PMXBOT Log file Viewer

Help | Karma | Search:

#pypa logs for Wednesday the 28th of May, 2014

(Back to #pypa overview) (Back to channel listing) (Animate logs)
[03:54:08] <radix> pip seems to not be picking up the "extras" dependency from testtools, specified in this setup.py: https://github.com/testing-cabal/testtools/blob/master/setup.py
[03:54:18] <radix> is there any obvious reason why it wouldn't? (maybe I'm completely misunderstanding how pip works)
[03:56:59] <Ivo> radix: could you paste the log of pip's output?
[03:57:17] <radix> I lost it, I'll try to reproduce in a new virtualenv
[03:58:24] <radix> oh.
[03:58:43] <radix> ok, I guess it worked, I think this only happened because I had a temporary error in some completely unrelated package during a pip install -r dev-requirements.txt
[03:58:55] <radix> sorry :)
[04:02:45] <dstufft> radix: probably the extras aren't installed if the package itself is already installed
[04:03:34] <dstufft> so likely the first time it added the extra packages to the list of things to install, installed testtools, then bailed out before installing the rest, then the second time around it saw testtools was already installed and didn't bother to do anything else
[04:03:40] <dstufft> probably we should attempt to fix that
[04:05:26] <radix> no, it was actually because twisted failed to install, because I didn't have python-dev installed
[04:05:45] <radix> my dev-requirements.txt just had testtools and twisted, and it installed testtools, then twisted, and didn't get around to installing extras yet
[04:05:54] <radix> when twisted failed
[07:44:31] <moldy> hi
[07:44:52] <moldy> which package is recommended for setting up a private package index? right now, i am trying to get pypiserver to work
[11:35:09] <agronholm> moldy: devpi?
[12:11:42] <moldy> agronholm: i took a look at devpi, but the docs were a little off-putting
[12:11:59] <agronholm> well, it's what people in the know recommend
[12:12:27] <agronholm> haven't used it myself
[12:12:34] <moldy> agronholm: the only documented deployment method seems to be deprecated
[12:12:41] <agronholm> heh
[12:12:53] <agronholm> sounds just like python packaging to me
[12:13:12] <moldy> pypiserver seems to work now. now i am looking into password-protecting it.
[14:23:32] <moldy> any idea how to make pypiserver auto-mirror pypi?
[14:25:41] <jezdez> moldy: I don't think it can mirror the whole index
[14:25:53] <jezdez> there is https://pypi.python.org/pypi/bandersnatch for that
[14:27:54] <moldy> jezdez: i don't need to mirror the whole index, but it would be nice if it mirrored "upon request". it seems it cannot do that either, though
[14:34:31] <jezdez> moldy: ah, yeah :-/
[14:58:20] <DanielHolth> moldy devpi is the favorite one now ?
[15:08:06] <moldy> DanielHolth: i guess i will look into it when i have some more time on my hands
[15:08:26] <moldy> DanielHolth: i'd be interested in hearing about the recommended way to deploy it.
[15:09:18] <moldy> if i understand it correctly, their official documentation describes one way, and that one is "deprecated" ;)
[15:30:47] <tomprince> moldy: One of the pypa people wrote https://pypi.python.org/pypi/pipa
[15:35:07] <vvladimirov> Hi. Could you please point me how can I build wheel package and include some external library which my code depends on into it?
[15:36:59] <xafer> so the 3.2 travis issue seems to come from scripttest being stuck on proc.communicate... I guess the PIPEs could be replaced by tempfile ??
[16:02:20] <radix> so I just uploaded the "effect" package to pypi, and I expected "pip install" to intsall the wheel version, but it seems to have downloaded the .tar.gz instead
[16:02:40] <Alex_Gaynor> radix: which version of pip?
[16:02:45] <dstufft> radix: what version of -- yea that
[16:02:52] <radix> hm, 1.4.1
[16:03:03] <Alex_Gaynor> radix: yeah, you need to enable wheels by default in versions before 1.5
[16:03:12] <Alex_Gaynor> with 1.5 it totally installs the wheel (if you provide --pre)
[16:03:46] <dstufft> note: pip 1.4 doens't install wheels quite right, it was experimental/opt in there
[16:03:48] <radix> ohh, the new version has colors too
[16:04:40] <dstufft> yes
[16:04:42] <dstufft> pretty colors
[16:04:52] <dstufft> makes stuff stand out more
[16:05:09] <radix> man it has been a super long time since I have actually released a package
[16:05:14] <radix> thanks dstufft, Alex_Gaynor
[16:46:18] <moldy> tomprince: thanks, i will take a look
[19:15:38] <miketheman> Ivo, tomprince : thanks for your help yesterday - I think we got to the root of the problem - the issue was indeed some sort of race condition where a background job was cleaning up /tmp/pip_build_root/ periodically on each node (at a random interval) and during installation, some nodes were being cleaned up post-unpack, leading to the invalid wheel message.
[19:55:52] <djmitche> hey folks, I've gotten myself stumped on testing a package that adds a setup() argument. It works on my system, but not on Travis. https://github.com/travis-ci/travis-ci/issues/2355
[19:56:21] <djmitche> Any setuptools gurus see anything familiar there? I've run out of debugging options (and the logging isn't telling me much)
[20:52:10] <ionelmc> is there a point to having pip issue warnings for unmatching "global-exclude *.py[co]" (or similar) from MANIFEST.in ?
[20:52:51] <ionelmc> to ask the other way, are those sort of exclude rules ever necessary ?
[21:02:21] <tos9> ionelmc: no, because pip only includes .pys by default
[21:02:30] <tos9> and you should be careful with overly general includes
[21:03:09] <tos9> djmitche: oh geez what the hell.
[21:05:54] <djmitche> that's what I'm saying :)
[21:08:25] <ionelmc> tos9: how about easy_install ?
[21:11:27] <ionelmc> actually, why does pip read the manifest anyway ?
[21:11:48] <ionelmc> it's supposed to be used when doing the sdist
[21:11:56] <ionelmc> s/manifest/MANIFEST.in/
[21:27:44] <djmitche> tos9: so I set it to run setup.py egg_info before the pip install -e, and now https://travis-ci.org/djmitche/build-relengapi#L143
[21:28:06] <djmitche> which I think indicates that it couldn't find an installed copy of Flask
[21:28:10] <djmitche> which makes sense since it hasn't installed it
[21:28:12] <djmitche> but .. why not?
[21:33:15] <djmitche> http://25.media.tumblr.com/tumblr_lm5bvelvTL1qercd3o1_500.gif
[21:35:40] <djmitche> https://travis-ci.org/djmitche/build-relengapi/builds/26254048 .. but, Flask is in install_requires for the app being tested, and for a bunch of its other requirements
[21:51:16] <dstufft> pip doesn't read the MANIFEST file
[21:51:22] <dstufft> setuptools/distuils does
[21:52:33] <aclark> djmitche!
[21:58:41] <djmitche> aclark: Alex!
[22:02:42] <djmitche> haven't seen you in a while :)
[22:21:42] <aclark> djmitche: hey still at Mozilla?
[22:21:53] <djmitche> I am!
[22:22:00] <aclark> Great!
[22:22:15] <djmitche> and learning Flask and such
[22:22:17] <djmitche> good times
[22:22:21] <aclark> Ah, nice