[21:53] <lifeless> mwhudson: hey
[21:53] <lifeless> lp:~lifeless/launchpad/test
[21:53] <lifeless> has some more work in it just fixing ec2 detected failures.
[21:53] <lifeless> more stale-librarian processes mainly.
[21:54] <lifeless> I'm sending it in to find more, but would like to know if the recent commits are clear to you
[21:54] <mwhudson> lifeless: wow, branch of doom
[21:54] <lifeless> mwhudson: I knew this would uncover a pile of mess.
[21:54] <lifeless> and it is
[21:55] <lifeless> its very clear to me why we were ending up with 4-5 librarians running mid-test-suite-run
[21:55] <lifeless> :)
[21:55] <mwhudson> yeah
[21:56] <mwhudson> lifeless: the changes look basically fine
[21:57] <mwhudson> lifeless: does LaunchpadScriptLayer.tearDown need to be conditional like that?
[21:59] <lifeless> well
[21:59] <lifeless> unregisterProvider returns False if it did not unregister.
[21:59] <lifeless> it might not unregister because:
[21:59] <lifeless>  - its not registered
[21:59] <lifeless>  - something prevents it being unregistered
[22:00] <lifeless> the former might happen if setUp wasn't called/blew up
[22:00] <lifeless> or if the base object got replaced with an empty registry.
[22:00] <lifeless> or the mailbox object got replaced.
[22:10] <mwhudson> lifeless: i wonder if noisly exploding would be the thing to do there?
[22:11] <mwhudson> (i'm not trying to ask leading questions, just encourage discussion/thinking)
[22:15] <lifeless> eventually, perhaps
[22:16] <mwhudson> lifeless: i guess it's not for this branch, but "ls --versioned --recursive --null --kind file | xargs -0 fgrep 'Setup().setUp()'" show more work to be done at some point :-)
[22:16] <mwhudson> lifeless: ok, i'm fine with the branch as-is, i hope there's not too much more to go before landing :-)
[22:16] <lifeless> mwhudson: yeah
[22:16] <lifeless> mwhudson: me too :P
[22:16] <lifeless> I'm only doing this cause I want parallel testing
[22:18] <lifeless> mwhudson: so we could raise LayerInvarient
[22:18] <lifeless> (sp?)
[22:19] <mwhudson> yeah, something like that
[22:19] <lifeless> I'd like to work on NotImplemented as a thing separately
[22:19] <mwhudson> (sorry, brain not working today)
[22:19] <lifeless> firstly I want to make it stateful
[22:19] <lifeless> then unique instances
[22:20] <lifeless> then parallel
[22:20] <mwhudson> lifeless: 'it' ?
[22:20] <lifeless> then cleanup
[22:20] <lifeless> the test environment
[22:20] <mwhudson> ah ok
[22:22] <mwhudson> lifeless: a problem that i've not yet had any good ideas about is the problem of hard coded paths/ports/etc in configs for tests that launch subprocesses
[22:22] <lifeless> mwhudson: See my thread where I describe how to fix that. I'd love a critical eye. So far its been 'mgmt' saying 'wow lovely' :P
[22:23] <mwhudson> lifeless: subject?
[22:23] <mwhudson> i have a feeling i only skimmed that mail
[22:23] <lifeless> stories for ...
[22:24] <mwhudson> ah right
[23:40] <mwhudson> lifeless: ok i replied to the mail
[23:40] <mwhudson> lifeless: slightly incoherently, sorry about that