Browser hang while watching video in PiP
Categories
(Core :: Panning and Zooming, defect)
Tracking
()
People
(Reporter: 1justinpeter, Unassigned)
References
Details
Attachments
(5 files)
profile of the hung process (NB: overall CPU usage is low)
4 or 5 times in the last two weeks I've had Nightly hang (Windows states "not responding", user input does nothing for at least tens of minutes) while watching video. In all cases, I was watching a youtube video at 3x speed (give or take) in picture-in-picture. However, I can't conclusively link any of those to the issue, as that describes most of the time I'm watching media in Firefox. (It describes a significant portion of the time I'm browsing, as well.) I attempted to get a stack trace and whatnot (I asked on Matrix for some direction), which I'll attach, but I really have no idea what's useful and what's not. about:support is also attached.
During the hang, video hangs like the rest of the browser but audio continues playing until some buffer runs out.
|
Reporter | |
Comment 1•2 years ago
|
||
|
|
Reporter | |
Comment 2•2 years ago
|
||
|
|
Reporter | |
Comment 3•2 years ago
|
||
|
|
Reporter | |
Comment 4•2 years ago
|
||
|
||
Comment 5•2 years ago
|
||
We have a little utility that will crash firefox and generate a crash report you can access from about:crashes. Would you mind trying to catch this hang with that if you can?
https://github.com/b0bh00d/crash-firefox
Another option would be to try and catch this in the profiler, but you'd need to be able to reliably reproduce. Visit https://profiler.firefox.com/ for more information. When capturing, please select the Media profile in the drop down.
|
|
Reporter | |
Comment 6•2 years ago
|
||
I can't reliably reproduce but I'll try that utility next time I run across the issue. I can't say when that'll be (again, not reliably reproducible) but I'm guessing based on its frequency in the past that it'll be sometime in the next few days.
|
|
Reporter | |
Comment 7•2 years ago
|
||
I think I ran into this this morning (it happened right at the beginning of the video, rather than in the middle, but otherwise all of the symptoms were identical) and ran that tool, but nothing showed up in about:crashes afterward. Perhaps this is because I just went with the process it defaulted to and then checked the box to kill all processes with the same name; next time I'll try getting the PID from task manager (which appears to be the main process) and specifically killing that one.
|
|
Reporter | |
Comment 8•2 years ago
|
||
Ah, per https://github.com/bsmedberg/crashfirefox-intentionally this probably defaults to the most recently created process, and per bug 1600145 this no longer works for content processes, so it may work better if I set the PID to the parent process.
|
|
Reporter | |
Comment 9•2 years ago
|
||
|
|
Reporter | |
Comment 10•2 years ago
|
||
|
|
||
Comment 11•2 years ago
|
||
In all of these the main thread appears to be sitting in a waitforsyncnotify call related to apz. Maybe apz folks can help diagnose here.
|
|
||
Updated•2 years ago
|
|
|
Reporter | |
Comment 12•2 years ago
|
||
Indeed, I just had this happen while not watching video. (It served as a great reminder to save things and not leave a half-dozen tabs open in the middle of stuff for a couple days, because data loss. In hindsight, I should've seen that coming.)
(As a side note: it seems more frequent when opening tabs, but happens without interaction as well.)
|
||
Comment 13•2 years ago
|
||
The severity field is not set for this bug.
:botond, could you have a look please?
For more information, please visit auto_nag documentation.
|
||
Comment 14•2 years ago
|
||
(In reply to Jim Mathies [:jimm] from comment #11)
In all of these the main thread appears to be sitting in a waitforsyncnotify call related to apz. Maybe apz folks can help diagnose here.
The parent process main thread performing a sync IPC call during event dispatch is expected. (It asks the GPU process to hit-test the event to determine, among other things, which content process needs to handle it.)
Is there a way to see in the crash report what the GPU process (in particular its main thread and RenderBackend thread) is doing at this time, i.e. why it's not responding to the message?
|
||
Comment 15•2 years ago
|
||
No, we'd need a separate crash report for the GPU process to see that.
|
|
||
Updated•2 years ago
|
Description
•