[06:49] wgrant: ta [06:53] * stub wonders how the .tac worked at all without os.path imported [06:56] stub: os.path is magical [06:57] stub: It's an alias for ${PLATFORM}path, which os/__init__.py determines and imports dynamically [06:57] wgrant: I've got the explicit wait-until-dead check, as I had a test confirming behavior when Swift is dead failing without it. I'm happy to move it to the base class if you want. [06:57] Well, except that I guess os is an extension, but same idea [06:58] stub: Hm, it seems odd that tearDown doesn't want until it's dead. [06:58] s/want/wait/ [06:58] I prefer the 'magical' explanation [06:58] Shutting down is normally a layer tear down thing - you don't usually care that it takes a little while to actually die. [06:59] Except that some of our layers use the same port each time [06:59] (or at least used to) [06:59] So things clearly shut down in relatively short order. [06:59] I guess the shutdown time < startup time :) [06:59] Yeah [09:03] stub: Oh, the diff in the MP must have been out of date [09:04] setup.py showed like 12 new deps, then your new commit removed one and now it only shows three new ones. [09:04] yeah, it got stuck at some point. [09:05] Unclogged now (sourcecode suppository?) [09:05] That makes a lot more sense, because I remember bringing that up a few weeks ago and you fixed it [09:07] So I should land that now? And do we still shuffle things through ec2test nowadays? [09:07] * stub hasn't inflicted reimbursement forms on his new management yet. [09:50] stub: We don't push stuff through ec2 very often at all nowadays, unless it has potential for widespread failure that might take a while to fix. [09:50] stub: mock-swift has no widespread impact, but I'd send the second one through ec2. [09:51] For small failures it's much quicker to get them from buildbot and then [testfix]. === marcoceppi_ is now known as marcoceppi === Ursinha is now known as Ursinha-afk === Ursinha-afk is now known as Ursinha