1057955 - Support MediaStreamTrack.stop()

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

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

To properly support uses of replaceTrack() (and per the spec), we need to be able to stop() a track in order to do things like avoid having two camera open at once:

  fake_track.enabled = false; // use black for transition
  old_track = sender.track;
  sender.replaceTrack(fake_track, function() {
    old_track.stop();
    navigator.getUserMedia(constraints, function(stream) {
      sender.replaceTrack(stream.getXxxxxTracks[0], function () {...}, fail);
    }, fail);
  }, fail);

Note that B2G camera doesn't seem to be able to open both at once; Not sure about android.  Even without the 2-step replacement, we'd like to be able to stop a track once it's replaced.

Comment 1

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

Attachment: Support MediaStreamTrack.stop()

Comment 2

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

Comment on attachment 8478104 [details] [diff] [review]
Support MediaStreamTrack.stop()

Need some more thought about the interaction with getUserMedia/TrackUnion (ugh)

Comment 3

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

Attachment: Stop the backend capture when the MSG track is stopped

Comment 4

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

Comment on attachment 8478104 [details] [diff] [review]
Support MediaStreamTrack.stop()

trivial DOM change (uncomment out a method we hadn't implemented before); simple change to DOMMediaStream

Comment 5

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

Attachment: Adjust TrackUnion TrackID mapping to be 1:1 whenever possible with source tracks

roc's patch via email; I r+'d it

Comment 6

[Mike Habicher [:mikeh] (high bugzilla latency)]

(In reply to Randell Jesup [:jesup] from comment #0)

> Note that B2G camera doesn't seem to be able to open both at once; Not sure
> about android.  Even without the 2-step replacement, we'd like to be able to
> stop a track once it's replaced.

The one-camera-at-a-time on B2G is a limitation of the AOSP layer, not Gecko, so I would expect Android to have the same behaviour.

Comment 7

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Comment on attachment 8479238 [details] [diff] [review]
Stop the backend capture when the MSG track is stopped

Review of attachment 8479238 [details] [diff] [review]:
-----------------------------------------------------------------

r- for potential null-dereference.

::: dom/media/MediaManager.cpp
@@ +546,5 @@
> +      mSourceStream->EndTrack(aTrackID);
> +      // We could override NotifyMediaStreamTrackEnded(), and maybe should, but it's
> +      // risky to do late in a release since that will affect all track ends, and not
> +      // just StopTrack()s.
> +      mListener->StopTrack(aTrackID, !!GetDOMTrackFor(aTrackID)->AsAudioStreamTrack());

GetDOMTrackFor may return nullptr which would cause a null dereference here.

http://mxr.mozilla.org/mozilla-central/source/content/media/DOMMediaStream.cpp#332 also has this comment:

  // We may add streams to our track list that are actually owned by
  // a different DOMMediaStream. Ignore those.

from which I infer that replaceTrack on b2g wont work more than once, i.e. it wont work to replace a track that's been replaced once already. Am I correct, and is that OK?

@@ +561,5 @@
> +    MOZ_ASSERT(aTrack->AsVideoStreamTrack() || aTrack->AsAudioStreamTrack());
> +    mListener->StopTrack(trackID, !!aTrack->AsAudioStreamTrack());
> +
> +    // forward to superclass
> +    DOMMediaStream::NotifyMediaStreamTrackEnded(aTrack);

Superclass is DOMLocalMediaStream

@@ +2328,5 @@
> +  {
> +    // XXX to support multiple tracks of a type in a stream, this should key off
> +    // the TrackID and not just the type
> +    nsRefPtr<MediaOperationRunnable> runnable(
> +      new MediaOperationRunnable(MEDIA_STOP_TRACK,

Is DOMMediaStream going to be sad without its last track?

I'm asking because the only other use of MEDIA_STOP_TRACK (which I know you added recently) seems to only send MEDIA_STOP_TRACK in the case where it knows there's another track left in the stream, otherwise it does MEDIA_STOP.

Comment 8

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

> ::: dom/media/MediaManager.cpp
> @@ +546,5 @@
> > +      mSourceStream->EndTrack(aTrackID);
> > +      // We could override NotifyMediaStreamTrackEnded(), and maybe should, but it's
> > +      // risky to do late in a release since that will affect all track ends, and not
> > +      // just StopTrack()s.
> > +      mListener->StopTrack(aTrackID, !!GetDOMTrackFor(aTrackID)->AsAudioStreamTrack());
> 
> GetDOMTrackFor may return nullptr which would cause a null dereference here.

