translates a sandboxed pypy and runs specific tests16:47
arigatobut it's not :-(16:48
arigatoah, there is test_pypy_interact.py16:48
arigatobut it works by building a "mini-pypy-like" executable instead16:49
mat^2 (~mathias@212.130.113.35) joined #pypy.16:49
nedbatarigato: why are these two things different?  1) importing site inside the full-blown sandboxed pypy, and 2) importing site inside one of these mini-pypy-like executables built by the unit tests?16:50
arigato1) imports "site" in a pypy interpreter.  2) is at RPython-level, where "import site" basically has no effect16:51
bbot24Failure: 15http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/624 [12hakanardo, jit-targets]16:51
arigatoin RPython, "import foo" is run at translation time and the resulting "foo" module plugged in place, as a constant16:51
nedbatarigato: right, ok.16:52
nedbatarigato: and a unit test that built a real sandboxed pypy would take an hour to run.16:52
arigatoyes16:53
arigatowe would like instead to have a set of tests that run with a previously-built sandboxed pypy16:53
arigatolike we have a set of tests that run with a previously-built pypy-with-a-jit16:54
arigatoand checks how it compiles small (pure Python) examples16:54
nedbatarigato: can you point me where that existing test is?16:54
arigatoyes: it's all files in pypy/module/pypyjit/test_pypy_c/16:55
arigatoit requires quite a bit of setup, unfortunately16:55
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer16:56
jterraceAlex_Gaynor: arigato: getting a translation error now that im not sure how to understand16:56
jterracethis is my current diff: http://pastebin.com/WFHF7HWB16:57
arigatomostly the run() method at test_00_model.py:2516:57
jterracethe error i'm getting (http://pastebin.com/raw.php?i=cFkRSnBT) doesn't make much sense to me16:57
asmeurer__ (~asmeurer@c-174-56-21-245.hsd1.nm.comcast.net) joined #pypy.16:59
arigatoI'm unsure to see why you added tolist()16:59
jterracei added it to the type itself where it could be unboxed16:59
jterraceinstead of having to check its type in genericbox16:59
jterracei thought that would be the right thing to do16:59
arigatobut why to the TypeDef? this adds it to app-level, i.e. it is user-visible16:59
jterraceyeah, it's supposed to be17:00
jterracenumpy.int32(5).tolist() == 517:00
arigatoah well, you're doing two things at the same time then17:00
arigatoyou're trying to fix translation errors, but at the same time writing new functionality17:00
kkris (~kris@80-123-38-195.adsl.highway.telekom.at) left irc: Read error: Connection reset by peer17:00
jterracethe rest of the tests pass though17:00
jterraceincluding testing the scalar tolist17:01
jterraceexcept for the translation17:01
arigatoI'm sure.  just saying that it would be great to focus on the translation issue instead17:01
jterracethe previous version had scalar tolist as well17:01
jterracethat's how the recursive tolist on arrays works17:01
jterracesince it calls the base version17:01
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.17:02
arigatofeel free to check in, if it passes the same amount of tests as before, plus the new ones17:02
jterraceit does, but the translate test still doesnt pass17:02
arigatoI know17:02
jterracelet me see if it's the translation of the scalar that's breaking it or the translation of the numarray tolist17:03
arigatoI can help you, but please first commit this changeset17:03
jterraceok17:03
jterraceone minute17:03
jterracearigato: i just pushed to jterrace/pypy with a branch name numpy_tolist17:10
jterrace(sorry realized after that convention is numpy-tolist)17:10
arigatook17:10
Alex_Gaynorjterrace: we'll live, I'll look after I get off my train :)17:11
tumbleweed (~stefanor@ubuntu/member/tumbleweed) joined #pypy.17:11
arigatofor me, running "py.test test_ztranslation.py", the first translation failure is:17:11
arigatohttp://paste.pocoo.org/show/518897/17:12
jterraceyeah17:12
jterracethat's what i pasted17:12
arigatosorry :-)17:12
jterracethat's ok17:12
arigatohum, a SomeObject17:12
