Closed
Bug 1401668
Opened 7 years ago
Closed 7 years ago
ImageBridgeParent acquires compositor thread reference too late
Categories
(Core :: Graphics, defect, P1)
Core
Graphics
Tracking
()
RESOLVED
FIXED
mozilla57
Tracking | Status | |
---|---|---|
firefox-esr52 | --- | unaffected |
firefox55 | --- | unaffected |
firefox56 | --- | wontfix |
firefox57 | --- | fixed |
People
(Reporter: aosmond, Assigned: aosmond)
References
Details
(Keywords: crash, Whiteboard: [gfx-noted])
Attachments
(1 file)
2.36 KB,
patch
|
dvander
:
review+
|
Details | Diff | Splinter Review |
As discovered in crash reports for bug 1398070, e.g.: https://crash-stats.mozilla.com/report/index/4a390f71-3770-4cf8-a353-c07a90170920 The compositor thread is getting shutdown too early because ImageBridgeParent is still alive. The working theory is that between ImageBridgeParent construction and ImageBridgeParent::OnChannelConnected, we call shutdown and we never acquire the reference to the thread. Whatever it is, PImageBridgeParent is the protocol still alive causing the crash.
Assignee | ||
Comment 1•7 years ago
|
||
try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=5b84767a41e44c075e4f5b74e06ceed5509f2967
Attachment #8910418 -
Flags: review?(dvander)
Assignee | ||
Updated•7 years ago
|
Assignee: nobody → aosmond
Status: NEW → ASSIGNED
Keywords: crash
Priority: -- → P3
Whiteboard: [gfx-noted]
Attachment #8910418 -
Flags: review?(dvander) → review+
Pushed by aosmond@gmail.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/74608fbc8d56 Change ImageBridgeParent to acquire a compositor thread reference on construction. r=dvander
Updated•7 years ago
|
Comment 3•7 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/74608fbc8d56
Status: ASSIGNED → RESOLVED
Closed: 7 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla57
Comment 4•7 years ago
|
||
Is this something we'd want to consider as an Fx56 dot-release ride-along candidate or should it just ride the 57 train?
Assignee | ||
Comment 5•7 years ago
|
||
It is probably safe to uplift, but I would only recommend it if we see crashes on 56. Since one of the patches that was uplifted as part of bug 1398021 was in the pushlog for when this signature appeared, I feared that it may introduce it somehow to 56. At this point, I'm not so sure -- looking at what fixed the issue, this problem had potential to be there for years. Something else probably made it more likely on 57.
Flags: needinfo?(aosmond)
Comment 6•7 years ago
|
||
Thanks for the context. I'm going to call this wontfix for 56 then and we can always revisit if it shows up after we ship next week.
You need to log in
before you can comment on or make changes to this bug.
Description
•