Bug 231286 – getDisplayMedia MediaStreamTrack.applyConstraints behavior regressed in Safari 15

• Component: WebRTC
• Status: RESOLVED FIXED

[Arjun Vade]

Created attachment 440353 [details]
issue_illustrator

A valid applyConstraints on a MediaStreamTrack is now throwing OverconstrainedError. This used to work in earlier versions of safari and has regressed only with Safari 15. Attaching a sample html page that illustrates the issue.

Steps to reproduce:
1. Open the html page attached in safari
2. Click on the start button
3. Notice that video 1 works, while video 2 doesn't. Open the dev tools console and notice the OverconstrainedError.

[Radar WebKit Bug Importer]

<rdar://problem/83930865>

[youenn fablet]

Created attachment 440360 [details]
Patch

[Eric Carlson]

Comment on attachment 440360 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=440360&action=review

It would be good to include tests the specify only width, only height, and neither.

> Source/WebCore/platform/mediastream/RealtimeVideoCaptureSource.cpp:261
> +        if (!requestedWidth &&  requestedHeight && size().isEmpty())
> +            return { };

Should this be `!requestedHeight`?

> Source/WebCore/platform/mediastream/RealtimeVideoCaptureSource.cpp:265
> +        if (!requestedWidth) {
> +            captureSize.setWidth(*requestedWidth);

This will crash, shouldn't it be `if (requestedWidth)`?

> Source/WebCore/platform/mediastream/RealtimeVideoCaptureSource.cpp:268
> +            captureSize.setHeight(*requestedHeight);

Can't requestedHeight be null here?

[youenn fablet]

Created attachment 440629 [details]
Patch

[Eric Carlson]

Comment on attachment 440629 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=440629&action=review

> Source/WebKit/WebProcess/cocoa/RemoteRealtimeAudioSource.cpp:49
> +    fprintf(stderr, "RemoteRealtimeAudioSource::create %d\n", (int)source->identifier().toUInt64());

Oops!

> Source/WebKit/WebProcess/cocoa/RemoteRealtimeDisplaySource.cpp:44
> +    fprintf(stderr, "RemoteRealtimeDisplaySource::create %d\n", (int)source->identifier().toUInt64());

Ditto

> Source/WebKit/WebProcess/cocoa/RemoteRealtimeVideoSource.cpp:51
> +    fprintf(stderr, "RemoteRealtimeVideoSource::create %d\n", (int)source->identifier().toUInt64());

Ditto

[youenn fablet]

Created attachment 440761 [details]
Patch

[youenn fablet]

Created attachment 441600 [details]
Patch

[youenn fablet]

Created attachment 441695 [details]
Patch

[EWS]

Committed r284444 (243206@main): <https://commits.webkit.org/243206@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 441695 [details].

[Philippe Normand]

Comment on attachment 441695 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=441695&action=review

> LayoutTests/fast/mediastream/getDisplayMedia-size-expected.txt:6
> +FAIL original source should stay at a small size assert_equals: expected 640 but got 320
> +FAIL original source should stay at a big size assert_equals: expected 640 but got 1280

This is a bit confusing. Should these tests be reworked to have PASS expectations?