Bug 206412

Summary:

[MSE] Decode glitches when watching videos on CNN.com

Product:

WebKit

Reporter:

Jer Noble <jer.noble>

Component:

New Bugs

Assignee:

Jer Noble <jer.noble>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

aboya, calvaris, commit-queue, eric.carlson, ews-watchlist, glenn, philipj, sergio

Priority:

Keywords:

InRadar

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	none
Patch for landing	none

Description Jer Noble 2020-01-17 06:41:42 PST

[MSE] Decode glitches when watching videos on CNN.com

Comment 1 Jer Noble 2020-01-17 06:42:49 PST

<rdar://problem/55685630>

Comment 2 Jer Noble 2020-01-17 06:56:55 PST

Created attachment 388034 [details]
Patch

Comment 3 Alicia Boya García 2020-01-17 08:08:47 PST

Comment on attachment 388034 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=388034&action=review

Looks good to me.

> Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1680
> +        // When appending media containing P-frames (media whose samples' presentation timestamps

I think you mean B-frames. A movie with only I-frames and P-frames has monotonic presentation timestamps, and usually PTS=DTS.

> Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1691
> +            auto nextDecodedSample = trackBuffer.samples.decodeOrder().findSampleAfterDecodeKey(decodeKey);

nextDecodeSample, nextSampleDTS, nextSampleInDecodeOrder or even just nextSample.

> Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1699
> +            if (nextSyncSample == nextDecodedSample)

The only way this could be true is if nextDecodeSample == end(), but that was checked already above.

Comment 4 Xabier Rodríguez Calvar 2020-01-17 08:13:40 PST

Comment on attachment 388034 [details]
Patch

Thanks Alicia!

Comment 5 Jer Noble 2020-01-17 09:11:33 PST

(In reply to Alicia Boya García from comment #3)
> Comment on attachment 388034 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=388034&action=review
> 
> Looks good to me.
> 
> > Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1680
> > +        // When appending media containing P-frames (media whose samples' presentation timestamps
> 
> I think you mean B-frames. A movie with only I-frames and P-frames has
> monotonic presentation timestamps, and usually PTS=DTS.

Whoops, looks like I got that... backwards.
> > Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1691
> > +            auto nextDecodedSample = trackBuffer.samples.decodeOrder().findSampleAfterDecodeKey(decodeKey);
> 
> nextDecodeSample, nextSampleDTS, nextSampleInDecodeOrder or even just
> nextSample.

As you can tell from SampleMap, I don’t mind long, descriptive names, so maybe I’ll go with “nextSampleInDecodeOrder”.

> > Source/WebCore/Modules/mediasource/SourceBuffer.cpp:1699
> > +            if (nextSyncSample == nextDecodedSample)
> 
> The only way this could be true is if nextDecodeSample == end(), but that
> was checked already above.

Good catch!

Comment 6 Jer Noble 2020-01-17 10:11:35 PST

Created attachment 388053 [details]
Patch for landing

Comment 7 WebKit Commit Bot 2020-01-17 13:03:59 PST

Comment on attachment 388053 [details]
Patch for landing

Clearing flags on attachment: 388053

Committed r254761: <https://trac.webkit.org/changeset/254761>

Comment 8 WebKit Commit Bot 2020-01-17 13:04:01 PST

All reviewed patches have been landed.  Closing bug.