Added various timestamps to FrameObject.

webrtc/modules/video_coding/frame_object.cc

Index: webrtc/modules/video_coding/frame_object.cc
diff --git a/webrtc/modules/video_coding/frame_object.cc b/webrtc/modules/video_coding/frame_object.cc
index c33fdf749b81b908fab9c29497c4d61308692c44..c99e7f56e291b85626b43a29d031a598960f67f2 100644
--- a/webrtc/modules/video_coding/frame_object.cc
+++ b/webrtc/modules/video_coding/frame_object.cc
@@ -26,10 +26,12 @@ RtpFrameObject::RtpFrameObject(PacketBuffer* packet_buffer,
                                uint16_t first_seq_num,
                                uint16_t last_seq_num,
                                size_t frame_size,
-                               int times_nacked)
+                               int times_nacked,
+                               int64_t received_timestamp)

[terelius, 2016/07/07 16:12:30]

If I understand the intended usage correctly, I would prefer calling this
variable frame_completed_timestamp.

[philipel, 2016/07/08 09:41:19]

This timestamp indicates when the last packet of the frame was received, so I
think |received_timestamp| is correct.

[stefan-webrtc, 2016/07/08 10:33:39]

Should it then be called last_packet_time or something? Looking at the  code
where this object  is constructed, it does look more like it's the time when the
frame was complete, as it may not necessarily be the last packet if there is
reordering.

[philipel, 2016/07/08 11:10:50]

Hmmm... naming here is kind of hard. I have chosen to not call a frame complete
until all its packets has been received AND all its references has been found,
so a frame that only has all its packets but not its references found is not a
complete frame.

Maybe it is better to change the convention so that a frame which has all its
packets is called complete, and after finding its references something else...

But just to be clear, I want to store when the last packet of this frame was
received regardless of which order they were received.

[stefan-webrtc, 2016/07/08 11:34:23]

Yes, it's a good idea to try to be clear. In the old JB we use "continuous" to
say that a frame has all its references.

[stefan-webrtc, 2016/07/08 11:35:27]

BTW, might be a good idea to add a comment about the chosen convention in the
header file.

[philipel, 2016/07/08 13:40:02]

Maybe this convention would work:

"received frame"
A frame which has received all its packets

"complete frame"
A frame which has received all its packets and all its references are known.

"continuous frame"
A frame which have all its packets, all its references known and all its  
referenced frames decoded.

     : packet_buffer_(packet_buffer),
       first_seq_num_(first_seq_num),
       last_seq_num_(last_seq_num),
+      received_timestamp_(received_timestamp),
       times_nacked_(times_nacked) {
   size = frame_size;
   VCMPacket* packet = packet_buffer_->GetPacket(first_seq_num);
@@ -84,6 +86,18 @@ bool RtpFrameObject::GetBitstream(uint8_t* destination) const {
   return packet_buffer_->GetBitstream(*this, destination);
 }
 
+uint32_t RtpFrameObject::Timestamp() const {
+  return timestamp_;
+}
+
+int64_t RtpFrameObject::ReceivedTimestamp() const {
+  return received_timestamp_;
+}
+
+int64_t RtpFrameObject::RenderTimestamp() const {
+  return _renderTimeMs;
+}
+
 RTPVideoTypeHeader* RtpFrameObject::GetCodecHeader() const {
   VCMPacket* packet = packet_buffer_->GetPacket(first_seq_num_);
   if (!packet)