add command line flags for webrtc

[kereliuk]

This is intended as a temporary fix to allow the webrtc tests in chrome to call getUserMedia without failing out.

[lukebjerring]

Subbing out for @mattl since he'll have more knowledge of flags for wpt run.

[foolip]

@mattl, let us know if you want to wait to merge this until the code freeze is over, which I believe is March 9. Up to you.

[rwaldron]

These can be combined

command.extend([
    '--binary-arg=--use-fake-ui-for-media-stream',
    '--binary-arg=--use-fake-device-for-media-stream'
])

[foolip]

Ping @kereliuk to combine these so we're all set for Monday.

[mattl]

I think we'll wait until the 9th, in the meantime this can easily be run on a feature branch if needed.

@rwaldron's suggestion is a good one.

[foolip]

@mattl, that sounds good, thanks! @kereliuk, anything we should do to increase confidence that this will in fact work once landed?

[foolip]

@kereliuk, a concrete idea is to run the webrtc/ directory with and without these flags and list and look at any results that change. If anything goes from passing to not passing, that's weird, and hopefully it's all in the other direction.

[mariestaver]

@foolip Thank you and yes, we'll pause this one until after 3/9 -- I've added it to our backlog to be sure it gets picked up then.

[foolip]

@alvestrand FYI, this will make the wpt.fyi results more meaningful for you, as promised

[kereliuk]

I ran diffs on the results of with and without these flags.

Got [2, 6] with flags for /webrtc/RTCPeerConnection-ontrack.https.html
Got [1, 6] without flags for /webrtc/RTCPeerConnection-ontrack.https.html

Got [2, 2] with flags for /webrtc/RTCDTMFSender-ontonechange-long.https.html
Got [0, 2] without flags for /webrtc/RTCDTMFSender-ontonechange-long.https.html

Got [8, 12] with flags for /webrtc/RTCDTMFSender-ontonechange.https.html
Got [0, 12] without flags for /webrtc/RTCDTMFSender-ontonechange.https.html

Got [1, 3] with flags for /webrtc/RTCRtpSender-getStats.https.html
Got [0, 3] without flags for /webrtc/RTCRtpSender-getStats.https.html

Got [6, 7] with flags for /webrtc/RTCPeerConnection-setRemoteDescription-replaceTrack.https.html
Got [0, 7] without flags for /webrtc/RTCPeerConnection-setRemoteDescription-replaceTrack.https.html

Got [1, 2] with flags for /webrtc/RTCRtpReceiver-getContributingSources.https.html
Got [0, 2] without flags for /webrtc/RTCRtpReceiver-getContributingSources.https.html

Got [3, 7] with flags for /webrtc/RTCDTMFSender-insertDTMF.https.html
Got [0, 7] without flags for /webrtc/RTCDTMFSender-insertDTMF.https.html

Got [400, 692] with flags for /webrtc/interfaces.https.html
Got [1, 5] without flags for /webrtc/interfaces.https.html

Got [11, 15] with flags for /webrtc/RTCPeerConnection-setRemoteDescription-tracks.https.html
Got [0, 15] without flags for /webrtc/RTCPeerConnection-setRemoteDescription-tracks.https.html

Got [2, 2] with flags for /webrtc/simplecall.https.html
Got [1, 2] without flags for /webrtc/simplecall.https.html

Got [3, 14] with flags for /webrtc/RTCPeerConnection-getStats.https.html
Got [1, 14] without flags for /webrtc/RTCPeerConnection-getStats.https.html

Got [5, 13] with flags for /webrtc/RTCPeerConnection-removeTrack.https.html
Got [0, 13] without flags for /webrtc/RTCPeerConnection-removeTrack.https.html

Got [4, 9] with flags for /webrtc/RTCPeerConnection-addTrack.https.html
Got [0, 9] without flags for /webrtc/RTCPeerConnection-addTrack.https.html

Got [12, 13] with flags for /webrtc/RTCPeerConnection-track-stats.https.html
Got [0, 13] without flags for /webrtc/RTCPeerConnection-track-stats.https.html

Got [1, 2] with flags for /webrtc/RTCRtpReceiver-getSynchronizationSources.https.html
Got [0, 2] without flags for /webrtc/RTCRtpReceiver-getSynchronizationSources.https.html

15 different results

This is just from the summary file, so I'm not 100% sure if anything fails that used to pass on the subtest level of granularity; however, it does increase the pass rate overall as above and the only differences noted only had at most 1 subtest passing originally, so I'm guessing not.

[foolip]

Sounds solid. Even if some subtest does go from passing to failing, that must be accidental and will just have to be fixed.

[fippo]

you might want to add https://peter.sh/experiments/chromium-command-line-switches/#mute-audio which (if I recall correctly) was added by the webrtc folks to avoid the beep sounds when running stuff locally