jterraceso im confused how to parse what this is saying?17:12
arigatolet me fix that in checkcompile(), it should have crashed earlier17:13
jterraceok17:13
Alex_Gaynorit's because your space.wrap() is on either a bool, float, or int17:13
arigatouh? no17:13
Alex_Gaynoryou need to move the space.wrap into the method thayt knows that type in itemtype17:13
Alex_Gaynorarigato: yes17:13
Alex_Gaynorat least, that's what the error is17:13
arigatowell, not the one I pasted17:13
Alex_Gaynorwell, I meant the issue with the code, I didn't look at the exact error17:14
jterraceoh because i'm calling self.unbox?17:14
arigatoif you say so, but I'm still confused, because you can call space.wrap() with various arguments17:14
arigatothe problem is not in space.wrap() itself17:14
arigatohttp://paste.pocoo.org/show/518900/17:16
jterracewhat's that do?17:17
arigatothat's a fix, gives better error messages, hopefully17:17
Alex_Gaynorarigato: you can't space.wrap(<float> or <int> or <bool>) since the types get merged17:17
Alex_Gaynoryou end up with a SomeObject17:17
arigatoAlex_Gaynor: my point is that you can't *have* an object that is <float> or <int> or <bool>17:18
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer17:18
Alex_Gaynorarigato: right17:18
arigatothat's a problem unrelated to calling space.wrap() on it17:18
Alex_Gaynortrue, I was just trying to be clear about where I was describing17:18
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.17:18
arigatosorry17:18
kenaan12arigo default 11969865e9cb30 15/pypy/objspace/fake/objspace.py: Forbid SomeObject there.17:19
kkris (~kris@93-82-47-109.adsl.highway.telekom.at) joined #pypy.17:19
gutworthshouldn't be allowing SomeObjects be False by default?17:20
arigatogutworth: yes, probably, nowadays17:20
arigatoit is, for stand-alone builds17:20
arigatojterrace: http://paste.pocoo.org/show/518904/17:20
overminder (~overminde@wf097-252.ust.hk) left irc: Ping timeout: 244 seconds17:21
arigatothis says that the various methods eval() are called at the place with the ==> marker17:21
jterraceso i cant call space.wrap(i) ?17:21
arigatobut they have incompatible signatures17:21
jterracespace.int(i) ?17:21
arigatothe problem is not space.wrap, it's the call to eval()17:21
jterraceoh17:22
arigatothis place can call (at least) the three methods listed in the beginning17:22
arigatoeval() of W_NDimArray, or of ViewArray, or of Scalar17:22
Alex_GaynorI think eval is fine, it's the descr_tolist on W_GenericBox that's fine17:22
arigatoindeed, this place is a line in descr_tolist, at interp_numarray.py line 92117:23
Alex_GaynorI'm talking about interp_boxes.py17:23
arigatoI think there are several errors, and we're talking about different ones17:25
jterracehave to run for a bit, thanks for that change17:26
jterraceill try and fix it later17:26
arigatothe issue is that eval() takes an instance as argument, e.g. a ConstantIterator, but not 017:29
jterrace (~Adium@dynamic-oit-vapornet-b-25.Princeton.EDU) left irc: Quit: Leaving.17:32
Rotund (~joe@184-158-101-126.dyn.centurytel.net) left irc: Ping timeout: 252 seconds17:34
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.17:35
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer17:35
fijal (~fijal@197.172.88.160) joined #pypy.17:37
fijal (~fijal@197.172.88.160) left irc: Read error: Connection reset by peer17:42
Action: arigato off17:43
arigato (~arigo@fwstups.cs.uni-duesseldorf.de) left irc: Quit: See you17:44
cheateradded x/y/@z support to etree17:50
cheaterit was very easy17:50
cheaterunder 20 loc :)17:50
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer17:50
DanKluev (~root@78-24-225-196-dsl.vntc.ru) left irc: Quit: Leaving.17:55
DanKluev (~root@78-24-225-196-dsl.vntc.ru) joined #pypy.17:57
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.17:58
jterrace (~Adium@dynamic-oit-vapornet-b-25.Princeton.EDU) joined #pypy.18:02
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Ping timeout: 240 seconds18:03
jterrace (~Adium@dynamic-oit-vapornet-b-25.Princeton.EDU) left irc: Quit: Leaving.18:11
