RealtimeOutgoingAudioSource and RealtimeOutgoingVideoSource should be destroyed on the main thread

• Component: WebRTC
• Status: RESOLVED FIXED

[youenn fablet]

RealtimeOutgoingAudioSource and RealtimeOutgoingVideoSource should be destroyed on the main thread

[youenn fablet]

<rdar://problem/38214152>

[youenn fablet]

Created attachment 335358 [details]
Patch

[EWS Watchlist]

Attachment 335358 [details] did not pass style-queue:


ERROR: Source/WebCore/platform/mediastream/RealtimeOutgoingVideoSource.h:39:  Alphabetical sorting problem.  [build/include_order] [4]
Total errors found: 1 in 8 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[youenn fablet]

Created attachment 335359 [details]
Patch

[Chris Dumez]

Comment on attachment 335359 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=335359&action=review

> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:79
> +    RefPtr<Thread> m_thread;

Is this OK? Are we sure this does not create a cycle?

> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:84
> +    // Free senders in a different thread like could be done with actual peer connection.

So we can just free senders on any thread but not the current one?

> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:86
> +    auto threadFunction =  [this, keeper = keeper.copyRef(), senders = WTFMove(m_senders)] { };

Why is this capture this? Any use of this would likely crash.

[youenn fablet]

Comment on attachment 335359 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=335359&action=review

>> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:79
>> +    RefPtr<Thread> m_thread;
> 
> Is this OK? Are we sure this does not create a cycle?

It should not create a cycle, since the thread entry function is deleted after being called.
That said, looking at it further, there is probably no need for ThreadKeeper at all.

>> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:84
>> +    // Free senders in a different thread like could be done with actual peer connection.
> 
> So we can just free senders on any thread but not the current one?

Since it is a mock, this would always be freed in the main thread.
And we want to test the case of sources being derefed in a non main thread.

>> Source/WebCore/testing/MockLibWebRTCPeerConnection.cpp:86
>> +    auto threadFunction =  [this, keeper = keeper.copyRef(), senders = WTFMove(m_senders)] { };
> 
> Why is this capture this? Any use of this would likely crash.

Not needed indeed.

[youenn fablet]

Created attachment 335427 [details]
Patch

[Eric Carlson]

Comment on attachment 335427 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=335427&action=review

> Source/WebCore/testing/MockLibWebRTCPeerConnection.h:217
> +    bool SetTrack(webrtc::MediaStreamTrackInterface* track) final
> +    {
> +        m_track = track;
> +        return true;
> +    }

Is this related? A note in the ChangeLog about why this was changed would be helpful.

[youenn fablet]

Created attachment 335432 [details]
Patch for landing

[youenn fablet]

(In reply to Eric Carlson from comment #8)
> Comment on attachment 335427 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=335427&action=review
> 
> > Source/WebCore/testing/MockLibWebRTCPeerConnection.h:217
> > +    bool SetTrack(webrtc::MediaStreamTrackInterface* track) final
> > +    {
> > +        m_track = track;
> > +        return true;
> > +    }
> 
> Is this related? A note in the ChangeLog about why this was changed would be
> helpful.

Thanks for the review.
The track change ensures that we keep a reference if we use SetTrack so that it can be freed later on a wrong thread. This is not strictly needed but might help in the future.

[WebKit Commit Bot]

Comment on attachment 335432 [details]
Patch for landing

Clearing flags on attachment: 335432

Committed r229491: <https://trac.webkit.org/changeset/229491>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.