Closed Bug 1644249 Opened 4 years ago Closed 4 years ago

Crash in [@ CrashReporter::OnChildProcessDumpWritten]

Categories

(Toolkit :: Crash Reporting, defect)

Unspecified
macOS
defect

Tracking

()

RESOLVED FIXED
85 Branch
Tracking Status
firefox-esr68 --- unaffected
firefox-esr78 --- fixed
firefox77 --- unaffected
firefox78 --- wontfix
firefox79 --- wontfix
firefox80 --- wontfix
firefox83 --- wontfix
firefox84 --- fixed
firefox85 --- fixed

People

(Reporter: philipp, Assigned: gsvelto)

References

(Regression)

Details

(Keywords: crash, regression)

Crash Data

This bug is for crash report bp-475517b7-a811-4f5d-8574-3ac070200608.

Top 7 frames of crashing thread:

0 XUL CrashReporter::OnChildProcessDumpWritten toolkit/crashreporter/nsExceptionHandler.cpp:3280
1 XUL google_breakpad::CrashGenerationServer::WaitForOneMessage toolkit/crashreporter/breakpad-client/mac/crash_generation/crash_generation_server.cc:151
2 XUL google_breakpad::CrashGenerationServer::WaitForMessages toolkit/crashreporter/breakpad-client/mac/crash_generation/crash_generation_server.cc:96
3 libsystem_pthread.dylib _pthread_body 
4 libsystem_pthread.dylib _pthread_start 
5 libsystem_pthread.dylib thread_start 
6 XUL XUL@0x3f9d75f 

this crash signature is starting to show up from macos builds in firefox 78.

Most likely too late for 78 at this stage.

Gabriele, any ideas?

Flags: needinfo?(gsvelto)

Yeah, this must be some kind of race I introduced while fixing the existing one in bug 1628399. What's odd is that we never tripped on the assertion before the crashing line. This will require some time to investigate so yeah, not going to make it for 78.

Flags: needinfo?(gsvelto)

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

For more information, please visit auto_nag documentation.

Flags: needinfo?(gsvelto)
Severity: -- → S2
Flags: needinfo?(gsvelto)
See Also: → 1663088

It looks like these started happening frequently on OSX Nightly (at least by the scale of the number of users on that branch). For instance, it was 27% of all OSX crashes on the 11/18 Nightlies. Looking at the table in this bug, that started with the 20201117094406 build.

This is the set of new changesets in that build: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=6b97acd45602162d54228a399c0b313af9ed2cfd&tochange=31d67eef91da183d26334f42e78ae41d9b37ff90

Any ideas, Gabriele? Thanks.

Flags: needinfo?(gsvelto)

It looks like the crash stats also shot up around the same time (11/17) at bug 1676102, which is Mac-specific.

Note, though, that they didn't shoot up on trunk, but only on beta (and maybe also release). So possibly a change to infrastructure is responsible (if only indirectly).

See Also: → 1676102

(In reply to Andrew McCreight [:mccr8] from comment #4)

Any ideas, Gabriele? Thanks.

This is a race and the spike in crashes seems to be happening on Big Sur. Maybe some change in the scheduler made it worse. I'll make this my priority for this week.

Assignee: nobody → gsvelto
Status: NEW → ASSIGNED
Flags: needinfo?(gsvelto)

The nightly crashes are all on ARM64 which won't help me debug it because I don't have an ARM Mac (yet?). My guess is that the spike isn't a real spike but rather a bunch of Mozillians now running nightly on ARM Macs because they're working on it. Anyway I'm trying to figure out why this is happening and I've got a hunch.

I have identified a potential race that should affect both Linux and macOS and I'm positive it's the cause of this crash. I had already filed bug 1624467 to investigate it but never got around to finish that investigation. I'm preparing a patch for that bug and after it lands we'll see if this is fixed.

Depends on: 1624467

Given the volume here this was clearly caused by bug 1624467 and has been fixed by the patch there. Shall we close this or do we wait until the fix makes it to ESR?

No more crashes until buildid 20201201093815 for nightly and buildid 20201203211213 for beta. Marking this as fixed.

Status: ASSIGNED → RESOLVED
Closed: 4 years ago
Resolution: --- → FIXED
Has Regression Range: --- → yes
You need to log in before you can comment on or make changes to this bug.