1481868 - Stop using media conduits and webrtc.org from main

Status: NEW

(Core :: WebRTC: Audio/Video, enhancement, P3)

Description

[Byron Campen [:bwc]]

The media conduit and webrtc.org code is deadlock-prone, so we should not be using it from main (or STS for that matter). It would be better to have a dedicated thread, one per PeerConnection.

Comment 1

[Dan Minor [:dminor]]

Makes sense to me, but I'd prefer to hold off until after the next webrtc.org update lands in Bug 1376873 because moving to the Call interface in AudioConduit required quite a few changes.

Comment 2

[Byron Campen [:bwc]]

It may end up being easiest to run the conduit code on this new thread too. I guess we'll see.

Comment 3

[Dan Minor [:dminor]]

(In reply to Byron Campen [:bwc] from comment #2)
> It may end up being easiest to run the conduit code on this new thread too.
> I guess we'll see.

Good point. I have a bunch of thread assertions in channel_proxy.cc in the update disabled for now. Moving the conduit to a new thread might be a closer match to how upstream runs things than what we have now.

Comment 4

[Byron Campen [:bwc]]

Running the conduit code on a separate thread would probably also allow us to eliminate some mutex locking, which might fix deadlock bugs too.