Enabling Firefox Profiler on large page crashes Firefox
Categories
(Core :: Gecko Profiler, defect, P2)
Tracking
()
People
(Reporter: tdulcet, Unassigned)
References
(Blocks 1 open bug)
Details
Attachments
(2 files)
User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:109.0) Gecko/20100101 Firefox/113.0
Steps to reproduce:
- Enable the Firefox Profiler.
- Load a large page like this: https://tdulcet.github.io/Missing-Words/
- Scroll up and down quickly until Firefox freezes.
Actual results:
Firefox freezes after a few seconds and never regains responsiveness.
Expected results:
Firefox should behave the same as when the profiler is not enabled. In other words, Firefox should not crash. I was able to reproduce this issue several times.
The above page has bad performance in Firefox (it has significantly better performance in Chrome), so I was trying to record a profile so I could report this performance issue to Mozilla. However, when Firefox is frozen, I of course have no way to get the resulting profile...
Reporter | ||
Updated•1 year ago
|
Updated•1 year ago
|
Comment 1•1 year ago
|
||
The severity field is not set for this bug.
:canova, could you have a look please?
For more information, please visit BugBot documentation.
I can't reproduce this on my machine but I have a rather powerful one. Will try to check it on Windows as well. Do you have any insights that could help me reproduce it?
Reporter | ||
Comment 3•1 year ago
|
||
I also have a relatively powerful system, with 12 CPU threads and 32 GiB of RAM. I used the default "Firefox" profiler settings. It takes around 30-45 seconds of scrolling for Firefox to fully freeze/crash. See the above screenshot from the latest Firefox Nightly 117 for example (note the "not responding" in the title bar).
Reporter | ||
Comment 4•1 year ago
|
||
If you enable the "Native Allocations" profiler setting, it does hard crash and open the crash reporter, but I suspect this is a different bug. See this crash report for example: https://crash-stats.mozilla.org/report/index/48e34530-6d9e-4427-931e-7d9890230621
Yeah that crash on "native allocations" sounds like a different issue. Could be worth to file a separate bug.
I managed to reproduce the hang locally after a few try now. Managed to sample using Activity Monitor on macOS. Attaching the sample file.
It looks like the main thread is mostly busy inside mozilla::detail::EventQueueInternal<16ul>::PutEvent
. Not sure about the couse of this though. Will investigate more.
(had to zip it because it was slightly larger than the 10mb limit.)
Description
•