Changed mixing api and removed resampler

webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h

Index: webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h
diff --git a/webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h b/webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h
index 844e9ae0fdb67d7d209f141c33e891370d750148..6d7bb2aefbb07cddbba85417d2e05b544b7fc856 100644
--- a/webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h
+++ b/webrtc/modules/audio_mixer/include/new_audio_conference_mixer.h
@@ -48,9 +48,12 @@ class NewAudioConferenceMixer {
                                                bool mixable) = 0;
 
   // Performs mixing by asking registered audio sources for audio. The
-  // mixed result is placed in the provided AudioFrame. Can only be
-  // called from a single thread.
-  virtual void Mix(AudioFrame* audio_frame_for_mixing) = 0;

[aleloi, 2016/08/05 09:52:21]

We are moving away from the AudioFrame. In the future, we want to communicate
audio with simpler structures. The 'void*' is a step in that direction. We also
may want to mix floating point audio.

This will we be main mixer function for clients of the mixer. The mixer will do
necessary resampling.

[the sun, 2016/08/05 12:26:00]

Then why not float* instead of void*?

[aleloi, 2016/08/05 12:37:14]

To do one change at a time, probably. Much of the mixer and surrounding code
needs to be changed to mix in floats, I think. In what order do you think it's
simpler to do the changes: rewrite mixer and surroundings to mix in floats, or
plug in the new mixer? Curious.

+  // mixed result is placed in the provided buffer, which is
+  // guaranteed to be large enough. Can only be called from a single
+  // thread.
+  virtual void Mix(int sample_rate,
+                   size_t number_of_channels,
+                   void* audio_data) = 0;
 
   // Set the minimum sampling frequency at which to mix. The mixing algorithm
   // may still choose to mix at a higher samling frequency to avoid