Bug 44928
| Summary: | commit-queue gets stuck if rollout fails and tree is red | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Eric Seidel (no email) <eric> |
| Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | Normal | CC: | mihaip |
| Priority: | P2 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | PC | ||
| OS: | OS X 10.5 | ||
| Bug Depends on: | |||
| Bug Blocks: | 44924 | ||
Eric Seidel (no email)
commit-queue gets stuck if rollout fails and tree is red
Recent Status
Patch 64903 from bug 44196: Landing patch 5 minutes ago
Unable to successfully build and test [results] 5 minutes ago
Patch 66003 from bug 44924: Landing rollout patch 17 minutes ago
Unable to successfully build and test [results] 17 minutes ago
Patch 66003 from bug 44924: Landing rollout patch 26 minutes ago
Unable to successfully build and test [results]
failure message:
http://webkit-commit-queue.appspot.com/results/3826166
I believe this is what's happening:
Normally the cq tries 3 times when landing a patch:
1. applies patch, tries to build, run tests, if successful lands, stops processing.
2. builds and tests w/o patch applied. if fails, assumes tree is busted, stops processing, sleeps 2 minutes.
3. applies patch, tries to build, run tests, if succeeds, lands. if fails, rejects patch.
this allows us to deal with flaky tests better.
I believe we already have a bug about svn-apply failures going through these steps when instad they should fail out early. Build errors should also fail early since they're never flaky. That would solve the case seen today where the apply failed.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Mihai Parparita
I think bug 37553 is about svn-apply failing.
Eric Seidel (no email)
Yup, this will basically end up being a duplicate of bug 37553. We just need bug 37553 to handle the build case too.
Eric Seidel (no email)
*** This bug has been marked as a duplicate of bug 37553 ***