Bug 116817 – [GStreamer] support preload="metadata"

• Component: Media
• Status: RESOLVED FIXED

[Balazs Kelemen]

Add appropriate behavior.

[Balazs Kelemen]

Created attachment 202964 [details]
Patch

[Philippe Normand]

Can it be rebased against ToT please? Would be nice to probe the new test on mac.

[Balazs Kelemen]

(In reply to comment #2)
> Can it be rebased against ToT please? Would be nice to probe the new test on mac.

It has a dependency. I will reupload after bug 116642 is resolved.

[Philippe Normand]

So this patch can be rebased now? :)

[Balazs Kelemen]

(In reply to comment #4)
> So this patch can be rebased now? :)

Ah, sure, I was forgetting my promise :)

[Balazs Kelemen]

Created attachment 205588 [details]
Patch

[Philippe Normand]

Comment on attachment 205588 [details]
Patch

No test results?

[Balazs Kelemen]

Created attachment 205607 [details]
Patch

[Balazs Kelemen]

(In reply to comment #7)
> (From update of attachment 205588 [details])
> No test results?

Fixed, sorry...

[Philippe Normand]

Comment on attachment 205607 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=205607&action=review

Looks good, I only have a couple of questions about the test

> LayoutTests/media/video-load-preload-metadata.html:22
> +                setTimeout(function () {

Why doing this with a timer?

> LayoutTests/media/video-load-preload-metadata.html:24
> +                    playbackStarted = true;

Perhaps this can be set when the play event is received.

[Eric Carlson]

Comment on attachment 205607 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=205607&action=review

> LayoutTests/media/video-load-preload-metadata.html:17
> +            function start()
> +            {
> +                findMediaElement();
> +                video.src = findMediaFile("video", "content/test");
> +
> +                testExpected("video.preload", "metadata");
> +                testExpected("video.readyState", HTMLMediaElement.HAVE_NOTHING);
> +                waitForEvent('loadedmetadata', onLoadedMetadata);
> +                waitForEvent('canplaythrough', onCanPlayThrough);
> +                failTestIn(3000);
> +            }

I am afraid this test will be flakey because there is no guarantee that 'load' will fire before the media engine starts loading data. 

If you want to test that readyState is HAVE_NOTHING, I think the only way to be sure is to have "preload=none" on the element and set preload to "metadata" in the load handler.

>> LayoutTests/media/video-load-preload-metadata.html:22
>> +                setTimeout(function () {
> 
> Why doing this with a timer?

Agreed, no timers unless absolutely necessary please.

[Build Bot]

Comment on attachment 205607 [details]
Patch

Attachment 205607 [details] did not pass win-ews (win):
Output: http://webkit-queues.appspot.com/results/965922

New failing tests:
media/video-load-preload-metadata.html

[Build Bot]

Created attachment 205671 [details]
Archive of layout-test-results from APPLE-EWS-3 for win-future

The attached test failures were seen while running run-webkit-tests on the win-ews.
Bot: APPLE-EWS-3  Port: win-future  Platform: CYGWIN_NT-6.1-WOW64-1.7.20-0.266-5-3-i686-32bit

[Balazs Kelemen]

(In reply to comment #11)
> (From update of attachment 205607 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=205607&action=review
> 
> > LayoutTests/media/video-load-preload-metadata.html:17
> > +            function start()
> > +            {
> > +                findMediaElement();
> > +                video.src = findMediaFile("video", "content/test");
> > +
> > +                testExpected("video.preload", "metadata");
> > +                testExpected("video.readyState", HTMLMediaElement.HAVE_NOTHING);
> > +                waitForEvent('loadedmetadata', onLoadedMetadata);
> > +                waitForEvent('canplaythrough', onCanPlayThrough);
> > +                failTestIn(3000);
> > +            }
> 
> I am afraid this test will be flakey because there is no guarantee that 'load' will fire before the media engine starts loading data. 
> 
> If you want to test that readyState is HAVE_NOTHING, I think the only way to be sure is to have "preload=none" on the element and set preload to "metadata" in the load handler.
> 

HAVE_NOTHING is not so important in this test, the important part is |testExpected("video.readyState", HTMLMediaElement.HAVE_ENOUGH_DATA, "<")| before playback is started. I will remove HAVE_NOTHING check.

