Bug 28222 - bugzilla-tool needs a way to ask build.webkit.org if the bots are passing
Summary: bugzilla-tool needs a way to ask build.webkit.org if the bots are passing
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: 528+ (Nightly build)
Hardware: PC OS X 10.5
: P2 Normal
Assignee: Nobody
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-08-12 11:36 PDT by Eric Seidel (no email)
Modified: 2009-08-19 17:13 PDT (History)
2 users (show)

See Also:


Attachments
Patch v1 (13.68 KB, patch)
2009-08-17 16:13 PDT, Eric Seidel (no email)
no flags Details | Formatted Diff | Diff
Updated patch which actually passed the unit tests (13.57 KB, patch)
2009-08-17 16:18 PDT, Eric Seidel (no email)
no flags Details | Formatted Diff | Diff
Patch v1 (16.71 KB, patch)
2009-08-19 16:41 PDT, Eric Seidel (no email)
abarth: review+
abarth: commit-queue+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Eric Seidel (no email) 2009-08-12 11:36:46 PDT
bugzilla-tool needs a way to ask build.webkit.org if the bots are passing

If they're not passing we should not run the --commit-queue, etc.

I'd rather not scrape the buildbot page, but we could.  I don't want to put too much extra load on the buildbot master.
Comment 1 Eric Seidel (no email) 2009-08-17 16:13:01 PDT
I have a patch which I'm testing now.  Will attach for others to see.
Comment 2 Eric Seidel (no email) 2009-08-17 16:13:19 PDT
Created attachment 34999 [details]
Patch v1
Comment 3 Eric Seidel (no email) 2009-08-17 16:18:12 PDT
Created attachment 35000 [details]
Updated patch which actually passed the unit tests
Comment 4 Eric Seidel (no email) 2009-08-17 16:21:31 PDT
I don't know much about buildbot's xmlrpc (or xmlrpc in general).  It might be possible to do all this and more via a few xmlrpc calls:
http://buildbot.net/apidocs/buildbot.status.web.xmlrpc.XMLRPCServer-class.html
Comment 5 Adam Barth 2009-08-17 17:25:12 PDT
This patch doesn't seem to call ensure_builders_are_green.  Did you mean to call that during land-diff and land-patches?
Comment 6 Eric Seidel (no email) 2009-08-17 17:26:23 PDT
Yes. :)  Will fix.
Comment 7 Eric Seidel (no email) 2009-08-17 22:48:28 PDT
Yay!  It's working!

Sleeping.... Mon Aug 17 22:33:06 PDT 2009
Fetching http://build.webkit.org/one_box_per_builder
ERROR: Builders at build.webkit.org are red, please do not commit.  Pass --ignore-builders to bypass this check.
Fetching http://build.webkit.org/one_box_per_builder
ERROR: Builders at build.webkit.org are red, please do not commit.  Pass --ignore-builders to bypass this check.
Sleeping.... Mon Aug 17 22:43:09 PDT 2009
Comment 8 Eric Seidel (no email) 2009-08-17 22:49:09 PDT
(The windows buildbots are currently red, thus the commit-queue is correctly pausing itself until they roll green again.  It wakes up every 10 minutes to check.)
Comment 9 Mark Rowe (bdash) 2009-08-17 22:53:44 PDT
Comment on attachment 35000 [details]
Updated patch which actually passed the unit tests

 47         # Other builders should be added to this list once they're known to be stable.
 48         self.core_builder_names_regexps = [ 'Leopard', "Windows.*Build" ]

Why is Tiger being left off?
Comment 10 Eric Seidel (no email) 2009-08-17 22:57:03 PDT
I just picked the bots which seemed the most consistent. If you'd like, I can add Tiger there too.  It seems that tiger breaks rather often, and we don't bother to "close" the tree when it does.  When Leopard or the Windows build breaks we seem to have fire-drills.  I wasn't really trying to make policy with this change, so much as teach the commit-queue about our existing policy.  I'm also happy to make the commit-queue block on all bots being green, but that's probably extreme.
Comment 11 Eric Seidel (no email) 2009-08-19 16:41:41 PDT
Created attachment 35159 [details]
Patch v1
Comment 12 Adam Barth 2009-08-19 16:50:23 PDT
Comment on attachment 35159 [details]
Patch v1

Great!  Mark, we can iterate on the list of builders if this list isn't sufficient.
Comment 13 Eric Seidel (no email) 2009-08-19 17:13:10 PDT
Committed r47532: <http://trac.webkit.org/changeset/47532>