1 2011-12-16T00:29:33 *** raignarok
2 2011-12-16T01:32:25 * ThomasWaldmann uploaded some tasks
3 2011-12-16T01:51:15 <ThomasWaldmann> izibi: for the "http performance" table, you need to tell how exactly you measure this
4 2011-12-16T01:51:59 <ThomasWaldmann> izibi: btw, when having finished, you can show the stuff to the guys on #pypy
5 2011-12-16T02:27:00 <ThomasWaldmann> http://windowsteamblog.com/ie/b/ie/archive/2011/12/15/ie-to-start-automatic-upgrades-across-windows-xp-windows-vista-and-windows-7.aspx
6 2011-12-16T03:09:48 *** yufra_
7 2011-12-16T03:23:09 *** yufra__
8 2011-12-16T03:24:18 *** yufra_
9 2011-12-16T03:39:06 *** yufra__
10 2011-12-16T05:14:43 *** qxcv
11 2011-12-16T09:31:38 *** raignarok
12 2011-12-16T10:28:15 *** greg_f
13 2011-12-16T10:32:29 *** raignarok
14 2011-12-16T12:49:55 *** greg_f
15 2011-12-16T12:51:55 *** greg_f
16 2011-12-16T13:48:46 *** dave_largo
17 2011-12-16T14:05:43 <ThomasWaldmann> moin
18 2011-12-16T14:09:22 *** qxcv
19 2011-12-16T14:09:38 *** qxcv
20 2011-12-16T14:22:29 <dreimark> moin
21 2011-12-16T14:31:33 <ThomasWaldmann> dreimark: http://www.google-melange.com/gci/task/view/google/gci2011/7180424 i just extended deadline, so there is enough time for review. can you do that?
22 2011-12-16T14:32:11 <ThomasWaldmann> (this task is taking much too long, hopefully he's finished soon)
23 2011-12-16T14:34:02 <ThomasWaldmann> and it looks like the 2nd wave has not been published yet, all new tasks are still in unpublished state
24 2011-12-16T14:39:02 <qxcv> ThomasWaldmann: about your comment on http://www.google-melange.com/gci/task/view/google/gci2011/7182222
25 2011-12-16T14:39:26 <qxcv> that particular .format() call isn't replacing printf-style string formatting
26 2011-12-16T14:40:34 <qxcv> should I actually make it printf-style, or I just keep it the way it is?
27 2011-12-16T14:40:51 <qxcv> initially it was just a static string
28 2011-12-16T14:41:40 <ThomasWaldmann> qxcv: well, if the other changes use % style, that change should be same style
29 2011-12-16T14:42:01 <ThomasWaldmann> and for your "...".format() changes: they are wrong
30 2011-12-16T14:42:35 <ThomasWaldmann> i also was not aware of this, but you cant just replace "..." % (...) by "...".format(...)
