Open Bug 1633019 Opened 7 months ago Updated 7 months ago

Crash in [@ mozilla::PresShell::EventHandler::MaybeFlushThrottledStyles]

Categories

(Core :: DOM: UI Events & Focus Handling, defect, P3)

defect

Tracking

()

Tracking Status
firefox75 --- affected
firefox76 --- affected

People

(Reporter: hiro, Unassigned)

References

Details

(Keywords: crash)

Crash Data

This bug is for crash report bp-a2a6ff81-8426-4a32-ac3e-294a70200424.

Top 10 frames of crashing thread:

0 XUL mozilla::PresShell::EventHandler::MaybeFlushThrottledStyles layout/base/PresShell.cpp:7283
1 XUL mozilla::PresShell::EventHandler::HandleEventUsingCoordinates layout/base/PresShell.cpp:6577
2 XUL mozilla::WidgetEvent::IsUsingCoordinates const widget/WidgetEventImpl.cpp:449
3 XUL mozilla::PresShell::EventHandler::HandleEvent layout/base/PresShell.cpp:6517
4 XUL mozilla::PresShell::HandleEvent layout/base/PresShell.cpp:6443
5 XUL mozilla::PresShell::EventHandler::MaybeHandleEventWithAnotherPresShell layout/base/PresShell.cpp:7172
6 XUL mozilla::PresShell::EventHandler::HandleEvent layout/base/PresShell.cpp:6505
7 XUL mozilla::PresShell::HandleEvent layout/base/PresShell.cpp:6443
8 XUL nsViewManager::DispatchEvent view/nsViewManager.cpp:751
9 XUL nsView::HandleEvent view/nsView.cpp:1135

Copying from bug 1555645 comment 9.

I haven't audited all crash reports, but as far as I can tell the call stack is including MaybeHandleEventWithAnotherPresShell which means the crash happened inside a nested document. And the crash happened at the code around we tried to get the root pres shell, it may be possible that the root pres shell is an ancestor of the initial target pres shell, I am quite unsure though.

Anyways, my question is "do we really need to flush throttled styles for the root pres shell?". My answer is "no". Given that we've already called MaybeFlushPendingNotifications just before we call MaybeFlushThrottledStyles, it would be sufficient.

As a side note about MaybeFlushThrottledStyles, emilo has a plan to drop the MaybeFlushThrottledStyles call.

Crash reports are quite annoying in this case. They don't clearly indicate where the crash is happening. The lines depend on the platform and often don't make quite sense.
This isn't very common crash, and it is null+offset (I just haven't found what is null).

Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.