Closed
Bug 614946
Opened 14 years ago
Closed 14 years ago
Unimportant l10n changes still get run eventually
Categories
(Release Engineering :: General, defect)
Release Engineering
General
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: catlee, Assigned: dustin)
References
()
Details
(Whiteboard: [l10n][buildbot])
Attachments
(2 files)
2.58 KB,
patch
|
catlee
:
review+
bhearsum
:
checked-in+
|
Details | Diff | Splinter Review |
1.07 KB,
patch
|
bhearsum
:
review+
bhearsum
:
checked-in+
|
Details | Diff | Splinter Review |
We have a fileIsImportant function for l10n changes (isImportantL10nFile) that restricts triggering of dependent builds to changes that touch certain modules. The problem is that the buildbot schedule doesn't throw away unimportant changes, it just keeps them for later. The first important change triggers builds for all accumulated unimportant changes since we have treeStableTimer=None. This could be considered an upstream bug, but a quick fix would be to write a custom scheduler that completely ignores unimportant changes.
Assignee | ||
Updated•14 years ago
|
Assignee: nobody → dustin
Assignee | ||
Updated•14 years ago
|
Assignee | ||
Comment 1•14 years ago
|
||
Attachment #494308 -
Flags: review?(catlee)
Assignee | ||
Comment 2•14 years ago
|
||
I think we'll need to upgrade the scheduler and release masters and restart them, right?
Reporter | ||
Comment 3•14 years ago
|
||
Comment on attachment 494308 [details] [diff] [review] 8cd63b53b93749fc487c.patch It might be worth going through buildbotcustom/scheduler.py to see if this will break anything there, or if we need to make similar changes to custom schedulers.
Attachment #494308 -
Flags: review?(catlee) → review+
Assignee | ||
Comment 4•14 years ago
|
||
Yes, it looks like scheduler.py needs similar changes. I'll put a patch up when I get a chance.
Assignee | ||
Comment 5•14 years ago
|
||
ben is staging both of these patches as well as the patch from bug 615559
Attachment #494756 -
Flags: review?(bhearsum)
Comment 6•14 years ago
|
||
I tested this in staging by applying both of the patches and doing the following; - resetting the db - pushing two changes to my user repository, which the poller picked up - sendchanging to start the staging release I ended up with a dummy repo_setup and tag run that only included the sendchange'd Change, which is what we want. When I tried the same experiment without the patches from here I got 3 pending builds when i sendchange'd, which were merged together randomly. One additional thing I noticed is that I got mail from the ChangeNotifier...which is not quite what we want. It seems to happen regardless of whether or not the patches from here are applied though, so I'll file it separately. This is ready to land as far as I'm concerned.
Flags: needs-treeclosure?
Updated•14 years ago
|
Attachment #494756 -
Flags: review?(bhearsum) → review+
Comment 7•14 years ago
|
||
Comment on attachment 494756 [details] [diff] [review] 614946-buildbotcustom-r1.patch changeset: 1251:a603fd12bf7c Landed on the default branch, will be merged into production and on the Buildbot masters tomorrow morning.
Attachment #494756 -
Flags: checked-in+
Comment 8•14 years ago
|
||
Comment on attachment 494308 [details] [diff] [review] 8cd63b53b93749fc487c.patch changeset: 111:e7d90277a2a9 Landed on the default branch, will be merged into production and on the Buildbot masters tomorrow morning.
Attachment #494308 -
Flags: checked-in+
Comment 9•14 years ago
|
||
Will be landing this in a rolling downtime tomorrow -- aka no tree closure needed....needs-reconfig+ is the best analogue.
Flags: needs-treeclosure? → needs-reconfig+
Comment 10•14 years ago
|
||
Should be all done here now.
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Updated•11 years ago
|
Product: mozilla.org → Release Engineering
You need to log in
before you can comment on or make changes to this bug.
Description
•