1870477 - Trigger GC soon after CC if CC found lots of GCed garbage

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, enhancement)

Description

[Olli Pettay [:smaug][[e-mail]]]

Currently it may take too much time to collect GCed objects after CC

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

Attachment: Bug 1870477 - Trigger GC soon after CC if CC found lots of GCed garbage, r=mccr8

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1b4285ba4f41
Trigger GC soon after CC if CC found lots of GCed garbage, r=mccr8

Comment 3

[Atila Butkovits]

https://hg.mozilla.org/mozilla-central/rev/1b4285ba4f41

Comment 4

[Byron Campen [:bwc]]

This seems to have dramatically increased the rates of intermittent timeouts in a handful of webrtc mochitests (eg; bug 1870769). I wonder if this is due to bug 1818283?

Comment 5

[Byron Campen [:bwc]]

Why is bugzilla automatically reassigning bugs when changing things like See Also?

Comment 6

[Byron Campen [:bwc]]

So, one of the commonalities of the tests that are failing is they all use a CaptureStream, and I see that there has already been one CaptureStream-related regression fixed. I don't think bug 1818283 was related to CaptureStream, so this might be a separate issue.

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

Garbage may indeed be collected now sooner. But it is still garbage, and should be collected. I hope media code doesn't for example keep weak references to unlinked objects around.
Bug 1870780 hints that it does that. But the fixes shouldn't be on GC/CC side but on media side.

Comment 8

[Byron Campen [:bwc]]

Yeah, it definitely highlights some sort of mistake, maybe a missing edge someplace.