1215195 - continued abort crash in mozilla::layers::PImageBridgeChild::Write

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Description

[Robert Kaiser]

[Tracking Requested - why for this release]:

+++ This bug was initially created as a clone of Bug #1208071 +++

Even after the fix in bug 1208071, we still have crashes with this signature, even though a reduced amount, those are 1.5% of 42.0b6 crashes right now.

Examples are bp-e8f793ed-457e-4e14-9496-1f1d32151014 or bp-9c634c8a-9e5c-4546-9f6d-214ef2151014

Nical, any idea what's up here?

Comment 1

[Sylvestre Ledru [:Sylvestre]]

Tracking as it is an important crash.

Comment 2

[Alice0775 White]

I have a STR, see Bug 1213344 Comment#9

Comment 3

[Milan Sreckovic [:milan] (needinfo for best results)]

Are all of the remaining crashes shutdown?

Comment 4

[Robert Kaiser]

(In reply to Milan Sreckovic [:milan] from comment #3)
> Are all of the remaining crashes shutdown?

No idea, I don't know how I could check for that.

Comment 5

[Milan Sreckovic [:milan] (needinfo for best results)]

(In reply to Robert Kaiser (:[e-mail]) from comment #4)
> (In reply to Milan Sreckovic [:milan] from comment #3)
> > Are all of the remaining crashes shutdown?
> 
> No idea, I don't know how I could check for that.

I was naively assuming that metadata ShutdownProgress set to xpcom-shutdown indicated that we were in shutdown procedure, perhaps that's not the case?

Comment 6

[Robert Kaiser]

Oh, OK, sounds like that, yes. And the shutdown progress facet on this search says that 99.92% of those crashes are there, yes:

https://crash-stats.mozilla.com/search/?product=Firefox&signature=%3Dmozalloc_abort+|+NS_DebugBreak+|+mozilla%3A%3Alayers%3A%3APImageBridgeChild%3A%3AWrite&_facets=signature&_facets=shutdown_progress&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-shutdown_progress

Comment 7

[Sylvestre Ledru [:Sylvestre]]

Shutdown hangs are sad but not big enough to track that late in the cycle. Untracking.

Comment 8

[Sylvestre Ledru [:Sylvestre]]

But tracking for 43.

Comment 9

[Liz Henry (:lizzard) (relman/hg->git project)]

Still a topcrash in 43 beta, but marking wontfix as we are a week from release.  
I don't see either of these signatures in 44 or 45, perhaps a good sign.

Comment 10

[Nicolas Silva [:nical]]

This is another symptom of bug 1207220: some video frames which hold on to gfx resources are still alive during/after gfx shutdown.

Comment 11

[Nicolas Silva [:nical]]

No crash since bug 1209039 landed (nightly and aurora). There's another bug that should have helped with this on beta as well but I don't have the bug number handy.