Closed Bug 1393934 Opened 7 years ago Closed 1 month ago

Intermittent Assertion failure: false (Ran out of memory while building cycle collector graph), at z:/build/build/src/xpcom/base/nsCycleCollector.cpp:959

Categories

(Core :: Cycle Collector, defect, P5)

defect

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: intermittent-bug-filer, Unassigned)

References

(Depends on 1 open bug)

Details

(Keywords: intermittent-failure, Whiteboard: [stockwell disabled])

Component: JavaScript: GC → XPCOM
ack, 37 failures in the last 7 days: https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1393934 this looks to be similar to bug 1347625 and the failure pattern is the same win7/debug non-e10s reftests.
Depends on: 1347625
Whiteboard: [stockwell needswork]
(In reply to Joel Maher ( :jmaher) (UTC-5) from comment #2) > this looks to be similar to bug 1347625 and the failure pattern is the same > win7/debug non-e10s reftests. Now see bug 1373578 for the latest discussion.
Depends on: 1373578
in bug 1373578 we are looking to turn off non-e10s reftests which would remove this failure
See Also: → 1373578
:froydnj, I see you as the triage owner here, can you take a look? Recent failures since going to 32 chunks seem to fail consistently on chunk 31 (492 tests in total), and in this chunk we seem to fail in the transform-3d manifest (near the end of the chunk)
Flags: needinfo?(nfroyd)
Andrew has more experience with the CC graph. But this *really* shouldn't be happening on tests...is there more logging or something that we can be adding here?
Flags: needinfo?(nfroyd) → needinfo?(continuation)
(In reply to Nathan Froyd [:froydnj] from comment #14) > Andrew has more experience with the CC graph. But this *really* shouldn't > be happening on tests...is there more logging or something that we can be > adding here? Win32 reftests have been an ongoing problem. If you have ever watched them run, they open and close pages extremely rapidly, which is hard for our GC/CC heuristics to deal with, so windows pile up. I'm tempted to blame bug 1377131, but that landed on the 6th, but this seems to have started happening a lot on the 4th or 5th. It looks like it got better around the 8th or 9th, so maybe bug 1368972 improved this.
Flags: needinfo?(continuation)
bug 1401228 has a disabled test, we shouldn't see more failures related to this bug.
Keywords: leave-open
Whiteboard: [stockwell disable-recommended] → [stockwell disabled]
Depends on: 1424349
Severity: normal → S3
Component: XPCOM → Cycle Collector
Status: NEW → RESOLVED
Closed: 1 month ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.