Filed by: wkocher [at] mozilla.com https://treeherder.mozilla.org/logviewer.html#?job_id=114498497&repo=autoland https://queue.taskcluster.net/v1/task/V7ciuC7DTtWTg6_paiGy4g/runs/0/artifacts/public/logs/live_backing.log https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/V7ciuC7DTtWTg6_paiGy4g/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1
This appears to have started with the landing of bug 1380025. The test varies, but the assertion is always Assertion failure: false (Ran out of memory while building cycle collector graph), at nsCycleCollector.cpp:930. (There's another frequent failure in these Win7debug Reftest-4 jobs with assertions about `Failed to create DrawTarget`, but those appear to be unrelated to this failure.
:jonco, as the author of the patch which triggered this failure please look into this- the longer we go with failures the less likely we are to fix them!
Hello Steve, do you have any thought about this failure?
No new failures reported since July 21.
Whiteboard: [stockwell needswork] → [stockwell unknown]
Whiteboard: [stockwell unknown] → [stockwell unknown], gfx-noted
I don't understand why bug 1380025 might have caused this but it seems to have stopped failing.
Huh. Yeah, I don't see how bug 1380025 would do this either, though it does require going through the logic for the gcPoke removal to see that, so maybe we missed something there. On the other hand, if we were to fix this OOM, it would probably be by adding an additional more robust mechanism. But this seems low enough volume right now to now bother with.
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.