[09:04:30] <zahlman> I tried to `pip install pywin32`, and received the error `Could not find any downloads that satisfy the requirement pywin32`. Is that simply because my version of python isn't supported?
[09:09:14] <mgedmin> older versions of setuptools would crawl all pages reachable from the pypi index looking for links that look like maybe they point to downloadable archives
[09:09:20] <mgedmin> newer versions of setuptools/pip stopped doing that
[09:09:30] <mgedmin> there was a transition and everything
[09:09:48] <mgedmin> but of course nobody updates dead projects :/
[09:30:41] <zahlman> excellent! thanks for the search-fu :)
[09:31:46] <zahlman> too bad that `pip install pywin32` doesn't find pypiwin32.
[09:35:05] <mgedmin> hmm, glyph is a pypi maintainer for both projects; he could upload the wheels into pywin32
[09:37:05] <zahlman> ...ahaha. pip claims it installed successfully (into my virtualenv, even, which apparently isn't supposed to be possible http://sourceforge.net/p/pywin32/bugs/669/)
[09:37:19] <zahlman> but it reported an error in 'compiling' the python scripts
[09:37:40] <zahlman> and I was rather amused by what tripped it up.
[09:41:17] <mgedmin> you're basically doing archaeology there
[09:41:27] <mgedmin> (btw script compiling errors during pip install are harmless)
[09:42:00] <zahlman> I figured. well, it still means I can't import win32comext.mapi.mapisend and expect it to work as is ;)
[09:42:59] <zahlman> I guess it would have worked to use `globals()` instead, since then the value would be found in global lookup on the old versions, and ignored on 3.x+ (and also not be invalid syntax)
[09:43:17] <zahlman> but you would have had to have been pretty psychic to anticipate that in 2.3-ish days :)
[09:45:24] <zahlman> (since it's in a venv anyway, I'm going to just delete those lines. although I suspect the installer modified my registry quite a bit anyway, which would make it not such a clean virtualization)
[12:03:16] <lumidee> hi, how do i write 'console_scripts': [ 'gymail-cli = gymail: argparse_entrypoint',] if my function argparse_entrypoint is in the file /gymail/gymail.py? /gymail is the top package directory, there's non above. i get this error message: ImportError: module 'gymail' has no attribute 'argparse_entrypoint'. strangely the cli command still works though this error appears. i also tried 'console_scripts': [ 'gymail-cli = gymail.gymail:
[14:01:35] <jml> If I have a project that needs --process-dependency-links to work (temporarily using a fork of a dependency while we wait for fixes to be merged), is there any way I can configure pip so that `pip install -e .[dev]` automatically uses --process-dependency-links?
[14:03:11] <mgedmin> something like export PIP_PROCESS_DEPENDENCY_LINKS=true?
[14:03:32] <mgedmin> maybe a config file in ~/.config/pip?
[14:03:48] <mgedmin> iirc every pip option can be set three ways: on the command line, via environtment, via config file
[14:16:08] <jml> but I think I'd still have to convince everyone in our dev team to do that.
[14:17:49] <mgedmin> do your devs do `pip install -e .[dev]` by hand?
[14:17:56] <mgedmin> maybe create a Makefile to automate that
[14:18:06] <mgedmin> have the makefile do `pip install --process-dependency-links -e .[dev]`
[14:18:12] <ronny> jml: do you have a own devpi instance, at my workplace we upload local versions to a devpi instance, pin those and require all devs to use the devpi
[14:24:29] <jml> ronny: yeah, we do have that, but we don't yet require all devs to do anything like that.
[14:24:53] <jml> sounds like no matter what I do, it'll involve a change that needs to be communicated to all devs.
[14:27:21] <mgedmin> buildout is nice for situations like this: you add a '[buildout] find-links = https://our.intranet/forked-packages/' to buildout.cfg and everyone will get the right bits next time they run 'make' (or bin/buildout)
[14:27:27] <mgedmin> shame about buildout not supporting wheels
[14:40:48] <a-tal> jml: you can put —process-dependency-links in your requirements.txt
[14:47:46] <ronny> mgedmin: buildout is a mess tho (it tries to do too much at once)
[14:48:06] <mgedmin> buildout is terribly undermaintained
[14:51:52] <ronny> there are by no means the human ressources behind it to do whatit set out to do
[20:45:59] <ronny> i managed to mess up by trying with git instead of hg
[20:46:30] <ronny> jaraco: btw, is it reasonable to do pr#s agaisnt your setuptools repo on gh?
[20:47:12] <jaraco> ronny: it’s acceptable, yes. In some ways it’s nicer (travis checks).
[20:48:01] <ronny> jaraco: im hoping i can finish m bb > gh importer bz the weekend (username mapping based on a actual human checked maping is missing)
[20:52:03] <ronny> jaraco: i udnerstood the bug, there is a hg template i should be giving with a newline and expect more than one line, i can make test
[21:07:47] <pmxbot> #pypa was once on Wheel of Fortune and was the first to spin. The next 29 minutes of the show consisted of everyone standing around awkwardly, waiting for the wheel to stop.
[21:07:50] <jaraco> Part of the ‘motivation’ package.