Thanks.  Better safe than sorry.

> http://mxr.mozilla.org/mozilla-central/source/content/media/DOMMediaStream.
> cpp#332 also has this comment:
> 
>   // We may add streams to our track list that are actually owned by
>   // a different DOMMediaStream. Ignore those.
> 
> from which I infer that replaceTrack on b2g wont work more than once, i.e.
> it wont work to replace a track that's been replaced once already. Am I
> correct, and is that OK?

I probably can just remove that restriction... 

> 
> @@ +561,5 @@
> > +    MOZ_ASSERT(aTrack->AsVideoStreamTrack() || aTrack->AsAudioStreamTrack());
> > +    mListener->StopTrack(trackID, !!aTrack->AsAudioStreamTrack());
> > +
> > +    // forward to superclass
> > +    DOMMediaStream::NotifyMediaStreamTrackEnded(aTrack);
> 
> Superclass is DOMLocalMediaStream

Right

> 
> @@ +2328,5 @@
> > +  {
> > +    // XXX to support multiple tracks of a type in a stream, this should key off
> > +    // the TrackID and not just the type
> > +    nsRefPtr<MediaOperationRunnable> runnable(
> > +      new MediaOperationRunnable(MEDIA_STOP_TRACK,
> 
> Is DOMMediaStream going to be sad without its last track?
> 
> I'm asking because the only other use of MEDIA_STOP_TRACK (which I know you
> added recently) seems to only send MEDIA_STOP_TRACK in the case where it
> knows there's another track left in the stream, otherwise it does MEDIA_STOP.

In this case, it's possible someone would stop() a track attached to a peerconnection, then replaceTrack() it with a different one.  I was worried that Finishing the stream there would make that impossible, and I don't think it hurts.

Comment 9

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

Attachment: Stop the backend capture when the MSG track is stopped

Comment 10

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Comment on attachment 8479584 [details] [diff] [review]
Stop the backend capture when the MSG track is stopped

Review of attachment 8479584 [details] [diff] [review]:
-----------------------------------------------------------------

r=me with more parens.

::: dom/media/MediaManager.cpp
@@ +548,5 @@
> +      // risky to do late in a release since that will affect all track ends, and not
> +      // just StopTrack()s.
> +      if (GetDOMTrackFor(aTrackID)) {
> +        mListener->StopTrack(aTrackID, !!GetDOMTrackFor(aTrackID)->AsAudioStreamTrack());
> +      }

So StopTrack on an invalid track completes silently?

@@ +2325,5 @@
> +void
> +GetUserMediaCallbackMediaStreamListener::StopTrack(TrackID aID, bool aIsAudio)
> +{
> +  if (((aIsAudio && mAudioSource) ||
> +       (!aIsAudio && mVideoSource) && !mStopped)

Unless mStopped is only for video, you need more parens here.

Comment 11

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/a7e4ef74441d
https://hg.mozilla.org/integration/mozilla-inbound/rev/ef1c321a4a32
https://hg.mozilla.org/integration/mozilla-inbound/rev/0edf5afdc13c

Comment 12

[Carsten Book [:Tomcat]]

sorry had to backout this changes for test failures like https://tbpl.mozilla.org/php/getParsedLog.php?id=46831559&tree=Mozilla-Inbound burning inbound to hell :)

Comment 13

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

Attachment: interdiffs to fix TrackID problems with crashtests

Comment 14

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

Comment on attachment 8479881 [details] [diff] [review]
interdiffs to fix TrackID problems with crashtests

Slight tweak from pastebin version - searches for empty slot in StreamBuffer instead of assuming local max+1 is free

Comment 15

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/3e5111c421e1
https://hg.mozilla.org/integration/mozilla-inbound/rev/c7683d561e01
https://hg.mozilla.org/integration/mozilla-inbound/rev/3a0146f0b776

Tries:
https://tbpl.mozilla.org/?tree=Try&rev=8bb8fe636201 (M3/7/10/Crash)
https://tbpl.mozilla.org/?tree=Try&rev=49c84780ea2a (for M1)

Comment 16

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/3e5111c421e1
https://hg.mozilla.org/mozilla-central/rev/c7683d561e01
https://hg.mozilla.org/mozilla-central/rev/3a0146f0b776

Comment 17

[Eric Shepherd [:sheppy]]

Documented this a few months ago as part of updating other material.