Closed Bug 1557695 Opened 6 years ago Closed 6 years ago

Getting crash when I start profiler (Or the page crashes without crashing, and reloads after I switch to it)

Categories

(Core :: Gecko Profiler, defect, P3)

defect

Tracking

()

RESOLVED WORKSFORME
Tracking Status
firefox69 --- affected

People

(Reporter: mayankleoboy1, Unassigned)

References

(Depends on 1 open bug, )

Details

(Keywords: regression, Whiteboard: crash, regression)

Crash Data

Attachments

(1 file)

definitely a recent regression

Crash Signature: [@ OOM | large | mozalloc_abort | mozalloc_handle_oom | moz_xmalloc | mozilla::MakeUnique<T> | ProfileBuffer::ProfileBuffer ]
Keywords: regression
Whiteboard: crash, regression

I am getting this after I go to http://dan-ball.jp/en/javagame/dust2/ and play for a bit, by increasing the number of particles inside the sandbox.

Crash Signature: [@ OOM | large | mozalloc_abort | mozalloc_handle_oom | moz_xmalloc | mozilla::MakeUnique<T> | ProfileBuffer::ProfileBuffer ] → [@ OOM | large | mozalloc_abort | mozalloc_handle_oom | moz_xmalloc | mozilla::MakeUnique<T> | ProfileBuffer::ProfileBuffer ] [@ OOM | large | mozalloc_abort | mozalloc_handle_oom | moz_xmalloc | ChunkedJSONWriteFunc::AllocChunk ]

Second symptom:

  1. Go to the URL
  2. Start the profiler.
  3. Play with the sandbox for 20-30 seconds
  4. Capture the profile
  5. Let it symbolicate
  6. Return back to the dust2 page

ER: Your page/content is there
AR: The page reloads when you switch to it

Summary: Getting crash when I start profiler → Getting crash when I start profiler (Or the page crashes without crashing, and reloads after I switch to it)

Not too sure how to help...

These are Out Of Memory (OOM) errors, either when the profiler or a process starts (signatures with "ProfilerBuffer") or when a profile is captured ("ChunkedJSONWriteFunc").
Looking at graphs of reports over time, there's no significant increase that's clearly visible. Maybe a few more crashes of the latter signature from 6 June, we should monitor this for a trend.

And I didn't see any obvious recent patches that could explain an increase in memory consumption. Could it be that the in-Firefox Profiler Toolbar Icon has made more people use the Profiler recently?

I'm not dismissing your experience Mayank, just that at the moment it's hard to see widespread issue (yet?), and find the cause so we can fix it.

On your side, could you please try reducing the "Buffer Size" in the Profiler toolbar icon settings?

One thing that we should do, is actually handle OOMs in the profiler, so we don't crash the process when there's not enough memory to start the profile or capture a profile.
Of course it would just wallpaper over the real issue if there's one. Nevertheless it would be the right thing to do, especially as the Profiler may use a lot of memory! I will open a separate bug for these, so we can keep this one here to monitor the potential issue that your first reported.

Priority: -- → P3

Adding dependence on bug 1558402, so if that one gets fixed first, this bug here may seem to disappear even if the root cause was not found.

Depends on: 1558402

Since this is canvas + WR, could this be another manifestation of bug 1558100 ?

See Also: → 1558100

cant repro anymore.

Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: