Improved readability and DCHECKing in AudioVector::[]

webrtc/modules/audio_coding/neteq/audio_vector.h

Index: webrtc/modules/audio_coding/neteq/audio_vector.h
diff --git a/webrtc/modules/audio_coding/neteq/audio_vector.h b/webrtc/modules/audio_coding/neteq/audio_vector.h
index 9100ff896893ebd43e53b7dbb4cfede2f10b48e5..57e01ab2e1290b64f2df40c3e7ac473690fb5cd5 100644
--- a/webrtc/modules/audio_coding/neteq/audio_vector.h
+++ b/webrtc/modules/audio_coding/neteq/audio_vector.h
@@ -127,9 +127,13 @@ class AudioVector {
   static inline size_t WrapIndex(size_t index,
                                  size_t begin_index,
                                  size_t capacity) {
-    RTC_DCHECK_GE(begin_index + index, index);  // Check for overflow.
-    const size_t ix =
-        begin_index + index - (begin_index + index >= capacity ? capacity : 0);
+    RTC_DCHECK_LT(index, capacity);
+    RTC_DCHECK_LT(begin_index, capacity);
+    size_t ix = begin_index + index;
+    RTC_DCHECK_GE(ix, index);  // Check for overflow.

[nisse-webrtc, 2017/02/22 07:48:40]

Given above DCHECKS, this could fail only if capacity > (maximum size_t) / 2.
Which is unlikely to happen (we'd run out of memory), but could be handled
correctly by changing the below condition to 

  if (ix < index || ix >= capacity)
    ix -= capacity;

[kwiberg-webrtc, 2017/02/22 08:25:48]

Aha. The two DCHECKs on line 130 and 131 guarantee that even if we do have
overflow, subtracting capacity will make everything work.

Doing what you suggest will however (1) cost extra at run-time (the ix < index
check) and (2) make the code harder to read, so given that it's very unlikely to
matter and that this piece of code is performance critical enough that we're
counting instructions for various alternatives, not acting on this clever idea
is probably the right thing to do.

+    if (ix >= capacity) {
+      ix -= capacity;
+    }
     RTC_DCHECK_LT(ix, capacity);
     return ix;
   }