Remove HTMLMediaElement::textTrackKindChanged

Source/core/html/track/TextTrack.cpp

Index: Source/core/html/track/TextTrack.cpp
diff --git a/Source/core/html/track/TextTrack.cpp b/Source/core/html/track/TextTrack.cpp
index 54f2cbdec176b2a2c6a4aefe59f6fb5adc5abd34..ceebcb4cd483321433f0bb44359b21587593c497 100644
--- a/Source/core/html/track/TextTrack.cpp
+++ b/Source/core/html/track/TextTrack.cpp
@@ -156,8 +156,11 @@ void TextTrack::setKind(const AtomicString& newKind)
     AtomicString oldKind = kind();
     TrackBase::setKind(newKind);
 
-    if (mediaElement() && oldKind != kind())
-        mediaElement()->textTrackKindChanged(this);
+    // If kind changes from visual to non-visual and mode is 'showing', then force mode to 'hidden'.

[philipj_slow, 2015/02/23 13:41:46]

Does this correspond to anything in the HTML spec? It seems reasonable, but
there's nothing equivalent when setting mode. I think this means that the follow
two bits of code won't result in the same state, if we start with kind
'captions' and mode 'hidden':

track.kind = 'metadata';
track.mode = 'showing';
// mode is showing

track.mode = 'showing';
track.kind = 'metadata';
// mode is hidden

That seems a bit odd.

[fs, 2015/02/23 14:00:02]

AFAIK, no. The spec says "The kind of track can change dynamically, in the case
of a text track corresponding to a track element." but doesn't prescribe any
specific handling for that case.
Odd indeed. I suspect what should really happen here (in TextTrack::setKind) is
that any currently rendered cues gets hidden.

+    if (oldKind != kind() && mode() == showingKeyword()) {
+        if (kind() != captionsKeyword() && kind() != subtitlesKeyword())
+            setMode(hiddenKeyword());
+    }
 }
 
 void TextTrack::setMode(const AtomicString& mode)