Allow mocking SendSideCongestionController for Call tests.

webrtc/call/rtp_transport_controller_send.h

 /*
  *  Copyright (c) 2017 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 #ifndef WEBRTC_CALL_RTP_TRANSPORT_CONTROLLER_SEND_H_
 #define WEBRTC_CALL_RTP_TRANSPORT_CONTROLLER_SEND_H_
 
 #include "webrtc/base/constructormagic.h"
 #include "webrtc/call/rtp_transport_controller_send_interface.h"
 #include "webrtc/modules/congestion_controller/include/send_side_congestion_controller.h"
 
 namespace webrtc {
 class Clock;
 class RtcEventLog;
 
 // TODO(nisse): When we get the underlying transports here, we should
 // have one object implementing RtpTransportControllerSendInterface
 // per transport, sharing the same congestion controller.
 class RtpTransportControllerSend : public RtpTransportControllerSendInterface {
  public:
   RtpTransportControllerSend(Clock* clock, webrtc::RtcEventLog* event_log);
 
-  void RegisterNetworkObserver(
-      SendSideCongestionController::Observer* observer);

[Stefan, 2017/05/01 11:17:06]

I think we have to decide if we want send_side_cc to be exposed, or if we should
only expose the necessary methods on RtpTransportControllerSend? This is a move
towards the former. As I see it, the RtpTransportSenders will talk to the
CongestionController by queueing packets to be sent etc, so the user of
RtpTransportControllerSend shouldn't have to know about it. That may be an
argument for not exposing it at all from RtpTransportControllerSend?

Nisse, what's your opinion?

[nisse-webrtc, 2017/05/02 07:01:29]

It's not clear to me how this will look in the end, but Call implements this
observer interface.

Also note that this is the implementation class. For the corresponding interface
(intended for use by send streams), it currently has a send_side_cc accessor,
which I think we should aim to delete.

[Zach Stein, 2017/05/04 21:30:02]

RegisterNetworkObserver would need to be added to
RtpTransportControllerSendInterface if send_side_cc was removed for this change
to work (Call's constructor takes the interface and currently registers the
network observer by accessing the send_side_cc). Should I go ahead and do that?

[nisse-webrtc, 2017/05/05 07:12:38]

No, I think it's better to leave send_side_cc() in the interface for the time
being.

[Zach Stein, 2017/05/05 23:28:12]

Acknowledged.

-
   // Implements RtpTransportControllerSendInterface
   PacketRouter* packet_router() override { return &packet_router_; }
   SendSideCongestionController* send_side_cc() override {
     return &send_side_cc_;
   }
   TransportFeedbackObserver* transport_feedback_observer() override {
     return &send_side_cc_;
   }
   RtpPacketSender* packet_sender() override { return send_side_cc_.pacer(); }
 
  private:
   PacketRouter packet_router_;
   SendSideCongestionController send_side_cc_;
 
   RTC_DISALLOW_COPY_AND_ASSIGN(RtpTransportControllerSend);
 };
 
 }  // namespace webrtc
 
 #endif  // WEBRTC_CALL_RTP_TRANSPORT_CONTROLLER_SEND_H_