Bug 31422 - Implement a try-queue
Summary: Implement a try-queue
Status: NEW
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Nobody
Depends on: 31513 31515
  Show dependency treegraph
Reported: 2009-11-12 10:59 PST by Adam Barth
Modified: 2013-04-17 22:23 PDT (History)
2 users (show)

See Also:

work-in-progress (19.00 KB, patch)
2009-11-12 11:02 PST, Adam Barth
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Adam Barth 2009-11-12 10:59:42 PST
It would be nice to be able to try patches via a bot.  Patch forthcoming.
Comment 1 Adam Barth 2009-11-12 11:02:49 PST
Created attachment 43078 [details]
Comment 2 Adam Barth 2009-11-12 11:05:06 PST
@Eric: Thoughts?
Comment 3 Brian Weinstein 2009-11-12 12:50:46 PST
This is really interesting. I had started looking at a Trybot implementation using something more like the Chromium try_change.py, but this seems like a really good way to make it work well with the Bugzilla workflow, and would be straightforward to restrict to committers or who we wanted to be able to use it.
Comment 4 Eric Seidel (no email) 2009-11-12 12:59:40 PST
I think BuildBot is going to be better designed for this.  I've even considered re-writing commit-queue on top of buildbot, because it would allow the commit to depend on the results of multiple bots.  This could be an interesting interim solution though.

I guess for this we'd have N bugzilla flags for the N "try-queues"?


And the "try bots" would run on those various platforms and set them to + on positive completion, or - on failure?

Or were you thinking we'd run this off of a single flag?  I guess we could run this off of the commit-queue=+ flag and set individual flags to mean that it passed the various platforms?
Comment 5 Eric Seidel (no email) 2009-11-12 13:23:57 PST
It would of course be possible to use comments instead of flags.  A flag to add it to the "try queue" and then comments to record success/failure from the various bots.  I'm not sure that would make things any less confusing, but it would be possible.  The comments would be "authenticated" in that they would be expected to come from certain users, and would contain which revision was used for testing, when it completed, sucessfully, etc.  We could also store this data outside of bugzilla in an AppEngine instance like how http://webkit-commit-queue.appspot.com/ works.
Comment 6 Adam Barth 2009-11-12 14:12:39 PST
> I guess for this we'd have N bugzilla flags for the N "try-queues"?

I'm inclined to solve the problem for N=1 first and worry about N=2 later.
Comment 7 Adam Barth 2009-11-14 23:21:44 PST
Comment on attachment 43078 [details]

I'm doing this in small pieces in dependent bugs instead of here.