PMXBOT Log file Viewer

Help | Karma | Search:

#pypa-dev logs for Friday the 9th of October, 2015

(Back to #pypa-dev overview) (Back to channel listing) (Animate logs)
[01:02:50] <lifeless> dstufft: so
[01:02:59] <lifeless> dstufft: would you be open to a pip thing like I raise?
[01:04:50] <dstufft> lifeless: how do you plan on determining if something is distro installed?
[01:07:29] <lifeless> dstufft: no idea :)
[01:07:53] <lifeless> dstufft: perhaps as crude as looking for the missing vendored urllib3
[01:08:08] <lifeless> dstufft: we could make it datadriven etc
[01:09:14] <dstufft> lifeless: so you want this to be specific to requests/urllib3 and not a generalized "x vendors y, reinstall x if you're installing y and x is from a distro?"
[01:14:19] <lifeless> dstufft: well, there's a limited number of vendoring cases
[01:14:41] <lifeless> dstufft: but that could work too
[01:14:50] <lifeless> dstufft: just refuse to consume anything in dist-packages
[01:16:51] <lifeless> dstufft: that might unfairly penalise numpy users though
[01:16:59] <lifeless> dstufft: (and other expensive-compile things)
[01:17:04] <dstufft> it's also debuntu specific
[01:17:13] <dstufft> other distros don't have a dist-packages
[01:17:20] <lifeless> Fedora sounds like they're fixing the situation anyhow
[01:17:38] <lifeless> do arch unbundle? Like, how far does the philosophical rot go?
[01:17:46] <dstufft> um
[01:17:51] <dstufft> I don't know what arch does tbh
[01:17:57] <dstufft> nobody has ever yelled at me about arch
[01:18:50] <dstufft> lifeless: I'm going to need to think about this some
[01:18:58] <dstufft> fixing this in pip feels really dirty
[01:19:02] <lifeless> dstufft: I agree
[01:19:04] <dstufft> but it might be the only effective thing :/
[01:19:15] <lifeless> dstufft: indeed
[01:19:22] <lifeless> dstufft: there's no panic over it
[01:19:34] <lifeless> dstufft: but its really sapping folks will to live a small slice at a time
[01:23:49] <dstufft> I might press my luck and see if I can convince requests to unbundle
[01:27:14] <lifeless> we'll still have to fix pip for that
[01:27:24] <lifeless> to make it honour their pins in all cases
[01:28:30] <dstufft> I'm not sure I understand why requests is special in that case
[01:31:29] <lifeless> its not, its just that most folk don't hard-pin
[01:31:47] <lifeless> but if they unbundle and want the same control over things they'll end up with a == pin on the urllib3 they work with
[01:32:14] <dstufft> I wonder if they even really need the ==
[01:32:24] <dstufft> urllib3 is pretty backwards compatible.
[01:32:40] <dstufft> I think the biggest thing is they just aggressively rely on new features in urllib3
[06:30:57] <Ivo> holy ****** Re: [Distutils] Towards a simple and standard sdist format that isn't intertwined with distutils is goddamn massive
[06:36:22] <Ivo> lifeless, arch has --without-ensurepip atm
[07:16:04] <lifeless> Ivo: thats unrelated to the requests issue
[07:34:11] <Ivo> sure, just wanted to mention
[19:42:59] <dstufft> lifeless: lol openstack-dev rejected me
[19:43:54] <dstufft> lifeless: anyways, I know about the resolver things, but I think we can get them fixed before Jessie+1 which is when requests would be upgraded to 2.7+ in Debian anyways
[23:49:04] <lifeless> dstufft: yeah, its in progress