Implement the NackModule as part of the new jitter buffer.

webrtc/modules/video_coding/include/video_coding_defines.h

Index: webrtc/modules/video_coding/include/video_coding_defines.h
diff --git a/webrtc/modules/video_coding/include/video_coding_defines.h b/webrtc/modules/video_coding/include/video_coding_defines.h
index 7d084c815b8f2ea2a92215bfc0a4394210783688..fc0f21cc1153b4aa32949458683db4dd2deeca60 100644
--- a/webrtc/modules/video_coding/include/video_coding_defines.h
+++ b/webrtc/modules/video_coding/include/video_coding_defines.h
@@ -160,6 +160,8 @@ class VCMFrameTypeCallback {
 // Callback class used for telling the user about which packet sequence numbers
 // are currently
 // missing and need to be resent.
+// TODO(philipel): Deprecate VCMPacketRequestCallback
+//                 and use NackSender instead.
 class VCMPacketRequestCallback {
  public:
   virtual int32_t ResendPackets(const uint16_t* sequenceNumbers,
@@ -169,6 +171,23 @@ class VCMPacketRequestCallback {
   virtual ~VCMPacketRequestCallback() {}
 };
 
+class NackSender {
+ public:
+  virtual int32_t SendNack(const uint16_t* sequence_numbers,
+                           uint16_t length) = 0;

[stefan-webrtc, 2016/03/01 12:16:34]

length should probably be size_t. Maybe we could even pass in a const
vector<uint16_t>& instead?

[philipel, 2016/03/01 13:31:08]

Agree, the vector is much cleaner, fixed.

+
+ protected:
+  virtual ~NackSender() {}
+};
+
+class KeyFrameRequestSender {
+ public:
+  virtual int32_t RequestKeyFrame() = 0;

[stefan-webrtc, 2016/03/01 12:16:34]

Return bool instead, or maybe even void? Same for SendNack.

[philipel, 2016/03/01 13:31:08]

Done.

+
+ protected:
+  virtual ~KeyFrameRequestSender() {}
+};
+
 // Callback used to inform the user of the the desired resolution
 // as subscribed by Media Optimization (Quality Modes)
 class VCMQMSettingsCallback {