1872850 - Nightly process became completely unresponsive during a build, and never recovered.

Status: NEW

(Core :: Widget: Cocoa, defect, P5)

Description

[Emilio Cobos Álvarez (:emilio)]

Attachment: Data from the macOS reporter.

I saw this happen twice so far: I'm doing a build and doing something else on my terminal, and going back to Nightly doesn't work at all.

I'm not sure if https://crash-stats.mozilla.org/report/index/bf183466-64c0-4ef1-af79-13b120240103 is related or not. Relevant code seems to be related to bug 1864641 so cc'ing relevant folks.

I attached the info I got from force-quitting. I also got a trace doing xctrace, but I basically got no info... Happy to attach it tho.

Comment 1

[Emilio Cobos Álvarez (:emilio)]

Attachment: traces.zip

Recorded with xctrace record --template 'Time Profiler' --attach 2813 so not sure if useful.

Comment 2

[Stephen A Pohl [:spohl]]

Are you saying that this is not a dupe of bug 1864641?

Comment 3

[Emilio Cobos Álvarez (:emilio)]

That is a shutdown hang. This is not, this is the parent process being stuck

Comment 4

[Emilio Cobos Álvarez (:emilio)]

My point is that I'm not sure whether they're related in the sense that the parent process seems to be killing content processes according to the attachment in comment 0. Those processes might not be responsive because they're on the background but this is causing the parent process to hang ~indefinitely or at least for a really long time

Comment 5

[Markus Stange [:mstange]]

Attachment 9370867 [details] shows the following:

1. On the main thread, ContentParent::KillHard blocks in ExceptionHandler::WriteMinidump
2. On the "Breakpad ExceptionHandler" thread, MinidumpGenerator::WriteBootargsStream blocks in malloc
3. On the "com.apple.NSEventThread" thread, malloc is suspended in the kernel: *11 ??? (kernel.release.t6020 + 734208) [0xfffffe0008613400] (suspended) 1-11

Not sure what to make of this.

Comment 6

[Emilio Cobos Álvarez (:emilio)]

ni? me to file a radar / maybe bring it up with apple folks, because ^ makes no sense to me either.

Comment 7

[Haik Aftandilian [:haik]]

(In reply to Markus Stange [:mstange] from comment #5)

Attachment 9370867 [details] shows the following:

1. On the main thread, ContentParent::KillHard blocks in ExceptionHandler::WriteMinidump
2. On the "Breakpad ExceptionHandler" thread, MinidumpGenerator::WriteBootargsStream blocks in malloc
3. On the "com.apple.NSEventThread" thread, malloc is suspended in the kernel: *11 ??? (kernel.release.t6020 + 734208) [0xfffffe0008613400] (suspended) 1-11

Not sure what to make of this.

Is this the Apple profiler freezing the parent process to get the thread stacks while the parent process was trying to write the minidump?

Am I right in thinking the KillHard stack is a result of the force quit? But we're interested in the original hang.

Comment 8

[BugBot [:suhaib / :marco/ :calixte]]

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

For more information, please visit BugBot documentation.

Comment 9

[Brad Werth [:bradwerth]]

I believe this is Bug 1864641 and that killing the parent process may result in differing outcomes and call stacks.