arigato (~arigo@82.113.121.130) joined #pypy.18:22
antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds18:23
DasIch_ (~dasich@p4FFDFC5D.dip.t-dialin.net) joined #pypy.18:37
DasIch (~dasich@p4FFDCD7B.dip.t-dialin.net) left irc: Ping timeout: 252 seconds18:40
mattip (~chatzilla@bzq-79-180-110-4.red.bezeqint.net) joined #pypy.18:42
cheater (~cheater@ip-80-226-24-11.vodafone-net.de) left irc: Ping timeout: 252 seconds18:44
mattipUhh, I think I stuffed up one of the build bots (bigboard?) with a win32 build18:45
mattipI started the build over a da ago, but it is still in the app-level (-A) test stage, and sending it an interrupt didn't kill the build.18:47
mattipcan anyone help?18:47
mattips/a da ago/a day ago/18:47
Alex_GaynorI'm not sure who, besides arigato, has admin on that box18:48
mattiparigato: ping18:48
arigatopong18:48
mattipHi. can you see what I said a minute ago or should I repost?18:48
arigatoI don't see any builder pending18:48
arigatoah yes18:49
arigatosorry18:49
arigatoI'm not an admin on that box, but I can at least log in as buildslave18:49
Alex_Gaynorwhoops18:50
mattiparigato: any idea what went wrong?18:51
arigatono18:51
arigatoah yes18:51
exarkunin general, interrupt doesn't work for windows slaves.18:51
arigatoit's blocked showing a dialog box, of course18:51
bbot213Exception: 15http://buildbot.pypy.org/builders/pypy-c-jit-win-x86-32/builds/288 [12mattip, default]18:51
Action: arigato clicked Ok18:51
Alex_Gaynorof coruse18:51
arigato:-(18:52
arigatook, you can restart your build if needed18:52
arigatoit's likely to get the same dialog box18:52
arigatobut if you don't interrupt but just ask me to click ok18:52
arigatoit should go past it18:52
mattipum, I guess I should find out why first. What dialog?18:52
arigato(as a temporary measure)18:52
arigatoit's the same old dialog, not something new, actually18:53
arigato"libexpat.dll not found"18:53
arigatoI never understood who was trying to load it, because libexpat.dll is properly copied in the same directory as pypy-c.exe18:53
mattipThat opens a dialog box? not nice.18:54
arigatowell it's windows18:54
Alex_Gaynorwindows has a tendency to do such things18:54
mattipIs it the runtime <arbitrary number> error or another one?18:54
arigatoI don't think it displays an arbitrary number18:54
arigatoto fix it, we'd need motivation and stakkars' instructions for how to set up the "auto-clicker" tool18:55
Alex_Gaynorit's some registery key I thikn18:55
arigatohah, naive18:55
mattipI have some experience with canceling windows dialog box using registry keys, but I would need a screen capture of the dialog box.18:55
arigatothe problem is that it's not just one registry key18:56
arigatoit's a mess of things that can each end up opening a dialog box, and each of them has some way or other to disable it18:56
arigatowe already set one or two registry keys18:56
arigatomattip: I'll take a screenshot next time18:57
mattipThanks. If I'm not around, send a link to matti <period> picus <at> gmail <period> com18:58
arigato(that's basically why I prefer an auto-clicker tool, which should just work in all cases)18:58
mattipOk, motivation I've got, I will try to talk to stakkers when he's around.18:58
arigato:-)18:59
cheater (~cheater@ip-80-226-24-12.vodafone-net.de) joined #pypy.18:59
RiaanFV (~riaanfv@41-132-169-26.dsl.mweb.co.za) joined #pypy.18:59
mattipI have another naive question: how much network traffic does a buildbot slave use for default builds of pypy?19:06
mattipDoes the slave need to clone the repository for each build?19:06
khs (~khs@2001:700:300:2120:725a:b6ff:fee5:a44) left irc: Quit: Leaving19:07
mattipstakkars: ping19:08
Action: arigato finds "clickoff"19:09
arigatohttp://www.johanneshuebner.com/en/clickoff.shtml19:09
arigatomattip: no, the slave can keep the already-cloned repo between builds19:09
arigatoand if you care, you can even manually initialize the few repos needed on a typical slave19:10
arigatoto avoid a few complete downloads19:10
