1329594 - Crash in cubeb_resampler_destroy

Status: RESOLVED FIXED

(Core :: Audio/Video: cubeb, defect, P2)

Description

[Nicholas Nethercote [inactive]]

This bug was filed from the Socorro interface and is 
report bp-213ba506-6105-4703-822a-b49cc2170108.
=============================================================

First occurrence of this crash appears to be on 20161022030204. All crashes are on Mac. Almost all crashes are on Aurora (52) with only a handful on Nightly (53), which is weird.

kinetik, can you please investigate?

Comment 1

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

Alex/Paul - anything you landed around that time, especially for Mac?

Pretty low-frequency crash, nullptr or near.
This likely is fixed in 53, since every crash I see is 52.  Perhaps we can uplift a fix, if it's not too large.

Comment 2

[Paul Adenot (:padenot)]

I this this will be fixed by Alex's patches in https://github.com/kinetiknz/cubeb/pull/208, which is currently pending review upstream. 

The crashes are coming from the event handler, which are known to be quite capricious, and the thread safety of the related API calls is not well documented.

I'll let Alex comment if he has more details.

Comment 3

[Matthew Gregan [:kinetik]]

Clearing NI based on comment 2.

Comment 4

[Alex Chronopoulos [:achronop]]

The crash should happen inside audiounit_property_listener_callback. I see a potential race case if the listener is called and the stream is destroyed in the same time. I will try to repro a crash like that.

Comment 5

[Julien Cristau [:jcristau] (back April 22)]

Mass wontfix for bugs affecting firefox 52.

Comment 6

[Paul Adenot (:padenot)]

Alex, I think that's been fixed by your patch in https://github.com/kinetiknz/cubeb/commit/00ee12037090f0d4fe7d75d9e36c86a150a18418, does that sound correct ?

Comment 7

[Alex Chronopoulos [:achronop]]

Yeah, that sounds about right. Given that it appears only in 52, it should be fixed by the commit mentioned in comment 6.