Sets the media element network state to idle until playback is requested.

content/renderer/media/android/webmediaplayer_android.cc

Index: content/renderer/media/android/webmediaplayer_android.cc
diff --git a/content/renderer/media/android/webmediaplayer_android.cc b/content/renderer/media/android/webmediaplayer_android.cc
index 5adbee7b4e482fd0ca667aea2bbb0a37e1490495..6484f1f61689ae5f7532e63bd8c5cd680b26ec5d 100644
--- a/content/renderer/media/android/webmediaplayer_android.cc
+++ b/content/renderer/media/android/webmediaplayer_android.cc
@@ -294,6 +294,9 @@ void WebMediaPlayerAndroid::DidLoadMediaInfo(
     UpdateReadyState(WebMediaPlayer::ReadyStateHaveMetadata);
     UpdateReadyState(WebMediaPlayer::ReadyStateHaveEnoughData);
   }
+  // Android doesn't start fetching resources until an implementation-defined
+  // event (e.g. playback request) occurs. Sets to the network state to IDLE.
+  UpdateNetworkState(WebMediaPlayer::NetworkStateIdle);

[qinmin, 2013/09/03 14:22:46]

this is an optional step in HTML5 spec, do we really need this? and this should
be done before CORS  fetch

[Jinsuk Kim, 2013/09/04 01:32:03]

Though optional, I see benefit in having this - it makes Clank handle the media
loading state same as desktop Chrome does. When tested with preload attr -
default/none/metadata/auto, this CL ensures both get same sorts of events and
reach the same networkState.

What I understand from the code flow is that the state gets set to loading
initially, and then set to idle only when it reaches the step getting metadata.
In case it fails the state will remain loading and stalled event will be fired. 

Not sure where this should be moved to address your comment. It seems that CORS
info is fetched by MediaInfoLoader which is initialized in
WebMediaPlayerAndroid::load(), but it is also in this method that the initial
state is set to loading as well. Any more hint as to where it should go?

[qinmin, 2013/09/04 14:10:12]

can you run all the layout test with content shell and make sure there is no
regressions? and also add a layout test to test the code path if there is no
such one available
On 2013/09/04 01:32:03, jindor wrote:

[Jinsuk Kim, 2013/09/13 05:50:06]

Ran the tests and confirmed there's no regression. 

Let me add the test for the new codepath in a follow-up CL which will go under
WebKit/LayoutTest.

 }
 
 void WebMediaPlayerAndroid::play() {
@@ -312,6 +315,7 @@ void WebMediaPlayerAndroid::play() {
   if (paused())
     proxy_->Start(player_id_);
   UpdatePlayingState(true);
+  UpdateNetworkState(WebMediaPlayer::NetworkStateLoading);
 }
 
 void WebMediaPlayerAndroid::pause() {