1322108 - Crash in webrtc::GofInfoVP9::CopyGofInfoVP9

Status: RESOLVED WORKSFORME

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

Description

[Philipp Hancke [:fippo]]

This bug was filed from the Socorro interface and is 
report bp-862a3d5d-6272-42bf-9cd8-886d22161205.
=============================================================

I started getting crashes in our appear.in smoke tests recently. Another example:
https://crash-stats.mozilla.com/report/index/2198bc0c-0dcc-42b9-ac2b-248f02161205

These happen only when I enable VP9 via media.peerconnection.video.vp9_enabled
In this case, both browsers end up using VP9 (why?)

The crashes only happen from selenium-runs but I can reproduce parts of the behaviour (no video, just audio) in production as well but it doesn't crash. Might be the combination of flags from https://github.com/fippo/testbed/blob/master/webdriver.js#L23, i'll see if I can make the exact smoke test public

Comment 1

[Philipp Hancke [:fippo]]

moar crashes, this time on https://webrtc.github.io/samples/src/content/peerconnection/constraints/ with media.peerconnection.video.vp9_enabled;true 
Remote picture took a while to appear, then froze immediately. Crash after a while:
https://crash-stats.mozilla.com/report/index/f5a5c869-3d08-4700-8901-4c6d12161205

Comment 2

[Philipp Hancke [:fippo]]

Attachment: vp9.html

i've attached a about:webrtc dump from the side that is _not_ crashing. This shows that VP9 is used -- see the 121 in the remote description

Comment 3

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

We didn't enable VP9 in 50; it's enabled in 51.  For 51, we fixed a bug in handling of VP9 where it barfs on input resolution changes (much like VP8 did before we fixed it).  I suspect this is the same problem - can you retry with 51?

Comment 4

[Philipp Hancke [:fippo]]

51.0b6 and 53 work for me. This wasn't a problem in 49 so if that bad was introduced in 50 and 51 this is not a concern. Thanks!