Build WebRTC with data channel only.

webrtc/call/call.h

Index: webrtc/call/call.h
diff --git a/webrtc/call/call.h b/webrtc/call/call.h
index caf0ee2d881951e5c28320086b6cf74c2c48ef92..75a9807af0f0ca414faf0e0dad7fd778863c455e 100644
--- a/webrtc/call/call.h
+++ b/webrtc/call/call.h
@@ -60,10 +60,16 @@ class Call {
  public:
   struct Config {
     explicit Config(RtcEventLog* event_log) : event_log(event_log) {
+#ifdef HAVE_MEDIA
       RTC_DCHECK(event_log);
+#endif

[pthatcher1, 2017/05/03 18:05:53]

Would it be easier to just not include this file if HAVE_MEDIA is false?

[Zhi Huang, 2017/05/04 01:08:02]

This is the base class of internal::Call and this header is part of the
"call_interface" target. The MediaController and WebRtcSession etc. are
depending on it. If removing this, there would be more ifdef in those files.

     }
 
+#ifdef HAVE_MEDIA
     static const int kDefaultStartBitrateBps;
+#else
+    static const int kDefaultStartBitrateBps = -1;

[pthatcher1, 2017/05/03 18:05:53]

It's confusing that we have an ifdef in two places for this (one in the .h and
one in the .cc).  Can you just put both one the .cc?

[Zhi Huang, 2017/05/04 01:08:02]

This is because call.h is an abstract class and it is part of the
"call_interface" build target while the call.cc is part of "call" target. When
building WebRTC without media, we don't build the "call" target so I have to
give the const member a value to avoid link errors. I can move both into .h file
but not sure if it is better to set this value in the abstract class.

+#endif
 
     // Bitrate config used until valid bitrate estimates are calculated. Also
     // used to cap total bitrate used.