Vanquish the remnants of media::MessageLoopFactory.

media/filters/gpu_video_decoder.h

Index: media/filters/gpu_video_decoder.h
diff --git a/media/filters/gpu_video_decoder.h b/media/filters/gpu_video_decoder.h
index 6c1fec3766022f0718a753f0772de01b6b295862..ae51189af890b75d976813e125a8c0f64c8ee209 100644
--- a/media/filters/gpu_video_decoder.h
+++ b/media/filters/gpu_video_decoder.h
@@ -55,13 +55,15 @@ class MEDIA_EXPORT GpuVideoDecoder
     // Close()ing the returned pointer.
     virtual base::SharedMemory* CreateSharedMemory(size_t size) = 0;
 
+    // Returns the message loop the VideoDecodeAccelerator runs on.
+    virtual scoped_refptr<base::MessageLoopProxy> GetMessageLoop() = 0;

[scherkus (not reviewing), 2012/12/07 23:54:11]

this is because we now create GpuVideoDecoder in webkit/ as opposed to content/
and can't access the proper loop

alternatively, we can create the thread + GpuVideoDecoder in content/ and
plumbing them into WMPI() (it better reflects the end goal of passing a
MessageLoopProxy of the shared media thread into WMPI) -- it gets a bit hairy
though with OverrideWMPI()

acowell: any thoughts / preferences?

+
    protected:
     friend class base::RefCountedThreadSafe<Factories>;
     virtual ~Factories();
   };
 
-  GpuVideoDecoder(const scoped_refptr<base::MessageLoopProxy>& gvd_loop_proxy,
-                  const scoped_refptr<base::MessageLoopProxy>& vda_loop_proxy,
+  GpuVideoDecoder(const scoped_refptr<base::MessageLoopProxy>& message_loop,
                   const scoped_refptr<Factories>& factories);
 
   // VideoDecoder implementation.