Log file Viewer

[01:56:59] <Somelauw> I wish pip would integrate with apt-get. It takes quite some effort to compile scipy.

[02:06:57] <[Tritium]> Somelauw: that would mean running pip as root

[02:07:00] <[Tritium]> which is bad

[02:07:36] <[Tritium]> sudo apt-get build-dep python-scipy && .venv/bin/pip install scipy

[02:08:44] <Somelauw> I forgot I could have done build-dep

[02:08:53] <Somelauw> so I installed openblas and fortran

[02:09:14] <[Tritium]> build-dep would have installed those too

[02:09:31] <[Tritium]> well, it would have installed a fortran and a blas

[02:10:35] <Somelauw> I don't mind the blas, but openblas was listed as optimized but still free somewhere

[02:11:41] <[Tritium]> all I know about blas is that its not a library, its a class of library, that implements the sameish api, and there are many implementations

[13:57:47] <iElectric> dstufft: around?

[13:58:41] <iElectric> dstufft: is there any reason we shouldn't implement https://github.com/pypa/pip/issues/2677

[13:58:46] <iElectric> or something that makes it hard?

[14:24:35] <iElectric> I see now

[14:24:55] <iElectric> dstufft: I'm assuming you'd prefer we have prefix as an explicit option

[14:25:11] <iElectric> rather than extracting it from install_option

[16:07:31] <dstufft> iElectric: sorry was away

[16:08:29] <dstufft> iElectric: and yes, --install-tion and such shouldn't be mapped to wheels, if we need something that provides that can be generically applied to wheels it should get it's own flag and then pip should map that new flag to the appropiate install option if need be

[16:08:31] <iElectric> dstufft: no problem

[16:08:42] <iElectric> I have changes ready to add --prefix

[16:08:48] <iElectric> just need to write unit/func tests

[17:48:30] <iElectric> dstufft: any chance you could review https://github.com/pypa/packaging/pull/36

[17:50:50] <dstufft> iElectric: I thought unittest2 got fixed? I don't really like that because it makes some errors (like passing a None into Version()) either silently pass or give a more confusing error. I don't think it's a good idea to pass anything but a str() into the setup(version=...) kwarg.

[17:51:57] <iElectric> but really that code uses regex and expects a str()

[17:52:34] <iElectric> you could also catch None and print a reasonable error

[17:53:33] <iElectric> dstufft: currently unittest2 can't be build as a wheel because of that (latest version on pypi)

[17:55:10] <iElectric> dstufft: do you know if wheels support setuptools namespaces?

[17:55:34] <dstufft> iElectric: yea but _everything_ has a str(), a typeerror is better than an implicit cast

[17:56:00] <dstufft> iElectric: what do you mean by "supports"? it'll drop the __init__.py in place if it's part of the wheel

[17:56:36] <iElectric> that's good enough, thanks

[18:06:45] <iElectric> dstufft: one more thing regarding wheels, does pip install bla.whl find existing dependencies from $PYTHONPATH

[18:07:05] <dstufft> iElectric: it should yes, as long as they satisify the constraints and you haven't used --upgrade

[18:07:14] <iElectric> hmmm

[18:07:50] <iElectric> it's trying to collect the dependency

[18:07:59] <iElectric> that shouldn't happen if six is in $PYTHONPATH somewhere right?

[18:11:07] <iElectric> dstufft: any idea what else could it be?

[18:11:32] <dstufft> does the six that is in Python path have a .dist-info or a .egg-info directory?

[18:11:36] <xafer> six metadata are also in the path ?

[18:12:03] <iElectric> yes

[18:12:36] <iElectric> /nix/store/jiavgr8kmg7hhnr22w9mapm3fb1p5942-python2.7-six-1.10.0/lib/python2.7/site-packages/six-1.10.0.dist-info/

[18:12:42] <iElectric> and /nix/store/jiavgr8kmg7hhnr22w9mapm3fb1p5942-python2.7-six-1.10.0/lib/python2.7/site-packages/ is in $PYTHONPATH

[18:12:54] <dstufft> iElectric: does ``pip list`` show it?

[18:13:04] <iElectric> no

[18:14:16] <iElectric> pip (6.1.1)

[18:14:18] <iElectric> setuptools (15.2)

[18:14:20] <iElectric> wheel (0.26.0)

[18:14:22] <iElectric> that's it

[18:14:49] <iElectric> so I'm using python setup.py bdist_wheel to build the wheel

[18:14:57] <iElectric> and then pip install dist/*.whl to install it

[18:15:16] <iElectric> while six dependency is in $PYTHONPATH

[18:15:20] <iElectric> built the same way

[18:16:16] <dstufft> iElectric: does PYTHONPATH=... python -c "import pkg_resources; pkg_resources.get_distribution('six')" raise an error?

[18:17:17] <iElectric> nope

[18:17:58] <xafer> and this happens inside a venv ?

[18:18:11] <iElectric> no, that's my Nix build

[18:18:43] <iElectric> but it really just uses pip and wheel

[18:18:52] <iElectric> and sets PYTHONPATH for built dependencies

[18:22:22] <iElectric> does it need any .pth files?

[22:07:02] <iElectric> got it.

[22:08:04] <ghsk> hi, i'm trying to get a local virtualenv install ( ~ python virtualenv.py test with python 2.6 ) . I'm getting AddressValueError excpetion when installing pip. Same cmd worked for 2.7, does anyone know how to deal with this ? thanks

[22:08:22] <ghsk> fyi https://virtualenv.readthedocs.org/en/latest/installation.html is what I'm using

[22:08:39] <ghsk> maybe I should try older virtualenv ?

[22:11:55] <iElectric> dstufft: so would pip detect if it had requests 1.x and requests 2.x in $PYTHONPATH?

[22:13:23] <iElectric> and raise a conflict

[22:18:27] <iElectric> I guess I'm asking what function figures out the conflict resolution