kenaan12ned nedbat-sandbox 117fa920a29974 15/pypy/translator/sandbox/: Implement from os: getuid, geteuid, getgid, getegid; in an attempt to get 'import site' working, but the probl...19:11
jterrace (~Adium@dynamic-oit-vapornet-b-25.Princeton.EDU) joined #pypy.19:14
mattiparigato: thanks19:15
mattiparigato: what would it take to get clickoff installed on bigboard?19:18
arigatovery naively I thought it wouldn't require an admin and installation19:20
arigatobut of course I was wrong19:20
Action: arigato looks if there is a Python program19:21
stakkarsno time right now19:33
stakkarsfinancial work, give me two hours19:34
derdon (~derdon@p5DE89AF6.dip.t-dialin.net) joined #pypy.19:34
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 240 seconds19:34
CIA-1303nedbat 07roundup * 10#949/1.7 Sandbox problems: 19:37
CIA-13About the "import site" problem: it seems to involve more than a few changes. The19:37
CIA-13best thing to do is to use "-S" in the sandbox arguments, to ... * 14https://bugs.pypy.org/issue94919:37
arigatook, found http://mail.python.org/pipermail/python-win32/2002-December/000635.html and tried it, it works nicely19:39
arigatonext time we get the dialog box, we can add its description to the program, and then we won't be annoyed any more by it19:39
arigatoI know it's not the perfect solution19:40
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.19:40
arigatobut with minimal extra hacking it can be ok19:42
arigato(a unix-like solution for a windows-only annoyance)19:42
mattiparigato: thanks. Sometime after the nightly build starts I will try to take a look and give you a heads up.19:44
InsectoidThis issue with random dll dialog boxes is a deeper problem, I think. At least, I'm seeing similar with my WX build. From what I was able to determine last night, it has no problem importing .dll (or .pyd files) from the same directory, but blows up when they're elsewhere.19:51
jterracethink i fixed tolist19:51
InsectoidFor instance, with WX, which has several pyd files. If I import, for instance, _core_ from within the wx directory it works. If I import wx, it blows up when importing _core_.pyd with a missing wx DLL.19:51
Alex_Gaynorjterrace: is it in the pull request?19:51
jterraceno, going to update it now19:51
Alex_Gaynork19:51
antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) joined #pypy.19:55
fijal (~fijal@41.49.224.41) joined #pypy.19:55
fijalhey19:58
fijalwtf is ITIN19:59
jterraceAlex_Gaynor: ok, pull request is updated19:59
jterraceall tests pass now20:00
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.20:00
mattipInsectoid: interesting. You have a working build of WX? Cool.20:00
mattipIs this for win32?20:00
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) left irc: Client Quit20:01
mattipFrom what I know, dlopen should be calling LoadLibrary with the name of the file to open.20:02
mattipIf the name has no slashes in it, the documentation of LoadLibrary claims that the dll (or pyd) can be anywhere on the system path.20:03
mattipMaybe pypy is adding a "." or a "./" to the path?20:03
mattipfijal: Individual Taxpayer Identification Number (ITIN)?20:04
fijalyes20:04
fijalmattip: I'm trying to understand if PSF is wanting me to do something illegal or not20:05
fijalvoidspace: ping20:05
mattipAnyone can get one, you do not have to be a citizen. 20:05
fijalright20:05
Action: fijal is super puzzled20:06
fijalthe main info about ITIN seems to be that they won't report that to homeland security20:06
fijal"In addition to use by unauthorized immigrants, ITINs are used by foreign investors in United States real estate. Such investors need ITINs to file federal and state tax returns to report rental income."20:07
Alex_Gaynorjterrace: ok on the wrong OS, I'll play with this and merge once I'm back under linux20:07
fijalI don't really feel like fitting in either of those categories20:07
jterracethanks20:07
mattipUS citizen -> ITIN == Social Security Number. Non-US citizen-> ITIN == random number issued to you for tax-collecting purposes.20:08
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy.20:08
nedbat (~nedbat@python/psf/nedbat) left irc: Ping timeout: 248 seconds20:09
