Miscellaneous changes split from https://codereview.webrtc.org/1230503003 .

webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc

Index: webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc
diff --git a/webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc b/webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc
index 99ff95a2ec1378d65bcdaa0bfa383e9f29697961..6a9b953f23a3a9d9db5f683a3820afebb9b49a2e 100644
--- a/webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc
+++ b/webrtc/modules/audio_coding/neteq/audio_decoder_impl.cc
@@ -256,16 +256,14 @@ int AudioDecoderG722Stereo::DecodeInternal(const uint8_t* encoded,
                             static_cast<int16_t>(encoded_len / 2),
                             &decoded[decoded_len], &temp_type);
     if (ret == decoded_len) {
-      decoded_len += ret;
+      ret += decoded_len;  // Return total number of samples.
       // Interleave output.
-      for (int k = decoded_len / 2; k < decoded_len; k++) {
+      for (int k = ret / 2; k < ret; k++) {
           int16_t temp = decoded[k];
-          memmove(&decoded[2 * k - decoded_len + 2],
-                  &decoded[2 * k - decoded_len + 1],
-                  (decoded_len - k - 1) * sizeof(int16_t));
-          decoded[2 * k - decoded_len + 1] = temp;
+          memmove(&decoded[2 * k - ret + 2], &decoded[2 * k - ret + 1],
+                  (ret - k - 1) * sizeof(int16_t));
+          decoded[2 * k - ret + 1] = temp;
       }

[kwiberg-webrtc, 2015/07/12 18:40:30]

Hmm. Maybe better, but still messy. How about using const int16_t ret_left for
the first call, const int16_t ret_right for the second call, and eliminate
decoded_len altogether?

[Peter Kasting, 2015/07/13 02:41:24]

I don't think that makes life better, due to the way the code wants to clean up
below this and then return |ret|, which will take on different values depending
on the result of the comparison.

I'm also going to be modifying this further, since the underlying function will
start returning a size_t and thus the ">= 0" check will disappear.  I would
suggest going with this as an improvement for now and then taking a look when I
touch it again at whether it's been made as good as possible or can still be
improved further.

-      ret = decoded_len;  // Return total number of samples.
     }
   }
   *speech_type = ConvertSpeechType(temp_type);