Log file Viewer

[06:20:11] <bsc> ping. Anyone here? Had a doubt regarding setuptools

[12:24:18] <ronny> bsc pong?

[12:24:36] <ronny> bsc: you should have written the doubt as well, ^^

[12:53:25] <mjacob> according to https://python-packaging-user-guide.readthedocs.io/en/latest/distributing/#platform-wheels uploading linux wheels to PyPI is not possible currently

[12:53:51] <mjacob> in a comment to a stackoverflow answer, someone said that might not be true anymore

[12:59:06] <apollo13> mjacob: no that is still the case

[12:59:11] <apollo13> you can upload manylinux wheels though

[12:59:12] <dstufft> no

[12:59:19] <dstufft> you can upload both kinds of wheels now I think

[12:59:23] <apollo13> but not the distro specific ones

[12:59:25] <apollo13> oh

[12:59:34] <apollo13> dstufft: when did you enable that :D

[12:59:39] <dstufft> there is no such thing as a "distro specific wheel" atm

[12:59:47] <dstufft> it's just "way too broad" and "manylinux"

[12:59:54] <dstufft> I think at the same time I enabled manylinux

[12:59:59] <apollo13> ah okay

[13:00:13] <apollo13> I thought you only enabled manylinux and still forbid the others cause the just don't work ;)

[13:00:31] <mjacob> ok, i'll search for manylinux

[13:00:45] <mjacob> that's a restriction only for the central pypi, right?

[13:01:05] <apollo13> yes, your own package index can host whatever it wants

[13:01:32] <apollo13> but installing linux wheels in general will only work if glibc version match etc… (unless it conforms to the manylinux spec)

[13:02:03] <mjacob> sure

[13:02:42] <mjacob> maybe it makes sense to update the note in https://python-packaging-user-guide.readthedocs.io/en/latest/distributing/#platform-wheels ?

[13:03:15] <apollo13> that would certainly be a good idea

[13:04:35] <mjacob> by the way, is it correct to say "according to the documentation" when talking about wheels and citing the python packaging user guide?

[13:29:11] <mjacob> dstufft: thank you for releasing virtualenv 15.0.2 btw

[13:38:05] <Vigdis> Hi, I'm a package maintainer for OpenBSD and I've a problem which seems to be related to https://bitbucket.org/pypa/pypi/issues/438/backwards-compatible-un-hashed-package

[13:39:11] <Vigdis> packages takes multiple hours to be reachable through the 'stable' url

[13:39:23] <Vigdis> the url for instance is https://pypi.io/packages/source/p/pycares/pycares-2.0.1.tar.gz

[13:40:43] <Vigdis> for instance, pycares was release on the 27th and yesterday I couldn't fetch the tar.gz through this url, but this morning it worked

[13:40:58] <Vigdis> is there a way to cut the delay?

[13:41:24] <Vigdis> poke dstufft ^

[13:43:01] <Vigdis> another example: now, if I try to fetch https://pypi.io/packages/source/F/Flask/Flask-0.11.tar.gz, I got redirected to https://files.pythonhosted.org/packages/source/F/Flask/Flask-0.11.tar.gz and then I get file not found

[13:43:18] <Vigdis> and I guess in a few hours I'll be able to get the file :)

[21:06:33] <bgold> I'm new to python packaging, and I'm trying to figure out the best way to package a module that has a dependency on a .jar file. Wheels look right, but if I package it as a wheel, will that affect users' ability to pip install?

[21:08:00] <tdsmith> pip likes wheels

[21:08:58] <bgold> :tdsmith so it would only affect it in a good way then? that sounds great.

[21:13:17] <gsnedders> I have a huge number of tests that I want to run against loads of old versions of setuptools. Is there any way I can sanely automate this? I guess assume the tests are stdlib unittest, and I just want to store the results of each test run in some machine readable format. Is there any sane way to do this? Have a script that creates virtualenvs and then runs the tests?

[21:18:58] <bgold> gsnedders: I'm not experienced with it, but I was just wondering a similar thing. Would using a free service like Travis CI with a build matrix work? You could give it a list of python versions (back to 2.6) and then use environmental variables to run against specific versions of setuptools. https://docs.travis-ci.com/user/languages/python#Build-Matrix

[21:19:44] <bgold> but that's not local

[21:20:45] <gsnedders> bgold: yeah, I could, and I think the config for travis ci would be way simpler than something equiv for tox, sadly

[21:21:05] <gsnedders> bgold: I mean I have no objection to doing it publicly, just…

[21:25:31] <bgold> gsnedders: out of curiosity what makes you hesitate? I'm not too experienced, so I wouldn't know if there were a common reason to not use it or to choose tox over travis.

[21:28:14] <bgold> gsnedders: does this do the same thing, more or less? https://tox.readthedocs.io/en/latest/example/basic.html#compressing-dependency-matrix

[21:28:19] <dstufft> gsnedders: tox?

[21:28:47] <dstufft> gsnedders: tox has matrix support fwiw

[21:29:17] <dstufft> gsnedders: the example uses Django, but: http://tox.readthedocs.io/en/latest/example/basic.html#compressing-dependency-matrix

[21:29:31] <gsnedders> bgold, dstufft: yeah, but then I end up with setuptools{18,19,20,21,22} and then below setuptools18: setuptools==18 etc., which gets pretty repetative

[21:31:49] <dstufft> yea I suppose

[21:35:45] <gsnedders> whereas on travis I can do something like have a load of environment variables with the version number and then not duplicate that

[21:36:12] <gsnedders> bgold: as for why hesitate about doing it remotely, it just slightly increases the latency when hacking on stuff

[21:37:37] <bgold> gsnedders: That makes total sense. Being able to test quickly and locally is such a boon for getting things done.

[23:23:53] <rardiol> does setuptools know enough to create .desktop files from entry_points?