Support start trimming post-decoding. Use it with FFmpegDemuxer.

media/formats/webm/webm_cluster_parser.cc

Index: media/formats/webm/webm_cluster_parser.cc
diff --git a/media/formats/webm/webm_cluster_parser.cc b/media/formats/webm/webm_cluster_parser.cc
index 3816fdb183442f17aca7e3e051eee193a92500d4..172eafafcf203d16309ff3a90d366a43222d2a71 100644
--- a/media/formats/webm/webm_cluster_parser.cc
+++ b/media/formats/webm/webm_cluster_parser.cc
@@ -416,8 +416,9 @@ bool WebMClusterParser::OnBlock(bool is_simple_block, int track_num,
   }
 
   if (discard_padding != 0) {
-    buffer->set_discard_padding(base::TimeDelta::FromMicroseconds(
-                                    discard_padding / 1000));
+    buffer->set_discard_padding(std::make_pair(
+        base::TimeDelta(),
+        base::TimeDelta::FromMicroseconds(discard_padding / 1000)));

[wolenetz, 2014/05/05 19:01:23]

Matroska allows for negative DiscardPadding to indicate start discard, and
positive for end discard.

Do any of our WebM-parsed formats possibly include negative DiscardPadding
(start discard)? If so, account for it here. If not, add DLOG(ERROR... or
CHECK(... ?

[DaleCurtis, 2014/05/05 20:16:03]

I didn't know that.  If WebM supports that, probably this code should be setting
start padding to -discard_padding then.  There'a a DCHECK in the
AudioDiscardHelper against negative values.

acolwell: WDYT?

   }
 
   return track->AddBuffer(buffer);