media: Set MediaClientAndroid in the GPU process.

content/public/common/content_client.h

Index: content/public/common/content_client.h
diff --git a/content/public/common/content_client.h b/content/public/common/content_client.h
index 243afccf4b0b361d539dab0fbddc8c9c8a268c62..129b3832e095f6663ccf052245a38a6b1f7ca3c2 100644
--- a/content/public/common/content_client.h
+++ b/content/public/common/content_client.h
@@ -34,6 +34,10 @@ namespace gpu {
 struct GPUInfo;
 }
 
+namespace media {
+class MediaClientAndroid;
+}
+
 namespace sandbox {
 class TargetPolicy;
 }
@@ -163,6 +167,9 @@ class CONTENT_EXPORT ContentClient {
   // compositor. Note setting this to true will permit synchronous IPCs from
   // the browser UI thread.
   virtual bool UsingSynchronousCompositing();
+
+  // Returns the MediaClientAndroid to be used by media code on Android.
+  virtual media::MediaClientAndroid* GetMediaClientAndroid();

[Lei Zhang, 2016/04/26 00:02:21]

Can this go into ContentGpuClient instead?

[xhwang, 2016/04/26 03:25:45]

Good question. Please see the CL description:

"Today in the browser, we call media::SetMediaClientAndroid() directly
from content embedders (e.g. ChromeBrowserMainPartsAndroid), which
seems hacky. We should use ContentClient::GetMediaClientAndroid() in
the browser process as well so that content will talk to the embedder,
and then talk to media. I'll work on it after this CL."

So this code will be called in both the browser process and gpu process.
Actually, in M53-ish we'll deprecate the browser part, at that point I should
be able to move this code to the GpuClient. Do you want me to do that
now and keep the browser code as is?

 #endif  // OS_ANDROID
 
  private: