Log file Viewer

[03:44:37] <lifeless> tchaypo: ok so

[03:45:33] <lifeless> looking in trello

[03:45:50] <lifeless> Nakato: https://github.com/pypa/pip/issues/2928 <- did we get a fix for that in a different # or something ?

[03:46:09] <Nakato> I'm chasing that up today.

[03:46:44] <lifeless> cool

[03:46:47] <lifeless> tchaypo: https://trello.com/c/0VCcCxsM/8-setuptools

[03:47:06] <Nakato> Lifeless: Need to respond to comment. What's your opinion on the comment from piotr in https://github.com/nakato/pip/commit/d266a5e1b9899a1a1e0a0b1868c258ca5db0ee4c ?

[03:47:30] <lifeless> tchaypo: seems to me that we should polish the setuptools marker support so that markers don't need the crazy indirect-via-extras thing

[03:49:33] <tchaypo> or switch from using tests-require to an extra named tests

[03:53:00] <lifeless> so thats not equivalent

[03:53:19] <lifeless> its a hack that works just for us today; part of the goal here was to do the ecosystem fix.

[03:53:45] <lifeless> e.g. having to workaround setuptools in pbr (we have a patch to strip markers from tests-require) is tech debt for us.

[03:53:59] <lifeless> likewise the mapping to crazy extras clauses is tech debt

[03:54:27] <lifeless> fixing that up will keep pbr lean and focused on value-add vs basic-functionality

[04:03:06] <lifeless> tchaypo: ^ ;0

[04:09:19] <tchaypo> specifically trying to figure out what env markers in install-requires look like if they’re entered by not-pbr

[04:12:26] <lifeless> tchaypo: right now they have to be entered as extras

[04:12:35] <lifeless> tchaypo: which is what you'd be fixing

[04:13:20] <tchaypo> Looking at the code

[04:13:42] <tchaypo> Setuptools/dist.py:check_extras is where they're looked for

[04:14:24] <tchaypo> Hence needing to be extras

[04:14:46] <lifeless> right

[04:15:01] <lifeless> from a backwards compat point of view

[04:15:45] <lifeless> that might be a necessary evil (and bodes poorly for tests_require, but from a ui perspective, setup(install_requires='argparse:python_version<"2.7"') is something we should deliver

[04:15:55] <tchaypo> If we tweak check_requirements we can have them supported in install_requires

[04:16:01] <tchaypo> Well

[04:16:21] <tchaypo> Install_requires=['...']

[04:16:41] <lifeless> we need to think at least a little about the impact of metadata in wheels landing on a system with an older pkg_resources

[04:16:52] <tchaypo> To get it working for tests_require though...

[04:17:00] <lifeless> tchaypo: install_requires can be a space delimited string or a list of strings

[04:17:11] <lifeless> tchaypo: also, lower case I :)

[04:21:05] <tchaypo> sigh.

[04:21:16] <tchaypo> is test_requires one of those things where it shells out to easy_install?

[04:23:56] <lifeless> setuptools doesn't shell out, it *is* easy_install

[04:24:23] <lifeless> tchaypo: tests_require btw, might help you find it in google

[04:24:28] <tchaypo> i hate these things

[04:24:32] <tchaypo> or in the code :)

[04:26:23] <tchaypo> so i see setuptools/commands/test.py:run passes the tests_require into self.distribution.fetch_build_eggs, which seems to end up calling into pkg_resources to get it downloaded and built

[04:27:14] <tchaypo> I’m not seeing how adding support for the markers on install-requires would help us with tests_require

[04:27:27] <lifeless> ok so lets step back

[04:27:41] <lifeless> there is tests_ setup_ install_ and extra_ requires

[04:27:56] <lifeless> of those four things, three don't yet support markers

[04:28:19] <lifeless> and the last supports it in an awkward fashion: you put a marker around the name of an extra, not in the reqs list.

[04:28:41] <lifeless> What I'm actually proposing is that setuptools be proffered a patch to do inline-with-requirements markers for all four of those things.

[04:29:04] <lifeless> -> picking up Cynthia

[04:29:28] <tchaypo> would a patch that modifies the first three to look for specially named extras satisfy you?

[04:29:53] <tchaypo> or to put it differently

[04:30:43] <tchaypo> it seems to me that it would make sense to have one set of resolving code for all of those cases. currently the most flexible resolving code is for extras, so it seems like the simplest thing to do might be to use extras for everything

[04:30:55] <lifeless> no :)

[04:31:01] <lifeless> because

[04:31:06] <lifeless> think about the UI

[04:31:30] <lifeless> setuptools is a UI

[04:32:02] <tchaypo> I have to go pick up my car as well. talk later?

[04:32:06] <lifeless> the implementation detail of 'look for all extras called ':THING' to find out what the actual install_requires are' is horrid

[04:32:09] <lifeless> yes

[22:23:31] <xafer> Hello, https://github.com/pypa/pip/pull/2995 looks ok to me. Maybe dstufft you could take a look some time, since you created the corresponding issue https://github.com/pypa/pip/issues/2643 ?

[22:34:02] <lifeless> tchaypo: ok so, you got your car ?

[22:36:07] <tchaypo> yes. I have an interview 9-10 and a meeting 10:30-11:30. Can possibly talk around them a bit but I don’t expect to be particularly responsive.

[22:47:33] <tchaypo> I have in my head a broad outline of the work that needs to happen, I think. when I’m free I’d like to run it past you

[23:08:33] <lifeless> tchaypo: sounds good; ping me when suits