media: Generalize MediaPermissionDispatcher.

content/renderer/render_frame_impl.h

Index: content/renderer/render_frame_impl.h
diff --git a/content/renderer/render_frame_impl.h b/content/renderer/render_frame_impl.h
index c3a6009c86359e7a2779a32b774b74494b6fd7d6..b247b78ea51385c04f60a7bcebc91a8bef7c8c10 100644
--- a/content/renderer/render_frame_impl.h
+++ b/content/renderer/render_frame_impl.h
@@ -887,6 +887,10 @@ class CONTENT_EXPORT RenderFrameImpl
 
   void RegisterMojoServices();
 
+  // Connect to an interface provided by the service registry.
+  template <typename Interface>
+  void ConnectToService(mojo::InterfaceRequest<Interface> request);
+
   // Connects to a Mojo application and returns a proxy to its exposed
   // ServiceProvider.
   mojo::ServiceProviderPtr ConnectToApplication(const GURL& url);
@@ -995,8 +999,8 @@ class CONTENT_EXPORT RenderFrameImpl
   // EncryptedMediaClient attached to this frame; lazily initialized.
   scoped_ptr<media::WebEncryptedMediaClientImpl> web_encrypted_media_client_;
 
-  // The media permission dispatcher attached to this frame, lazily initialized.
-  MediaPermissionDispatcher* media_permission_dispatcher_;
+  // The media permission dispatcher attached to this frame.
+  scoped_ptr<MediaPermissionDispatcher> media_permission_dispatcher_;

[xhwang, 2015/11/24 08:28:05]

We don't care about any Frame events. So using a scoped_ptr makes more sense. Do
I miss anything?

FWIW, I also want to use MediaPermissionDispatch in the GPU process where a
RenderFrameObserver doesn't make sense.

[nasko, 2015/11/25 15:04:27]

Makes sense to me. I now wonder why it was an observer to begin with :).

 
 #if defined(ENABLE_MOJO_MEDIA)
   // The media factory attached to this frame, lazily initialized.