Upstream implementation for embedded video for WebMediaPlayerAndroid

webkit/media/android/webmediaplayer_manager_android.h

+// Copyright (c) 2012 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef WEBKIT_MEDIA_ANDROID_WEBMEDIAPLAYER_MANAGER_ANDROID_H_
+#define WEBKIT_MEDIA_ANDROID_WEBMEDIAPLAYER_MANAGER_ANDROID_H_
+
+#include <jni.h>
+#include <map>
+
+#include "base/basictypes.h"
+
+namespace webkit_media {
+
+class WebMediaPlayerAndroid;
+
+// Class for managing all the WebMediaPlayerAndroid objects in a renderer
+// process.
+class WebMediaPlayerManagerAndroid {
+ public:
+  WebMediaPlayerManagerAndroid();
+  virtual ~WebMediaPlayerManagerAndroid();

[scherkus (not reviewing), 2012/05/24 21:36:32]

why are all these methods virtual?

[qinmin, 2012/05/25 01:04:21]

removed all the virtual keyword
On 2012/05/24 21:36:32, scherkus wrote:

+
+  // Register and unregister a WebMediaPlayerAndroid object.
+  virtual int RegisterMediaPlayer(WebMediaPlayerAndroid* player);
+  virtual void UnRegisterMediaPlayer(int player_id);

[scherkus (not reviewing), 2012/05/24 21:36:32]

nit: "Un" isn't a word! let's have Unregister

[qinmin, 2012/05/25 01:04:21]

Done.

+
+  // Release all the media resources on the renderer process.
+  virtual void ReleaseMediaResources();
+
+  // Pass the video surface to a particular WebMediaPlayerAndroid object given
+  // by the player_id.
+  // TODO(qinmin): Upstream the code for passing Java surface object to the
+  // renderer thread
+  virtual void SetVideoSurface(jobject j_surface, int player_id);

[scherkus (not reviewing), 2012/05/24 21:36:32]

who calls this method?

I'm not sure how many more methods we're going to have like this, but does it
make more sense to have a method that returns a WMPA* given a player_id so that
we don't need to add forwarding methods?

Is it illegal to call this method with an invalid player_id?

Note that if you want to keep this method I'd suggest swapping the parameter
order (i.e., objects that are getting operated on [player_id] typically go first
followed by parameters to that object)

[qinmin, 2012/05/25 01:04:21]

When creating a video surface, the renderer process sends an IPC to the browser
process. After creating the video surface in the java side, the browser process
will pass the surface to the renderer process through binder
(http://developer.android.com/reference/android/os/Binder.html).  Surface is the
only object that we cannot pass through  chromium IPC. 
When the renderer process receives the surface through binder, it will call
setVideoSurface in our own RendererMainPlatformDelegate.

I changed this function to GetWebMediaPlayer() so that we only return a WMPA* to
the RendererMainPlatformDelegate. And then it can call setVideoSurface Directly

If the player_id does not exist (player got deleted before surface is created),
then nothing should happen.

On 2012/05/24 21:36:32, scherkus wrote:

+
+ private:
+  // Information needed to manage WebMediaPlayerAndroid.
+  // TODO(qinmin): more informations will be added here for resource management.
+  struct MediaPlayerInfo {
+    webkit_media::WebMediaPlayerAndroid* player;
+  };
+
+  // Info for all available WebMediaPlayerAndroid on a page; kept so that
+  // we can enumerate them to send updates about tab focus and visibily.
+  std::map<int32, MediaPlayerInfo> media_players_;
+
+  // ID used to create the next media player for passing the Surface.

[scherkus (not reviewing), 2012/05/24 21:36:32]

"for passing the Surface" doesn't read very well -- perhaps just remove this
comment -- the variable name is pretty obvious

[qinmin, 2012/05/25 01:04:21]

Done.

+  int32 next_media_player_id_;
+
+  DISALLOW_COPY_AND_ASSIGN(WebMediaPlayerManagerAndroid);
+};
+
+}  // namespace webkit_media
+
+#endif  // WEBKIT_MEDIA_ANDROID_WEBMEDIAPLAYER_MANAGER_ANDROID_H_