Log file Viewer

[00:36:50] <lifeless> righto

[00:36:54] <lifeless> marginally functional again

[00:37:02] <lifeless> dstufft: so, saw a bunch of stuff in 7.0.1

[00:37:37] <lifeless> dstufft: I presume someone hit a fs limit with the cache? That surprised me turning up so quickly :)

[00:51:05] <dstufft> lifeless: hm, don't think so? I don't think 7.0.1 had any fs stuff

[00:51:21] <lifeless> dstufft: you changed the cache path algorithm

[00:51:39] <lifeless> dstufft: the commit says its was a file system limit that drove it

[00:54:52] <dstufft> lifeless: oh, I changed that before 7.0, mostly to use the the hash(full url + "#<hashtype>=<hash>") as the key instead of the filename. Thought about it for awhile and I decided that the risk of people using the same version to mean two different things was big enough compared to the relatively small reward of sharing a cache between two mirrors that have different URLs wasn't worth it

[01:00:11] <lifeless> dstufft: ah. So - FWIW I think that means that e.g. openstack CI will have vey noisy benefits

[01:00:46] <lifeless> dstufft: because they have a local mirror

[01:01:23] <lifeless> dstufft: I would have suggested hashing the artifact as a way to defend against that without giving up on the benfits

[01:01:37] <dstufft> lifeless: we don't have the artifact without downloading it

[01:01:57] <dstufft> lifeless: oh, so my plan btw to handle that is I have some plans to enable better support for multi repositories, and provide a way to override the cache key for that repository

[01:02:11] <dstufft> if you don't say an explicit cache key, it'll use the base url

[01:02:23] <lifeless> dstufft: yeah, but we have the artifact in the cache

[01:02:24] <dstufft> if you do, it'll use that explicit cache key instead of the base url

[01:02:32] <lifeless> dstufft: (or we couldn't have built the cached wheel in the first place)

[01:03:03] <lifeless> dstufft: [and on the second url, we'd only download the artifact, not compile it].

[01:03:08] <lifeless> anyhoo

[01:03:17] <lifeless> dstufft: more important - the setuptools/distribute thunk.

[01:03:30] <lifeless> dstufft: if you want to put it back, I'm entirely happy to preserve it in the resolver branch

[01:03:53] <dstufft> lifeless: I think barry warsaw is going to try and backport a newer virtualenv to wheezy

[01:04:08] <dstufft> so I think it's OK to just forget about it

[01:04:14] <lifeless> dstufft: there are older RHELs around too, but your call.

[01:04:23] <dstufft> I have a COPR for them

[01:04:24] <lifeless> just saying - as the person likely to be most impacted - I'm fine if you want it back

[01:04:29] <dstufft> which I need to update

[01:05:00] <dstufft> I tihnk it's OK, we've only had 1 person notice so far

[01:06:53] <lifeless> k

[01:07:02] <lifeless> well, I'm fading, catch you tomorrowish

[01:07:41] <dstufft> lifeless: see ya!

[22:44:25] <pmxbot> jaraco pushed 4 commits to setuptools (https://bitbucket.org/pypa/setuptools/) :

[22:44:25] <pmxbot> extract variable for nicer indentation

[22:44:26] <pmxbot> Update docstring to include the actual entry point names (for hits when grepping).

[22:44:26] <pmxbot> Disallow path separators in script names. Fixes #390

[22:44:26] <pmxbot> Update changelog