mattipAll income generated in the US needs to be taxed there. You will need to file a US tax report.20:09
fijalmattip: google soc has not been taxed20:10
fijalin the states20:10
mattipThe wierder thing is that US citizens must file tax reports even if they live 365 days a year outside the US.20:10
fijalmattip: that's normal, polish have to do the same20:10
mattipYes but even if they have no US based income.20:11
fijalin theory I should even pay taxes in poland sometimes20:11
Alex_Gaynorfijal: really?  I thought US was one of few countries that taxed based on citizenship and not living location20:11
fijalAlex_Gaynor: one of the few :)20:12
fijalso I guess poland is another20:12
Action: lucian wonders if he should be paying taxes in romania20:12
mattipMaybe SOC has some tax exempt status as it is somehow connected to students, but IANAL20:12
lucianit sounds stupid, double-dipping like that20:12
mattipwhoops IANAA (accountant)20:12
jerithPerhaps they just want a return stating that you have no taxable incamoe?20:12
jerith*income20:12
lucianmattip: GSoC participants pay taxes in the country they work in20:13
Fluxid< mattip> The wierder thing is that US citizens must file tax reports even if they live 365 days a year outside the US.20:13
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 255 seconds20:13
Fluxidwhat, really?20:13
mattipBut they are employees of Google USA?20:13
Fluxiddamn, i have double citizenship..20:13
lucianmattip: technically, they're contractors of Google, yes20:13
mattiplucian: but Google has offices locally in many countries.20:14
luciani know non-US students pay no US taxes (i didn't)20:14
lucianmattip: regardless, everyone deals with Google US for GSoC20:14
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.20:15
mattipOK, I dunno. I only know that it was a real pain to find an accountant here (Israel) that understands dual citizenship and taxes.20:15
Alex_Gaynorhuh, I'm a little surprised, that's a pretty sizable group of people, though I know at least one person who uses an american accountant because that's easier20:16
RiaanFV (~riaanfv@41-132-169-26.dsl.mweb.co.za) left irc: Quit: Leaving20:16
mattipWell, alot of people (like Fluxid) don't really know/care about the obligations. I happen to be a stickler for filing correctly. The IRS only takes notice if you are a big fish.20:18
Action: fijal is not particularly big20:18
Fluxidwell i live outside us since i'm 3, and work/pay taxes in poland, i thought i don't need to do anythin. mattip how can i find some info about this?20:19
mattipI even make a FBAR declaration http://www.irs.gov/businesses/small/article/0,,id=210244,00.html20:21
Action: mattip looking at irs website for Fluxid20:21
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 252 seconds20:23
mattipFluxid: start here http://www.irs.gov/businesses/small/international/article/0,,id=97324,00.html20:23
fijalAlex_Gaynor: checkins like 6c4b9c3bc0d6 are utter nonsense20:23
fijalI want my way to check loops back20:23
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.20:24
Fluxidmattip: thanks20:24
Alex_Gaynorfijal: well, not totally useless20:24
fijalno?20:24
fijalI mean we'll break such things all the time20:24
Alex_Gaynorthe test still makes the assertion we want20:24
fijaldid you look at traces?20:24
fijaldoes it?20:25
Alex_Gaynorit also makes a bunch of other unimportant ones20:25
fijalwhat does it mean we have 2 extra guards class?20:25
Alex_Gaynoryes, it has float_add: 4, it would be float_add: 8 if it was wrong20:25
Alex_Gaynoror something like that20:25
mattip (chatzilla@bzq-79-180-110-4.red.bezeqint.net) left #pypy.20:25
Alex_Gaynorfijal: that's the unimportant extra assertion20:25
Action: fijal heavily doubts that20:25
fijalI just skip them20:25
fijalbecause it's nonsense to keep them around20:25
Alex_Gaynorwell, then we don't know if it's broken20:25
fijalwe don't know either way20:26
Alex_Gaynoryes you do, because you can tell by reading traces20:26