31 2011-12-16T14:42:51 <ThomasWaldmann> there is a difference how unicode is handled
32 2011-12-16T14:44:27 <ThomasWaldmann> you can just try "%s%s" % (u'ä', u'ö')
33 2011-12-16T14:45:04 <ThomasWaldmann> and "{foo}{bar}".format(foo=u'ä', bar=u'ö')
34 2011-12-16T14:45:09 <ThomasWaldmann> bbl
35 2011-12-16T14:51:10 <qxcv> that's funny
36 2011-12-16T14:51:35 <qxcv> looks like it's a bug, not a feature :( http://bugs.python.org/issue7300
37 2011-12-16T14:52:43 <qxcv> I googled it hoping for some wonderful explanation of why mixing strings/bytes is harmful, but all I found was that :/
38 2011-12-16T14:55:23 *** qxcv
39 2011-12-16T16:05:49 * dreimark looks at 7180424
40 2011-12-16T17:32:51 *** MattMaker
41 2011-12-16T17:37:14 *** MattMaker
42 2011-12-16T17:49:18 *** greg_f
43 2011-12-16T17:54:36 *** raignarok
44 2011-12-16T18:24:11 *** raignarok
45 2011-12-16T18:49:37 *** pkumar
46 2011-12-16T19:20:55 *** raignarok
47 2011-12-16T19:26:55 *** raignarok
48 2011-12-16T21:10:14 <ThomasWaldmann> re
49 2011-12-16T22:15:47 *** dave_largo
50 2011-12-16T22:20:55 *** raignarok
51 2011-12-16T22:31:24 *** raignarok_
52 2011-12-16T22:31:47 *** raignarok
53 2011-12-16T22:41:48 *** raignarok_
54 2011-12-16T22:55:47 <izibi> looks like some files aren't closed properly (at least on pypy 1.6)
55 2011-12-16T22:56:07 <izibi> i get errors like this: OSError: [Errno 24] Too many open files: '/home/moin/moin-2.0/wiki/index'
56 2011-12-16T22:56:09 *** MattMaker
57 2011-12-16T22:58:15 *** pkumar
58 2011-12-16T23:00:24 <ThomasWaldmann> izibi: yeah, we have an issue for that
59 2011-12-16T23:05:52 <ThomasWaldmann> ah, new werkzeug release
60 2011-12-16T23:06:13 <izibi> and on pypy it just crashed after a random number of requests. very strange (no error message/traceback)
61 2011-12-16T23:06:38 <izibi> exits with error code 245
62 2011-12-16T23:07:50 <ThomasWaldmann> izibi: i ran it on 1.6 for quite many requests
63 2011-12-16T23:09:51 <izibi> what are "quite many requests"?
64 2011-12-16T23:10:06 <izibi> only happens with >1000 request for me
65 2011-12-16T23:10:15 <izibi> probably because the file limit is 1024
66 2011-12-16T23:10:19 <ThomasWaldmann> for sure it was 1000
67 2011-12-16T23:10:26 <ThomasWaldmann> not sure whether it was more
68 2011-12-16T23:10:51 <ThomasWaldmann> btw, you can increase the limit :)
69 2011-12-16T23:11:03 <izibi> i know
70 2011-12-16T23:11:32 <ThomasWaldmann> if you want to work on the open files issue, you can claim a hard bug fix for it
71 2011-12-16T23:12:23 <izibi> did you execute all requests in one ab call?
72 2011-12-16T23:12:39 <ThomasWaldmann> maybe not
73 2011-12-16T23:12:51 <ThomasWaldmann> i wanted to now how fast it is after N request
74 2011-12-16T23:13:15 <ThomasWaldmann> so i executed 100, 100, 100, 100, 100 and looked how request rate climbed
75 2011-12-16T23:14:31 <izibi> looks like it's also an open files issue
76 2011-12-16T23:14:41 <izibi> at least the output of lsof get's pretty long
77 2011-12-16T23:15:24 *** MattMaker
78 2011-12-16T23:17:12 <ThomasWaldmann> my impression was that it is caused to a big part by whoosh
79 2011-12-16T23:17:24 <ThomasWaldmann> but i reviewed whoosh code and didn't find something definite
80 2011-12-16T23:18:21 <izibi> probably. part of the lsof output: http://paste.pocoo.org/show/521905/
81 2011-12-16T23:20:20 <izibi> this makes benchmarking with ab practically impossible
82 2011-12-16T23:21:04 <izibi> as it already crashes after 10-30 requests on pypy 1.7
83 2011-12-16T23:21:20 <ThomasWaldmann> that's rather fast
84 2011-12-16T23:21:30 <ThomasWaldmann> and on 1.6?
85 2011-12-16T23:21:42 <izibi> i could also crash it by holding down F5 in the browser
86 2011-12-16T23:22:07 <izibi> on pypy only after more than 1000 requests (had no problem with running with -n 1000)
87 2011-12-16T23:22:15 <ThomasWaldmann> you're sure that it is due to 1.7 and not due to your build?
