Bug 1548827 Comment 14 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

(In reply to Jon Coppeard (:jonco) from comment #9)
> Oh, I meant to talk to you about this before.  I saw that there was a large build up of live windows when running the JS reftests and I looked into whether we were failing to trigger GC.  I found that GC was running but was not collecting anything.  I tried putting in extra CCs but this didn't help.  I came to the conclusion that we are somehow leaking all these windows until the end of the test.  This could explain why we are getting these OOMs.

It is possible that the test harness is leaking windows. Like maybe the main page of the test test opens iframes, and somehow keeps references to them. There was an issue like this in some other test framework a while ago. To investigate this, you'd need to do something like get GC/CC logs from while the test is running, and see what is holding alive a window that you'd have expected to be cleaned up already.
(In reply to Jon Coppeard (:jonco) from comment #9)
> Oh, I meant to talk to you about this before.  I saw that there was a large build up of live windows when running the JS reftests and I looked into whether we were failing to trigger GC.  I found that GC was running but was not collecting anything.  I tried putting in extra CCs but this didn't help.  I came to the conclusion that we are somehow leaking all these windows until the end of the test.  This could explain why we are getting these OOMs.

It is possible that the test harness is leaking windows. Like maybe the main page of the test opens iframes, and somehow keeps references to them. There was an issue like this in some other test framework a while ago. To investigate this, you'd need to do something like get GC/CC logs from while the test is running, and see what is holding alive a window that you'd have expected to be cleaned up already.

Back to Bug 1548827 Comment 14