fijalif some operation is no longer moved to the preamble only, the number stays the same, no?20:26
fijalthere are many ways you can organize the same number of resops20:26
Alex_Gaynoryes, but that's not what this test is about20:26
fijaland not all of them are equally good20:26
Alex_GaynorYes, we have other tests for that stuff though20:27
fijalit used to be about that, it's not about that any more20:27
Alex_Gaynorno it's not20:27
Alex_Gaynorthis one is specifically about what do I do if part of the tree was precomputed20:27
fijaleh20:27
fijalthen make it float_add=420:28
fijalif this is what you're checking20:28
linq (~ident@24-246-25-39.cable.teksavvy.com) joined #pypy.20:32
Nick change: Gulaway -> Gulopine20:52
Da_Blitz (~Da_Blitz@203.56.250.63) left irc: Ping timeout: 268 seconds20:57
arigatoAlex_Gaynor, fijal: hakanardo said the jit-targets branch is getting ready to be merged21:01
Alex_GaynorCool21:01
arigatojust wanted to tell you :-)21:01
arigatothe micronumpy test with obscure "getfield_gc=41" can then be made more precise21:02
arigatoI mean, we need to invent something else, but at least it's better than now because it's about to break due to the jit-targets merge21:02
fijalcool21:03
etrepum (~bob@69-9-57-57.static.logl001.digis.net) left irc: Quit: etrepum21:10
Rotund (~joe@184-158-101-126.dyn.centurytel.net) joined #pypy.21:12
Da_Blitz (~Da_Blitz@203.56.250.63) joined #pypy.21:13
fijalarigato: just merge it then maybe?21:20
fijalarigato: I'm fine with skipping the obscure merge for now21:20
fijaler21:20
fijalthe obscure failure21:20
fijalI'm refactoring stuff so it would not be so reliant on preamble anyway21:20
arigato:-)21:24
arigatoI'll leave hakanardo decide when it's ready for merge21:25
arigatobut today we fixed probably the last failures21:25
arigatoit'll make bivab unhappy though, because it's yet another change for the arm backend :-/21:26
mtigas (~mtigas@c-67-185-214-42.hsd1.wa.comcast.net) joined #pypy.21:27
mtigas (~mtigas@c-67-185-214-42.hsd1.wa.comcast.net) left irc: Client Quit21:28
tumbleweedwhy does pypy ship two unicodedata versions?21:28
mtigas (~mtigas@c-67-185-214-42.hsd1.wa.comcast.net) joined #pypy.21:29
etrepum (~bob@c-76-27-62-239.hsd1.ut.comcast.net) joined #pypy.21:31
fijalarigato: and ppc one21:40
fijalarigato: what we miss for having arm in trunk though?21:40
fijalother than a buildbot21:40
ramusara (~ramusara@220.156.210.236.user.e-catv.ne.jp) left irc: Quit: Leaving...21:40
fijal (~fijal@41.49.224.41) left irc: Read error: Connection reset by peer21:41
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.21:47
mtigas (~mtigas@c-67-185-214-42.hsd1.wa.comcast.net) left irc: Quit: mtigas21:48
arigatofijal (for the logs): we fixed friday what was probably the last bug in arm, 21:57
arigatoso now it should land in trunk very soon21:57
Alex_GaynorDo we have a buildbot?21:57
arigato(of course, the last bug before the next one)21:57
arigatono, that's annoying too21:58
Alex_Gaynor:/ if we have no buildbot it'll go out of date21:58
arigatoyes21:58
arigatomaybe it's possible to set up something completely based on a virtual machine21:58
Alex_Gaynorthat'd be ok, we couldn't test on speed, but we could do everything else21:58
Alex_Gaynorwhich is good enough IMO21:59
arigatobivab was using a combination of a VM and of a real beegle board21:59
arigatoyes, hopefully21:59
arigatobut of course just installing the VM is a major issue and depends on the platform21:59
Rhyolitearigato: there probably is more opportunity for PyPy with 64-bit ARMv8, but ARMv8 ISA is a complete rewrite21:59
arigato64-bit ARMv8 are still future hardware only, or are they available now?22:00
Rhyolitenot available now22:01
Rhyolitebut some vendors say they have designs22:01
Rhyolitethat will be much more powerful for self-hosted systems22:01
Action: lucian has an arm laptop he might be able to set up as a buildslave22:02
arigatowell, anyway, what we would really like is someone that knows how to install and maintain a VM on Linux, for ARMv7 for now22:02
luciani'd rather sell it though, frankly22:02