[Balazs Kelemen]

(In reply to comment #10)
> (From update of attachment 205607 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=205607&action=review
> 
> Looks good, I only have a couple of questions about the test
> 
> > LayoutTests/media/video-load-preload-metadata.html:22
> > +                setTimeout(function () {
> 
> Why doing this with a timer?

Without the timer we can't be sure that canplaythrough not fires too early.
Basically the problem is that there is no way to test that smg _not_ fires without a timer.

> 
> > LayoutTests/media/video-load-preload-metadata.html:24
> > +                    playbackStarted = true;
> 
> Perhaps this can be set when the play event is received.

We need it in the canPlayThrough callback and canplaythrough fires earlier than play.

If you have a good idea how to avoid the timer without loosing testing completeness I'm open to that, but currently I would rather keep it.

[Balazs Kelemen]

Created attachment 205999 [details]
Patch

[Balazs Kelemen]

(In reply to comment #14)
> (In reply to comment #11)
> > (From update of attachment 205607 [details] [details])
> > View in context: https://bugs.webkit.org/attachment.cgi?id=205607&action=review
> > 
> > > LayoutTests/media/video-load-preload-metadata.html:17
> > > +            function start()
> > > +            {
> > > +                findMediaElement();
> > > +                video.src = findMediaFile("video", "content/test");
> > > +
> > > +                testExpected("video.preload", "metadata");
> > > +                testExpected("video.readyState", HTMLMediaElement.HAVE_NOTHING);
> > > +                waitForEvent('loadedmetadata', onLoadedMetadata);
> > > +                waitForEvent('canplaythrough', onCanPlayThrough);
> > > +                failTestIn(3000);
> > > +            }
> > 
> > I am afraid this test will be flakey because there is no guarantee that 'load' will fire before the media engine starts loading data. 
> > 
> > If you want to test that readyState is HAVE_NOTHING, I think the only way to be sure is to have "preload=none" on the element and set preload to "metadata" in the load handler.
> > 
> 
> HAVE_NOTHING is not so important in this test, the important part is |testExpected("video.readyState", HTMLMediaElement.HAVE_ENOUGH_DATA, "<")| before playback is started. I will remove HAVE_NOTHING check.

I removed this, otherwise it's the same patch.

[Eric Carlson]

Comment on attachment 205999 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=205999&action=review

> Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:1523
> +    bool shouldDownload = !isLiveStream() && m_preload == MediaPlayer::Auto;
>      if (shouldDownload) {

Do you really want to clear the GST_PLAY_FLAG_DOWNLOAD flag when m_preload changes from "auto" to "metadata" or "none" (eg. after downloading has already started)?

[Balazs Kelemen]

(In reply to comment #18)
> (From update of attachment 205999 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=205999&action=review
> 
> > Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:1523
> > +    bool shouldDownload = !isLiveStream() && m_preload == MediaPlayer::Auto;
> >      if (shouldDownload) {
> 
> Do you really want to clear the GST_PLAY_FLAG_DOWNLOAD flag when m_preload changes from "auto" to "metadata" or "none" (eg. after downloading has already started)?

Good point. It needs a bit of additional logic so I think it's better to do one more round.

[Balazs Kelemen]

Created attachment 206028 [details]
Patch

[Balazs Kelemen]

Comment on attachment 206028 [details]
Patch

Clearing flags on attachment: 206028

Committed r152391: <http://trac.webkit.org/changeset/152391>

[Balazs Kelemen]

All reviewed patches have been landed.  Closing bug.