Closed Bug 1179826 Opened 9 years ago Closed 5 years ago

Shutdown assert in MediaStreamGraph EnsureRunInStableState

Categories

(Core :: Audio/Video: MediaStreamGraph, defect, P3)

defect

Tracking

()

RESOLVED WORKSFORME
Tracking Status
firefox42 --- affected

People

(Reporter: jesup, Unassigned)

References

Details

(Keywords: assertion)

Attachments

(2 files)

inbound from yesterday, with bug 950660 patches (should be irrelevant).  e10s active, linux64 debug build.

Assertion failure in RunInStableState called from MSG - too late in shutdown (we're cleaning out the last CC run).  

Attached is stack, info on threads, dump of the Graph object, etc.

STR:
Loaded, started talky.io call with another local machine (using TCP-TURN), ended call, selected Leave, filled out and submitted feedback, then clicked close box on window.
bug 1182467 may provide a reliable way to reproduce this race/assert
Blocks: 1162003
Flags: needinfo?(padenot)
Bug 1179909 is going to change this code so that the final cycle collection is not "too late" to do RunInStableState (although I probably need to reorder things in that patch so that its also after the final GC).
Clearing NEEDINFO in light of comment 3. If this still happens again, please NI again.
Flags: needinfo?(padenot)
Component: Audio/Video → Audio/Video: Playback
Component: Audio/Video: Playback → Audio/Video: MSG/cubeb/GMP
We do have test_ipc.html disabled as a result of this bug:
https://dxr.mozilla.org/mozilla-central/source/dom/media/tests/mochitest/ipc/mochitest.ini?offset=100#9

I am not sure how we got here, it looks like the bug filed for the issue (bug 1182467) was marked as a duplicate of this bug.

:padenot, is there somebody who could own working on this?  I am not sure if this is a known issue to anyone who cares about this leak or test case.
Flags: needinfo?(padenot)
It would be worth it to do another try push, we changed a lot of things. Then we can find an owner if there still are problems.
Flags: needinfo?(padenot)
Changes in Bug 1179909 would have changed the behavior here I expect, but it may now be a shutdown leak.

mDetectedNotRunning is not set in the AppendMessage() call, I guess because threaded-ml did not finish before the main thread stopped receiving events.
Now that the MSG thread is not necessarily an XPCOM thread, XPCOM will not wait for that thread to shut down before closing the main thread.
this is really unrelated to this bug, but when bug 1182467 was filed, we filed it and disabled test_ipc.html- now it appears that we don't have a problem with it:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=a7e3d0db3db4
Attachment #8667139 - Flags: review?(padenot)
Comment on attachment 8667139 [details] [diff] [review]
enable dom/media/.../test_ipc.html again (1.0)

Review of attachment 8667139 [details] [diff] [review]:
-----------------------------------------------------------------

Let's land this. If something goes wrong again, we'll have a more in depth look.
Attachment #8667139 - Flags: review?(padenot) → review+
No longer blocks: 1162003
Rank: 24
Priority: -- → P2
Mass change P2->P3 to align with new Mozilla triage process.
Priority: P2 → P3
The leave-open keyword is there and there is no activity for 6 months.
:drno, maybe it's time to close this bug?
Flags: needinfo?(drno)
Status: NEW → RESOLVED
Closed: 5 years ago
Flags: needinfo?(drno)
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: