Open Bug 1908257 Opened 7 months ago Updated 5 months ago

15+ second janks of Main Process Mainthread when scrolling with mouse wheel

Categories

(Core :: Panning and Zooming, defect, P3)

x86_64
Linux
defect

Tracking

()

People

(Reporter: jesup, Unassigned)

References

(Blocks 1 open bug)

Details

I was interacting with ChatGPT in a tab, and tried to scroll down through a large amount of output it had given. I'm not sure what happened, but after that until I restarted, any attempt to scroll with the mouse wheel would lock the browser for ~17-19 seconds. (See https://share.firefox.dev/3y7jMne)

Page up/down were fine. Closing and switching tabs had no impact.
This was with Nightly Linux, probably from mid-June, on Fedora 39, on a threadripper and Radeon GPU

It's spending a long time in SimulateUntil(aTime);.

aTime comes from mApzc.GetFrameTime().Time(); and it's simulating since mLastSimulatedTime.
So either mLastSimulatedTime is long in the past for some reason, or aTime is far in the future for some reason. I think the former is more likely.

This sounds a bit like Bug 1877972, but it seems like this issue is in the smooth scroll animation implementation. If this is reasonably reproducible, could you snag a profile with the same logging settings used there? See bug 1877972 comment 2. Since this is a MSD scroll animation issue, this only impacts nightly (see pref definition).

See Also: → 1877972
Blocks: 1760372

Regardless of the root cause, we may want to have an iteration limit in this loop, since scrolling by an incorrect amount seems like a less bad outcome than hanging while we iterate millions of times.

The severity field is not set for this bug.
:botond, could you have a look please?

For more information, please visit BugBot documentation.

Flags: needinfo?(botond)
Severity: -- → S3
Flags: needinfo?(botond)
Priority: -- → P3
You need to log in before you can comment on or make changes to this bug.