Adding new function ReadFrames() that returns the audio frame in planar

media/base/audio_buffer.h

Index: media/base/audio_buffer.h
diff --git a/media/base/audio_buffer.h b/media/base/audio_buffer.h
index a07985c9fd8ca9fb4f3545f429ee851c5cdad363..93aa70b2080b0c057c5b980d175314e117c5cc85 100644
--- a/media/base/audio_buffer.h
+++ b/media/base/audio_buffer.h
@@ -73,6 +73,11 @@ class MEDIA_EXPORT AudioBuffer
                   int dest_frame_offset,
                   AudioBus* dest);
 
+  // Copy frames into |std::vector<int32>|. The frames are converted from their
+  // source format into planar int32; note that while each frame use 32 bits
+  // each fame is actually PCM signed 24-bit little-endian.
+  std::vector<int32> ReadFrames();

[DaleCurtis, 2014/09/24 22:30:17]

We don't have a SampleFormat for planar 24 bit, only interleaved 32 bit. Which
do you actually need? Also, I was thinking this would be two methods like:

void ReadFramesInterleaved(int frames_to_copy, SampleFormat output_format, void*
dest);
void ReadFramesPlanar(int frames_to_copy, int plane_length, SampleFormat
output_format, void* dest);

Note, you'll need to be careful about adjusted_frame_count_ and trim_start_.
Implementing a generic method may be overkill if you only expect to output
24-bit PCM, so I'd be fine with something like:

void ReadFramesPlanarS24(int frames_to_copy, int plane_length, int32* dest);

If you always know your input sample format I'd be fine with CHECK_EQ()'ing that
and titling the method appropriately.

+
   // Trim an AudioBuffer by removing |frames_to_trim| frames from the start.
   // Timestamp and duration are adjusted to reflect the fewer frames.
   // Note that repeated calls to TrimStart() may result in timestamp() and