WeakPtr destruction order cleanup: media edition.

media/cast/audio_receiver/audio_receiver.h

Index: media/cast/audio_receiver/audio_receiver.h
diff --git a/media/cast/audio_receiver/audio_receiver.h b/media/cast/audio_receiver/audio_receiver.h
index b2abfc2f4b04d86c10f8b309234b6370a0aa513f..1503cee7e5ff6351209a8cf1e3fe93d5a9db64de 100644
--- a/media/cast/audio_receiver/audio_receiver.h
+++ b/media/cast/audio_receiver/audio_receiver.h
@@ -115,8 +115,6 @@ class AudioReceiver : public base::NonThreadSafe,
   // Processes raw audio events to be sent over to the cast sender via RTCP.
   ReceiverRtcpEventSubscriber event_subscriber_;
 
-  base::WeakPtrFactory<AudioReceiver> weak_factory_;
-
   const transport::AudioCodec codec_;
   const int frequency_;
   base::TimeDelta target_delay_delta_;
@@ -140,6 +138,10 @@ class AudioReceiver : public base::NonThreadSafe,
   // it allows the event to be transmitted via RTCP.
   RtpTimestamp frame_id_to_rtp_timestamp_[256];
 
+  // NOTE: This must always be last so it's invalidated before other member
+  // variables might be accessed.

[Ami GONE FROM CHROMIUM, 2014/03/10 22:02:08]

Call me crazy but this comment is starting to seem repetitive :)
I'd much rather, for example, know which thread this factory was bound to...

[DaleCurtis, 2014/03/10 22:22:40]

Seemingly it's only repetitive because you're looking at a CL full of them :) 
Digging through every one of these changes to figure out the threading story
appears non-trivial at first glance.

If you're strongly opposed I can remove the text and we can hopefully rely on
the clang-plugin in the future.

+  base::WeakPtrFactory<AudioReceiver> weak_factory_;
+
   DISALLOW_COPY_AND_ASSIGN(AudioReceiver);
 };