arigatoa VM on a Linux server machine is the best way to go22:03
Rhyolitearigato: yes.  It will be easier to set up build bots for PPC22:03
arigatoI'm rather unhappy even about the Macs: although we have a buildbot for OS/X 64, it's not a server, so we can't completely rely on it and log in at any time to try stuff22:04
arigatoRhyolite: yes22:04
Alex_Gaynorarigato: we should talk to whoever does the CPython buildbots about sharing them22:04
arigatomy opinion is still that a VM is best, and once we have a VM, installing it on this or that machine doesn't really change much I hope22:06
arigatoin other words the issue is not about sharing machines with the CPython buildbots, but just having someone that cares about the VM22:06
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) joined #pypy.22:07
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy22:10
Ademan (~dan@adsl-71-141-224-79.dsl.snfc21.pacbell.net) left irc: Quit: Lost terminal22:11
jterraceAlex_Gaynor: added another pull request for uints and fromstring: https://bitbucket.org/pypy/pypy/pull-request/16/22:11
jterracepromise all the tests pass this time22:11
jterracethe formatting of my list got messed up, seems like i cant update the description?22:11
lucianarigato: an emulated vm you mean?22:12
arigatolucian: yes22:12
lucianwon't that be too slow to be practical?22:13
lucianthe android emulator is stupid slow, for example22:13
lucianand that's just qemu22:13
gsneddersarigato: 64-bit ARM hardware is a fair way off shipping. Cortex A15 is still to start shipping, and the 64-bit stuff is the generation after.22:15
Action: tumbleweed debugs arm stuff under qemu-user chroots. It's slow, but esp. for things that parellise, it's not that bad.22:16
tumbleweedI'm happy to try a pypy build and see how long it takes :P22:16
gsneddersQEMU runs at about half the speed on a BeagleBoard on a fairly high-end Core 2 Duo, IIRC22:16
gsnedderstumbleweed: For building you're better off cross-compiling, the only thing you want to do on ARM is test the build.22:16
tumbleweedgsnedders: right, I do distro-packaging stuff, so we are always doing everything native22:17
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer22:17
gsneddersAlso for testing ARM you have the question of whether you test on hardware with NEON/VFP22:17
gsnedderstumbleweed: What's the gain in doing it on the actual hardware?22:17
tumbleweedgsnedders: we want every architecture to be self-hosting22:18
gsnedderstumbleweed: Ah, so it's just a matter of making sure it is self-hosting.22:18
amaury_Hi, I'm still trying to embed 2 object spaces in a single translation22:18
gsneddersIn the more general case, it doesn't seem worthwhile checking a compiler on ARM itself can compile the ARM build.22:19
tumbleweedpractically, that means we (debian) have the resources to do everything native (porterboxes and buildds for every arch)22:19
amaury_But it fails around multimethods22:19
luciangsnedders: for now22:19
amaury_similar to this test: http://paste.pocoo.org/show/519024/22:19
gsnedderslucian: Which part?22:19
luciangsnedders: self-hostiness. it will matter if arm becomes mainstream for desktop-ish machines22:20
amaury_can PBCs be inherited?22:20
gsnedderslucian: It doesn't for PyPy's case, surely? Doesn't it seem reasonable to assume an ARM build of gcc will work just as well as an x86_64 build of gcc, both with an ARM-generating backend?22:21
gsneddersIf there's any issue on ARM but not when cross-compiling, that's a gcc bug.22:22
amaury_hmm, MixedModules can be subclassed22:22
amaury_but their _freeze_() method returns False22:22
Rhyolitegsnedders: not completely22:22
luciangsnedders: there are always edge cases22:22
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.22:22
luciangsnedders: when PyPy devs develop on ARM hardware, it will matter. but of course, it'll solve itself. i was just nitpicking22:23
Rhyolitegsnedders: there still may be some differences with 32 bit or 64 bit host of gcc22:23
gsneddersWell, I guess there's the case of when you accidentally rely upon something on the host system instead of the arch you are compiling for.22:23
