Log file Viewer
#pypa-dev logs for Friday the 19th of February, 2016
[00:53:35] <lifeless> njs: putting each thing in its own tree and just setting PATH and PYTHONPATH is effectively free
[00:57:08] <dstufft> besides the fact that pip has to be installed *into* a virtual environment, so you'd have to invoke a subprocess pip inside of the virtual environment which may or may not be the same version as the currently executing pip
[00:59:18] <lifeless> njs: (Also consider that the wheel from C may differ depending on the version of B - as I believe the numpy ABI stuff implies?
[01:18:51] <njs> lifeless: if you have a single empty virtualenv to work with + PYTHONPATH/PATH manipulations to manipulate it, I think that gives you the best of both worlds
[01:28:27] <dstufft> seems like it'd be easier to just use python -S instead of a empty virtual environment
[02:51:38] <njs> meh, no, there's PYTHONNOUSERSITE but that only applies to the user directory, it doesn't disable site.py as a whole
[14:42:47] <pmxbot> jaraco pushed 2 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :
[16:03:23] <pmxbot> jaraco pushed 2 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :
[16:03:23] <pmxbot> Fix typo, correcting failures on late Pythons when mock is not already installed.
[16:26:36] <pmxbot> jaraco pushed 6 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :
[16:26:37] <pmxbot> Merge pull request series from Steve Kowalik for updated and unified marker implementation. Fixes #122.
[16:50:59] <pmxbot> jaraco pushed 3 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :
[17:37:38] <xafer> Strange, packaging.requirements.Requirement('django==1.6+azeze') is accepted but packaging.requirements.Requirement('django>=1.6+azeze') is not
[17:38:23] <xafer> (The >= one raises InvalidRequirement: Invalid requirement, parse error at "'+azeze'") cc StevenK
[19:01:33] <xafer> Well, https://www.python.org/dev/peps/pep-0440/#inclusive-ordered-comparison seems to disallow that :-/
[19:35:47] <xafer> Include the specified version with this local label and local label strictly bigger + the version strictly bigger to the version without label
[19:36:44] <xafer> Since "Comparison and ordering of local versions" is defined in the PEP, this could make sense
[19:49:23] <xafer> Since there is a whole paragraphs on local version comparison and ordering in the PEP440 maybe it would make sense to allow them ?
[22:30:07] <pmxbot> jaraco pushed 2 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :
[22:44:29] <xafer> Well most have already been commited to develop and would need to be cherry-picked in master I guess ?
[22:45:03] <njs> https://github.com/pypa/pip/issues/3486 is still on my todo list, I want to get to it soon but given how "wanting" works I'm reluctant to suggest holding up a release for it
[22:46:20] <xafer> dstufft: I guess I could probably do the cherry-picking but I'm not sure out you want to handle maintenance releases
[23:08:52] <xafer> njs: I feel the main change for the issue 3486 is to reset show_stdout back to True or am I missing something ?
[23:09:54] <njs> xafer: show_stdout is set in for subprocess runs that are supposed to unconditionally show stdout -- the only thing that uses it is some (random?) subset of the calls to VCS commands
[23:11:00] <njs> xafer: for the main uses of that function (like calling setup.py), what's actually supposed to happen is that output is visible if the log level is DEBUG, and otherwise you get the spinner
[23:11:23] <njs> I didn't understand this correctly at all when I implemented the spinner stuff so the logic that's in there right now doesn't make any sense
[23:13:16] <xafer> Well https://github.com/pypa/pip/blob/develop/pip/utils/__init__.py#L665-L666 with show_stdout=True will show the output if running with --verbose
[23:16:41] <njs> also the logic that shows the output iff the command fails should probably not show the output a second tiem if it's already been shown
[23:18:06] <njs> ideally the show_output logic should be removed from this function altogether I think :-/ and the use of the logging module replaced with some bespoke output routines that are tailored to do what we actually want
[23:24:46] <njs> also: https://git.io/v2UOq is ready to merge. would be interested to know if there's anything we can do to get that done and get an 8.1 release with it out as soon as possible, because we are extremely eager for it to become real :-)
[23:26:39] <xafer> njs but if you remove the show_stdout, what do you show when the command is executing ?
[23:27:38] <njs> xafer: what I mean is that the current show_stdout= argument, which is only used by a few obscure places, should not be here at all, because this function is trying to serve too many masters
[23:28:06] <njs> and those few obscure places should be adapted to do something else (maybe use the same process spinner logic as everything else, or just use subprocess directly, or I dunno)