88 2011-12-16T23:23:02 <izibi> well i hope that the build from the arch repo is working ;)
89 2011-12-16T23:23:43 <ThomasWaldmann> (23:20) < izibi> on pypy only after more than 1000 requests (had no problem with running with -n 1000)
90 2011-12-16T23:23:46 <ThomasWaldmann> what pypy?
91 2011-12-16T23:24:12 <izibi> 1.6
92 2011-12-16T23:25:06 <izibi> i need a system with old software to run the official build :D
93 2011-12-16T23:25:54 <ThomasWaldmann> ok, so it looks like a regression of 1.7? did you ask the pypy guys?
94 2011-12-16T23:26:54 <izibi> do you have pypy 1.7 available? would be easier than setting up a vm for that i guess :D
95 2011-12-16T23:27:33 <ThomasWaldmann> well, i didn't try it yet, otherwise there would've been no task :D
96 2011-12-16T23:28:47 <izibi> i guess they'll just say that moin and/or whoosh is/are crap :D
97 2011-12-16T23:28:49 <izibi> "The effect that you get in CPython has clearly been described as a side-effect of the implementation and not a language design decision: programs relying on this are basically bogus."
98 2011-12-16T23:28:53 <izibi> ;)
99 2011-12-16T23:28:56 <izibi> http://doc.pypy.org/en/latest/cpython_differences.html#differences-related-to-garbage-collection-strategies
100 2011-12-16T23:30:44 <ThomasWaldmann> well, but you found that 1.6 and 1.7 is different, right?
101 2011-12-16T23:31:02 <izibi> yeah, it got worse :D
102 2011-12-16T23:32:22 <izibi> i could write a custom benchmark tool that waits for a short time before doing the next request
103 2011-12-16T23:32:39 <izibi> ab doesn't have an option for that as far as i know
104 2011-12-16T23:39:19 <ThomasWaldmann> hmm, tried with 1.7, doesn't work at all for me
105 2011-12-16T23:40:12 <ThomasWaldmann> i don't think waiting time helps
106 2011-12-16T23:40:32 <izibi> what do you mean by "not at all"? doesn't even start?
107 2011-12-16T23:40:58 <ThomasWaldmann> ehrm, with 127.0.0.1 it works
108 2011-12-16T23:42:27 <izibi> i just start it with 0.0.0.0 to do testing over the network. but this doesn't really make a difference
109 2011-12-16T23:43:11 <ThomasWaldmann> your ab commandline?
110 2011-12-16T23:44:24 <izibi> ab -n 100 -c 1 http://127.0.0.1:5000/Home
111 2011-12-16T23:44:43 <izibi> didn't test if it crashed if there's no Home item
112 2011-12-16T23:46:30 <ThomasWaldmann> it survived
113 2011-12-16T23:46:42 <izibi> all 100 requests?
114 2011-12-16T23:47:04 <izibi> with a Home item existing?
115 2011-12-16T23:47:41 <izibi> and what's your open file limit? (ulimit -n)
116 2011-12-16T23:48:16 <ThomasWaldmann> OSError: [Errno 24] Too many open files: '/home/tw/w/moin-2.0/wiki/index'
117 2011-12-16T23:48:34 <ThomasWaldmann> i mean it did not crash/terminate
118 2011-12-16T23:48:38 <ThomasWaldmann> but now out of files
119 2011-12-16T23:48:56 <izibi> that's what i get with pypy 1.6
120 2011-12-16T23:49:28 <izibi> maybe whoosh does some stange stuff if it reaches the limit?
121 2011-12-16T23:50:42 <ThomasWaldmann> ok, i suggest you raise your file limit to a very high value and just ignore that issue
122 2011-12-16T23:59:25 <izibi> still crashes with a 2 second delay between the request