Split VTTCue from TextTrackCue

Split VTTCue from TextTrackCue

VTTCue has been split out from TextTrackCue in the specs:
http://whatwg.org/html#texttrackcue
http://dev.w3.org/html5/webvtt/#vttcue-interface

The Opera tests were fixed upstream and re-imported:
https://github.com/w3c/web-platform-tests/commit/8c08468d5494c9026c2a857f366c631651c413fd
https://github.com/w3c/web-platform-tests/commit/da34397c2c16170da60e5c7638e00c0edadcf0be

Don't import the async_test from cues.html since it requires a more
recent testharness.js than we have and wouldn't pass anyway:
https://github.com/w3c/web-platform-tests/pull/435

The TextTrackCue constructor was removed to verify that the tests
work as expected without it. A follow-up commit will bring it back
as a temporary alias for VTTCue, to isolate everything that should
eventually be removed into a single easy-to-revert package.

Internal refactoring has been kept to an absolute minimum, future
commits should move more code from TextTrackCue to VTTCue.

BUG=270340

[philipj_slow, 2013-11-12 14:44:05]

Part 2 to extend the lifespan of the TextTrackCue constructor is still WIP, but
I think the review of this can begin first.

https://codereview.chromium.org/69993003/diff/1/Source/core/html/track/TextTr...
File Source/core/html/track/TextTrackCue.h (left):

https://codereview.chromium.org/69993003/diff/1/Source/core/html/track/TextTr...
Source/core/html/track/TextTrackCue.h:79: UseCounter::count(document,
UseCounter::TextTrackCueConstructor);
The attentive reader will notice that this was wrong, since it counted not only
script-created cues but also those created by the WebVTT parser and from in-band
tracks. Oops.

[gasubic, 2013-11-12 20:48:43]

On 2013/11/12 14:44:05, philipj wrote:
> Part 2 to extend the lifespan of the TextTrackCue constructor is still WIP,
but
> I think the review of this can begin first.
> 
>
https://codereview.chromium.org/69993003/diff/1/Source/core/html/track/TextTr...
> File Source/core/html/track/TextTrackCue.h (left):
> 
>
https://codereview.chromium.org/69993003/diff/1/Source/core/html/track/TextTr...
> Source/core/html/track/TextTrackCue.h:79: UseCounter::count(document,
> UseCounter::TextTrackCueConstructor);
> The attentive reader will notice that this was wrong, since it counted not
only
> script-created cues but also those created by the WebVTT parser and from
in-band
> tracks. Oops.

My original thinking was to implement this change without changing any of the
tests to prove that there was full backward compatibility. And then implement
the test changes as a subsequent patch after being satisfied on backward
compatibility. You might want to consider doing it that way.

[philipj_slow, 2013-11-12 22:24:27]

On 2013/11/12 20:48:43, gasubic wrote:

> My original thinking was to implement this change without changing any of the
> tests to prove that there was full backward compatibility. And then implement
> the test changes as a subsequent patch after being satisfied on backward
> compatibility. You might want to consider doing it that way.

I can see the value in that, if the reviewers would prefer that I could split it
that way. The downside is that the TextTrackCue constructor hack would be part
of the huge patch instead of in a single revert-friendly commit.

[philipj_slow, 2013-11-13 12:42:42]

Part 2: https://codereview.chromium.org/59623008/

[philipj_slow, 2013-11-14 08:53:26]

Closing in favor of https://codereview.chromium.org/63173020/