1270004 - Crash in winmm_stream_init

Status: RESOLVED FIXED

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

Description

[Nicholas Nethercote [inactive]]

This bug was filed from the Socorro interface and is 
report bp-8e8217ab-be36-424b-a1e8-950a82160503.
=============================================================

New crash in Nightly 20160502030207. 4 occurrences across 2 installations, though judging by the crash URLs it's the same user for both installations.

Crash address is 0x0. Top frame of the stack trace corresponds to this line in cubeb_winmm.c, which is a really odd-looking assertion:

  XASSERT(!input_stream_params && "not supported.");

Comment 1

[Nicholas Nethercote [inactive]]

kinetik, can you please take a look?

Comment 2

[Matthew Gregan [:kinetik]]

That assert is to catch a cubeb_stream_init requesting an input or full-duplex stream from a backend that hasn't (yet) implemented the feature.  It should probably just return CUBEB_ERROR_NOT_SUPPORTED instead.

I guess what's happening here is that cubeb_init fails for the WASAPI backend for some reason, so falls back to the WinMM backend.  The full-duplex MSG code knows to only request full-duplex streams on Vista+, but probably nobody considered that Vista+ can fall back to WinMM in unusual situations.

So, to fix this: we can make the assert an error return, but that'll probably still leave the user with a broken WebRTC, so we'll probably need to fix that up to deal with falling back to the old (non-libcubeb) input/full-duplex codepath.

Comment 3

[Matthew Gregan [:kinetik]]

Attachment: fix

This fixes the crash (assert), but as mentioned it'll still leave WebRTC's capture broken in the situation where the fallback happens.

Comment 4

[Matthew Gregan [:kinetik]]

09:36 < jesup> kinetik: in that case, we may want to merge achronops patch and not wait for winmm
               (padenot is on holiday)
09:39 < kinetik> jesup: we want the assert removed, at least, so if you can r+ that in padenot's
                 absence i'll merge it
09:39 < kinetik> i merged achronop's fix, minus the stringfying bit
09:48 < jesup> kinetik: r+

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/cc9051869d1a

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/cc9051869d1a