media: Support transporting 0-byte DecoderBuffer over mojo

media/mojo/interfaces/media_types.mojom

Index: media/mojo/interfaces/media_types.mojom
diff --git a/media/mojo/interfaces/media_types.mojom b/media/mojo/interfaces/media_types.mojom
index e9b679699b79dcfab73a787bcbe121091a0281fe..cd098adb5aa9452cea4616a62476ea260c96b0ba 100644
--- a/media/mojo/interfaces/media_types.mojom
+++ b/media/mojo/interfaces/media_types.mojom
@@ -104,8 +104,13 @@ struct DecoderBuffer {
   mojo.common.mojom.TimeDelta timestamp;
   mojo.common.mojom.TimeDelta duration;
 
-  // The number of bytes present in this buffer.  The data is not serialized
+  // Whether the buffer is an end-of-stream (EOS) buffer.
+  bool is_end_of_stream;

[alokp, 2016/11/08 21:34:08]

If "data_size == 0 && side_data_size == 0", would it be correct to conclude that
this is eos buffer?

[xhwang, 2016/11/08 22:13:51]

We could, but I feel it's really hacky and confusing. A dedicated flag here
seems more clear to me.

[alokp, 2016/11/08 22:17:30]

I agree. Should we add this flag to media::DecoderBuffer as well to keep things
consistent? "data_ == NULL" is equally hacky.

[xhwang, 2016/11/08 22:28:35]

Agreed. I tried refactor DecoderBuffer last week but it's more involved than
what I thought. This code is pretty old and needs some non-trivial work.

I added a comment in the bug:
https://bugs.chromium.org/p/chromium/issues/detail?id=663438#c2

+
+  // The number of bytes present in this buffer. The data is not serialized
   // along with this structure and must be read from a separate DataPipe.
+  // Note that |data_size| could be zero even for a non-EOS buffer (e.g.
+  // with non-empty |size_data|). See http://crbug.com/663438
   uint32 data_size;
 
   // Indicates whether or not this buffer is a random access point.