[00:45] <bigjools> wgrant/StevenK: did the recipe builds change recently to append ~ubuntuNN instead of ~series?
[00:47] <wgrant> bigjools: Yes
[00:47] <wgrant> Ubuntu's used the version number for backports for a while, and we need to switch before u-series
[00:47] <bigjools> wgrant: ok.  you broke juju-core folks. haha :)
[00:47] <wgrant> Why are they depending on the version number?
[00:47] <wgrant> s/version number/structure of the version number/
[00:47] <wgrant> That's pretty broken
[00:47] <bigjools> stupidity I think :)
[00:47] <wgrant> Still
[00:48] <wgrant> That recipe is already pretty broken
[00:48] <wgrant> I'm going to close that tab and pretend I never saw it.
[00:48] <bigjools> heh
[04:00] <StevenK> wgrant: The next branches for buildd-manager are due RSN?
[04:00] <wgrant> Indeed
[07:44] <wgrant> StevenK: https://code.launchpad.net/~wgrant/launchpad/buildd-manager-inlineCallbacks/+merge/183578
[07:55] <StevenK> wgrant: I'm wary of the the commit in the manager code
[07:57] <wgrant> StevenK: It's committing after it comes back from a Deferred
[07:57] <wgrant> Which doesn't make much sense.
[07:58] <wgrant> Though I'm surprised you noticed that in the merged diff.
[07:59] <StevenK> It's five lines into a green diff
[07:59] <wgrant> In fact, you couldn't have noticed. Which commit are you talking bout?
[07:59] <StevenK> 455	+ # Commit the changes done while possibly rescuing jobs, to
[07:59] <StevenK> 456	+ # avoid holding table locks.
[07:59] <StevenK> 457	+ transaction.commit()
[07:59] <wgrant> That's not new
[08:00] <wgrant> And it shouldn't do anything, but it's there to ensure that builder.currentjob is as up to date as possible.
[08:00] <StevenK> Oh, it's just moved from other twisted garbage
[08:00] <wgrant> I didn't change any code except for removing two xmlrpc.Fault wrappers.
[08:00] <StevenK> And I guess that is one of the commits that is not long for this world?
[08:01] <wgrant> Correct.
[08:01] <wgrant> We can't afford to calculate currentjob there.
[08:01] <wgrant> It will be precalculated, and that method won't touch the DB until further down.
[08:02] <StevenK> So thanks to a day of JavaScript and Twisted, my brain is now leaking out of my ears
[08:05] <StevenK> wgrant: r=me
[08:05] <StevenK> wgrant: Sorry, I got distracted
[08:07] <wgrant> Thanks
[09:01] <cjwatson> wgrant: Your buildd-manager-inlineCallbacks branch and my destroy-builder-aborted branch clash.  Which would you prefer to have landed first?
[09:06] <wgrant> cjwatson: I was deliberately waiting for yours to land, sorry.
[09:07] <cjwatson> OK, landing
[09:08] <wgrant> During the refactoring I discvoered something that I thought was just a production glitch a few months ago.
[09:08] <wgrant> A manualled, disabled builder won't get its current build terminated.
[09:09] <wgrant> Probably missed originally because of the convoluted callbacks. It makes no sense otherwise.
[09:16] <wgrant> cjwatson: test_binarypackagebuildbehavior still references AbortedSlave
[09:18] <cjwatson> what.  I grepped.
[09:19] <cjwatson> one moment :(
[09:19] <cjwatson> Oh, I must have only grepped buildmaster, damn
[09:19] <wgrant> Ah, yeah
[09:21] <cjwatson> Very confused by what buildbot just did
[09:21] <cjwatson> It successfully ran no tests
[09:22] <wgrant> It doesn't check the exit code due to an old lxc-start-ephemeral bug
[09:22] <wgrant> The testrunner probably died badly due to the import.
[09:29] <cjwatson> wgrant: https://code.launchpad.net/~cjwatson/launchpad/testfix-destroy-builder-aborted/+merge/183598
[09:33] <wgrant> cjwatson: r=me
[09:34] <cjwatson> landing
[09:57] <stub> Anyone up for a Swift branch review? More /srv and /srv2 juggling underway it seems.
[10:02] <wgrant> I really should get to that eventually
[10:02] <wgrant> But the /srv and /srv2 juggling can't stop for at least a few months after we start using Swift