Log file Viewer

[00:00:58] <sontek> adourado: Yeah, so you just have a custom build_ext class that builds up the extensions you want inside it

[00:01:11] <sontek> adourado: In your specific case you would return no extensions on windows

[00:01:16] <sontek> and you can do that with sys.platform

[00:03:20] <adourado> sontek: this will always ignore the erros, though, right?

[00:03:32] <adourado> I would rather only do that if specified

[00:10:28] <sontek> adourado: You can support random arguments, so you could add something like this for --no-extensions: https://github.com/datastax/python-driver/blob/master/setup.py#L169

[00:11:32] <adourado> sontek: I can't put that on a install_requires/extra_requires, though

[00:11:52] <adourado> so there would be no way to another package to require the extensionless version

[00:12:07] <adourado> and that is already the problem I have

[00:14:32] <sontek> adourado: You could do it in extras

[00:14:43] <sontek> so if you wanted mylib[with_extensions]

[00:15:49] <adourado> extra and install don't accept custom parameters to setup.py. Just regular requirements specifiers

[15:04:43] <muelli> hey hye. I'm wondering whether http://paste.ubuntu.com/12321300/ is the expected behaviour. In short: I cannot update libraries which are already installed on the system when I've created virtualenv with --system-site-package.

[15:10:43] <ionelmc> muelli: does --ignore-installed help?

[15:10:51] <ionelmc> eg, pip install -I -U stuff

[15:40:13] <muelli> ionelmc: hm. doesn't seem to be the case: http://paste.ubuntu.com/12321521/

[15:52:35] <ionelmc> must be one of those things that only happen on debian :)

[15:53:23] <muelli> ionelmc: hm. I don't think I have a special setup. It's an ordinary ubuntu vivid :-\

[15:55:47] <ionelmc> muelli: i can't reproduce on trusty, dunno what's going on

[15:57:22] <tdsmith> is there a pip/setuptools/? API for normalizing version numbers? i'm running into a situation where pip.get_installed_distributions() is giving me a normalized version number which is not identical to the canonical version number, but i can't use that version to interact with the pypi JSON API because it only handles exact version matching, not equivalent version matching, so it gives me 404s

[15:57:35] <tdsmith> so i figure i'll just ask pypi about all versions and then do the matching myself

[15:58:58] <ionelmc> tdsmith: afaik there's a package for that (one of the vendored deps of pip/setuptoools)

[16:03:18] <tdsmith> looks like pkg_resources.safe_version will do it

[16:17:22] <muelli> ionelmc: hm. funny. Are you running the exact same commands as I do?

[16:17:42] <ionelmc> muelli: i think so :)

[16:18:03] <ionelmc> i got an urlib3 installed globally

[16:18:17] <ionelmc> is your urllib provided by apt?

[16:19:15] <muelli> ionelmc: seems so. At least apt-cache policy python-urllib3 shows 1.9.1. so does pip.

[16:22:27] <ionelmc> lol, in trusty there's python-urllib3-whl and python-urllib3

[16:22:31] <ionelmc> what gives

[16:22:52] <ionelmc> "This package contains the universal wheel."

[16:23:03] <ionelmc> lol

[16:30:23] <ionelmc> looks like it put a whl file in /usr/share/python-wheels

[16:30:28] <ionelmc> what's going on here?

[16:36:47] <Wooble> I don't know but https://www.debian.org/doc/packaging-manuals/python-policy/ch-module_packages.html makes me think they don't understand what wheels are for to begin with? :/

[16:40:09] <ionelmc> Wooble: could it be that they provide a wheel cache?

[16:40:44] <ionelmc> sounds interesting but doesn't it mean that they also have to patch pip and virtualenv more than they already do?

[16:41:05] <Wooble> apparently those wheels only exist for "enabling the pip tool" but I have no idea what that means or why it would need distro-installed wheels. :/

[16:42:32] <ronny> ionelmc: even better, their rule for wheels ensures they build broken ones (enforcement of the uiversal flag)

[16:51:33] <ionelmc> oh

[16:51:59] <ionelmc> i had a good feeling about it for a brief moment

[16:55:40] <ronny> ionelmc: thats the uplifts will fal harder syndrome to help you better appreciate their technical fail

[17:11:04] <dstufft> tdsmith: fwiw pkg_resources uses the packaging lib for that

[17:11:14] <dstufft> if you want to be able to control what version of packaging you get easier

[17:11:29] <dstufft> https://pypi.python.org/pypi/packaging

[17:12:16] <ronny> dstufft: btw, i'd like to discuss a way to give linux distros sane multi version cappable de-vendoring and a preinstalled/symlinkable package repo format

[17:16:07] <ronny> dstufft: but i suppose you are stripped for time for a while longer?

[17:48:35] <tdsmith> ty dstufft

[20:44:24] <moldy> hiiii

[20:44:35] <moldy> opps, sorry. keyboard got stuck :)

[20:44:56] <moldy> is there a way to make pip cache tar.gz downloads?

[20:45:10] <moldy> and also git checkouts?

[21:40:02] <ionelmc> moldy: doesn't it already cache the sdists (in the http cache thing)?

[21:43:46] <moldy> ionelmc: i need to experiment more to be sure, but so far my impression was that it does not cache tar.gzs and/or git checkouts

[21:44:08] <moldy> ionelmc: let me see if i can confirm this...

[21:46:00] <ionelmc> git checkout might be cached, for editable installs

[21:46:06] <ionelmc> hmmm

[21:48:47] <moldy> as far as i see here, pip https://whatever never gets cached

[21:50:08] <moldy> https://dpaste.de/6n9H

[22:28:50] <tdsmith> is there a good best-practice example for installing documentation?

[22:39:20] <agronholm> what kind of documentation? what format?

[22:39:51] <tdsmith> just a README.txt, let's say

[22:40:07] <tdsmith> something prerendered

[22:41:07] <tdsmith> using data_files with a "share/doc/myproject" suffix seems like an option

[23:51:25] <ionelmc> data_files is not very consistent iirc

[23:51:59] <ionelmc> there was a ticket somewhere with list of weird differencess

[23:52:45] <ionelmc> * For relative paths:

[23:52:45] <ionelmc> * ``setup.py install`` with ``setuptools`` put them inside the egg zip/dir

[23:52:45] <ionelmc> * ``setup.py install`` with ``distutils`` and ``pip install`` put them in ``sys.prefix`` or ``sys.exec_prefix``