Crash in [@ nsObserverService::NotifyObservers] called from nsChildView::PaintWindow
Categories
(Core :: Widget: Cocoa, defect)
Tracking
()
People
(Reporter: marcia, Assigned: mstange)
References
(Blocks 1 open bug)
Details
(Keywords: crash, regression)
Crash Data
This bug is for crash report bp-4e0833d9-3949-49ef-a52d-c73c10190311.
Seen while looking at 65 release crash stats - smaller volume crash which appears to affect Mac 10.14 more than the other platforms: https://bit.ly/2TBC6Qg. What concerned me was several comments mentioned "Fresh install of Firefox crashes on startup" as well as "under mojave 10.14.4 no possible opening of firefox"
Code was touched in Bug 1502284 in the 65 timeframe
ni on :gsvelto for ideas
Top 10 frames of crashing thread:
0 XUL nsObserverService::NotifyObservers xpcom/ds/nsObserverService.cpp:173
1 XUL mozilla::PresShell::DidPaintWindow layout/base/PresShell.cpp:8078
2 XUL nsView::DidPaintWindow view/nsViewManager.cpp:703
3 XUL nsChildView::PaintWindow widget/cocoa/nsChildView.mm:1514
4 libobjc.A.dylib objc_msgLookup_fpret
5 XUL -[ChildView drawRect:] widget/cocoa/nsChildView.mm:3658
6 AppKit GCC_except_table7
7 libobjc.A.dylib _class_initialize
8 AppKit GCC_except_table7
9 libobjc.A.dylib _class_resolveInstanceMethod
Comment 1•6 years ago
|
||
The crash reason is:
MOZ_CRASH(Using observer service off the main thread!)
It looks to me like gsvelto's patch just moved around where this check was done, rather than added any checks.
Comment 2•6 years ago
|
||
It looks like this crash also existed in 64: bp-9c6be3f5-f6d4-4310-933f-302a70190105
In fact, I can see similar crashes going back to 60 ESR: https://crash-stats.mozilla.com/search/?moz_crash_reason=~Using%20observer%20service%20off%20the%20main%20thread&proto_signature=~nsChildView%3A%3APaintWindow&product=Firefox&platform=Mac%20OS%20X&date=%3E%3D2018-12-11T20%3A00%3A00.000Z&date=%3C2019-03-11T20%3A00%3A00.000Z&_facets=signature&_sort=version&_sort=-date&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#crash-reports
Markus, any ideas?
FWIW, this crash from 60 ESR has a better looking stack than the more recent crashes I saw: bp-6b3d176d-7d99-4722-a531-6407f0181215
Comment 3•6 years ago
•
|
||
There are a couple of crashes talking about crashing in light mode.
"Still crashes on startup in Light Mode."
bp-2948c019-df1e-4c38-b5db-3fcdd0181213
This crash report maybe has some STR:
Happens after setting:
> defaults write -g NSRequiresAquaSystemAppearance -bool NO
Updated•6 years ago
|
Updated•6 years ago
|
| Assignee | ||
Comment 4•6 years ago
|
||
We added a main thread check to this method in bug 1515183. Should we uplift that patch?
Comment 5•6 years ago
|
||
I think so.
| Assignee | ||
Comment 6•6 years ago
|
||
I'll request uplift.
Comment 7•6 years ago
|
||
Thanks!
Comment 8•6 years ago
|
||
Looks like the remaining crashes with this signature on 66 are on Windows. Calling this Mac-specific variant fixed by bug 1515183.
Description
•