Log file Viewer

[13:59:45] <jbaiter> Is it possible to create 'universal' Python2/Python3 wheels on Mac OSX that are nevertheless restricted to x64/i686?

[14:00:33] <jbaiter> I have a package that ships some native code that does not link against Python (but is accessed with cffi), so it's not necessary to compile it for different versions of CPython

[14:02:09] <jbaiter> I've tried with 'mypackage-0.1-py2.py3-none-macosx_10_9_x86_64.whl', but it's not accepted since the only architecture 'py2' allows is 'macosx_10_0_universal'

[14:03:29] <jbaiter> the same pattern works on windows and linux, i.e. 'mypackage-0.1-py2.py3-none-{win32,win_amd64,manylinux1_x86_64,manylinux1_i686}.whl' works like a charm

[14:06:36] <dstufft> jbaiter: might just be a missing platform in pip/pep425tags.py

[14:09:14] <jbaiter> dstufft: i'll check, thanks!

[14:54:54] <jbaiter> ah, it's already fixed in 8.1 :-) i was using 8.0

[16:20:20] <jaraco> dstufft: As I’m looking through pip for how it handles `simple` links (aka PEP 503), I’m frustrated that there’s also a `simple` package used in a lot of the tests, making it difficult to search the code.

[16:21:21] <jaraco> How would you feel about an issue/PR to rename that `simple` package to maybe `sample` or `sandbox` or similar to disentangle it from a “simple index”?

[17:02:47] <dstufft> jaraco: fine with me

[17:03:00] <dstufft> jaraco: fwiw pip/index.py should be all that logic

[17:03:39] <jaraco> dstufft: Thanks for the tip. I’ve found it!

[17:05:37] <jaraco> The code is very readable and logical.

[17:05:44] <jaraco> !m dstufft

[17:05:44] <pmxbot> you're doing good work, dstufft!

[17:15:53] <dstufft> jaraco: it *should* match what's in PEP 503- if you find something that doesn't it's a bug :]

[17:17:18] <jaraco> What’s confused me is the behavior when using devpi. https://bitbucket.org/hpk42/devpi/issues/354/pip-and-easy_install-require-different

[17:18:07] <dstufft> oh

[17:18:11] <dstufft> I think devpi does something fancy there

[17:18:29] <dstufft> I *think* devpi inspects the UA and if it's pip it serves the /simple/ regardless

[17:18:30] <dstufft> or something

[17:19:13] <dstufft> but whatever works for easy_install should work perfectly fine for pip... modulo some stuff that we removed that I don't think easy_install ever removed

[17:19:21] <jaraco> dstufft: I’m pretty sure pip is traversing links from the non-simple page… which leads to pages with the download links.

[17:20:07] <jaraco> oh, fschulze confirms - devpi is doing user-agent detection and redirecting.

[17:20:16] <dstufft> If we're given an URL like https://devpi.yougov.net/root/yg and we're looking for supervisor, we should request https://devpi.yougov.net/root/yg/supervisor/ and that's it

[17:20:30] <dstufft> thought so :]

[17:20:39] <dstufft> I would recommend they use the same URL for easy_install and pip

[17:20:44] <dstufft> and not do the redirect trick at all

[17:20:49] <jaraco> That’s the route we’re going.

[17:20:59] <jaraco> !schneier dstufft

[17:21:00] <pmxbot> SSL is invulnerable to man-in-the-middle attacks. Unless that man is dstufft.

[17:21:14] <dstufft> pip shouldn't be special, anything that implements PEP 503 should be able to be dropped in, in place

[17:21:31] <dstufft> that's the goal anyways :]

[17:23:56] <dstufft> jaraco: I was thinking about spinning out the PEP 503 stuff in pip/index.py and seeing about adding it to packaging-- not sure if it'd make sense or not though it might be higher level than what packaging should be. Any interest in something like that from setuptools side? Or prefer ot keep the same code you're using now?

[17:24:34] <jaraco> I’m eager to harmonize in every way possible.

[17:25:47] <dstufft> I might sketch up some API ideas and post them to pypa-dev then- It's a bit different than the other stuff in there in that we want it to be flexibile enough that setuptools can still discover eggs and such

[23:38:57] <Nakato> dstufft: Is pip not going to be using a develop branch anymore?

[23:39:18] <dstufft> Nakato: correct

[23:39:28] <dstufft> master branch lik e98% of all other git projects :]

[23:42:41] <Nakato> Cool, just wanted to double check before I started to assume that. Gotta fix PBR's gate. :) I like the change, it kept biting me locally.