[13:01] <cjwatson> wgrant: Did you have a chance to look at my series-alias branch while I was on holiday?
[21:36] <cjwatson> wgrant: I'm trying to sort out the Deferred handling in https://code.launchpad.net/~cjwatson/launchpad/buildmaster-cancel-properly/+merge/177580.  Why would requestAbort fail if the slave is already ABORTING?  My reading of the slave code is that it should just return immediately.
[21:37] <cjwatson> wgrant: My inclination is that if requestAbort fails then I should immediately enter the resume-slave-host-or-fail-builder path.  But I think I can only take that approach if such a failure isn't going to happen any time somebody restarts buildd-manager in the middle of a cancellation.
[21:49] <cjwatson> I believe that http://paste.ubuntu.com/6034419/ implements this if it's a tenable approach.
[23:28] <wgrant> cjwatson: Ah, you're right, abort() shouldn't normally fail.
[23:29] <wgrant> resetting is indeed probably sensible, but the CancellationError due to a manager restart is indeed an issue.