1068981 - Crash in MediaStreamGraphImpl::CollectReports when collecting DMD reports

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Randall Barker [:rbarker]]

Running DMD on MacOS X will crash on an assert in when MediaStreamGraphImpl::CollectReports calls CurrentDriver()->WakeUp();

Comment 1

[Randall Barker [:rbarker]]

I have been unable to reproduce. I did not save the stack trace when I first encountered this and now that I'm trying to reproduce it, I am unable to.

Comment 2

[Eric Rahm [:erahm]]

Randall, do you have any sort of STR for this? Maybe a specific page or test?

Comment 3

[Randall Barker [:rbarker]]

(In reply to Eric Rahm [:erahm] from comment #2)
> Randall, do you have any sort of STR for this? Maybe a specific page or test?

I was debugging a memory leak while streaming a tab. I didn't save the stack trace and when I later tried to reproduce for the bug, I was unable to. I may have been the victim of a timing related bug that is difficult to reproduce? The media pipeline was busy encoding a tab while I was attempting to save a report so maybe that increased the chance of hitting the mGraphImpl->GetMonitor().AssertCurrentThreadOwns(); ?

Comment 4

[Randall Barker [:rbarker]]

Attachment: Stack Trace

Realized I forgot to uncomment the line causing problems yesterday when trying to reproduce. After uncommenting, reproduced on first try.

Comment 5

[Randall Barker [:rbarker]]

Attachment: .mozconfig used to build

Comment 6

[Eric Rahm [:erahm]]

This appears to be some sort of fallout from bug 848954, previously we would grab a lock before calling |WakeUp|.

Comment 7

[Eric Rahm [:erahm]]

Attachment: Acquire monitor before calling WakeUp

Comment 8

[Eric Rahm [:erahm]]

Randall, would you mind testing this patch to see if it fixes your issue? I've confirmed it builds, runs, and an about:memory report worked w/o crashing.

Comment 9

[Randall Barker [:rbarker]]

(In reply to Eric Rahm [:erahm] from comment #8)
> Randall, would you mind testing this patch to see if it fixes your issue?
> I've confirmed it builds, runs, and an about:memory report worked w/o
> crashing.

It appears to fix the crash I was seeing. I ran without the patch and crashed as before. After applying the patch was able to run multiple times without issue.

Comment 10

[Eric Rahm [:erahm]]

remote:   https://hg.mozilla.org/integration/mozilla-inbound/rev/c37936058203

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/c37936058203

Comment 13

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 8495632 [details] [diff] [review]
Acquire monitor before calling WakeUp

This is blocking CAF stability testing. a=me for uplift.

Comment 14

[Randell Jesup [:jesup] (needinfo me)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/32f99eac065c