Action: gsnedders has most certainly done that22:23
Rhyoliteand possibly some OS features not detected by cross-build environment22:23
Action: gsnedders hasn't touched any cross-compilers since setting one up a couple of years back22:25
strtok (strtok@thales.strtok.net) left #pypy.22:27
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer22:27
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.22:27
luciancan PyPy be cross-compiled? doesn't it depend on the target during translation?22:28
lucianor am i being stupid22:28
amaury_no cross-compilation at the moment I'm afraid22:30
gsneddersHave fun finding an ARM machine with enough RAM to do the translation! :P22:31
tumbleweedyeah. 2G is not hard to find. More ... :/22:32
tumbleweedoh, before I forget. I was able to build pypy fine on my 64bit desktop with 6G ram, 2 months ago. But not recently.22:32
Shanita (~John@osbk-4db06a58.pool.mediaWays.net) left irc: Read error: Connection reset by peer22:34
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.22:35
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer22:35
bbot2Started: 15http://buildbot.pypy.org/builders/own-macosx-x86-32/builds/74022:36
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-64/builds/16622:36
bbot2Started: 15http://buildbot.pypy.org/builders/jit-benchmark-linux-x86-32/builds/97422:36
Kaskuka (~John@osbk-4db06a58.pool.mediaWays.net) joined #pypy.22:37
Rotund (~joe@184-158-101-126.dyn.centurytel.net) left irc: Quit: Ex-Chat22:37
arigatono, pypy-for-arm is cross-compiled right now22:40
arigatoyou need a VM and a way for the translator to inspect the VM's environment instead of the host by running gcc22:40
arigatoplus a host that is not completely unrelated to the target22:41
arigatoe.g. a 32-bit host, not a 64-bit22:41
amaury_ (~amaury_@46-127-23-192.dynamic.hispeed.ch) left irc: Ping timeout: 244 seconds22:46
fox__ (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer22:46
lucianarigato: i see22:49
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.22:52
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Read error: Connection reset by peer22:56
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) joined #pypy.23:00
fijal (~fijal@197.172.198.224) joined #pypy.23:01
Nick change: DasIch_ -> DasIch23:05
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) joined #pypy.23:06
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.23:11
lucian (~lucian@cpc1-newc15-2-0-cust84.gate.cable.virginmedia.com) left irc: Ping timeout: 244 seconds23:16
Nick change: Gulopine -> Gulaway23:17
`fox` (~fox@host48-59-dynamic.48-82-r.retail.telecomitalia.it) left irc: Ping timeout: 244 seconds23:21
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) left irc: Quit: Computer has gone to sleep.23:21
dracman (~draco@212.255.27.92) left irc: Ping timeout: 252 seconds23:21
dracman (~draco@212.255.19.163) joined #pypy.23:22
kkris (~kris@93-82-47-109.adsl.highway.telekom.at) left irc: Quit: Leaving.23:27
papercrane (~papercran@c-71-202-211-7.hsd1.ca.comcast.net) joined #pypy.23:31
tumbleweedlooks like somone forgot to bump the version number to 1.7 in the docs...23:39
fijal (~fijal@197.172.198.224) left irc: Ping timeout: 268 seconds23:39
kenaan12alex_gaynor numpy-dtype-refactor-complex 114c4c94b41e24 15/pypy/doc/conf.py: update the version number (technically we're 1.8-pre or something now, but this is at le...23:42
Alex_Gaynorcrap, wrong branch23:45
fzzzy (~donovan@76-198-130-19.lightspeed.mtvwca.sbcglobal.net) left irc: Quit: fzzzy23:45
Alex_Gaynorah crap wrong branch23:45
tumbleweedheh23:46
kenaan12alex_gaynor default 1116f29d272911 15/pypy/doc/conf.py: update the version number (technically we're 1.8-pre or something now, but this is at least *more* accureate)23:46
nedbat (~nedbat@python/psf/nedbat) joined #pypy.23:50
antocuni (~antocuni@host117-71-dynamic.58-82-r.retail.telecomitalia.it) left irc: Ping timeout: 252 seconds23:51
--- Sun Dec 11 201100:00

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