Fix for integer overflow in NetEq.

webrtc/modules/audio_coding/neteq/expand.cc

Index: webrtc/modules/audio_coding/neteq/expand.cc
diff --git a/webrtc/modules/audio_coding/neteq/expand.cc b/webrtc/modules/audio_coding/neteq/expand.cc
index ffe4370e4e61e8766c50ce8efcf22c96bd9723bf..b640f268650e9f24189f0e6b43651b768f5c18dc 100644
--- a/webrtc/modules/audio_coding/neteq/expand.cc
+++ b/webrtc/modules/audio_coding/neteq/expand.cc
@@ -675,12 +675,13 @@ void Expand::AnalyzeSignal(int16_t* random_vector) {
                               parameters.ar_filter,
                               kUnvoicedLpcOrder + 1,
                               128);
-    int16_t unvoiced_prescale;
-    if (WebRtcSpl_MaxAbsValueW16(unvoiced_vector, 128) > 4000) {
-      unvoiced_prescale = 4;
-    } else {
-      unvoiced_prescale = 0;
-    }
+    int16_t unvoiced_max_abs = WebRtcSpl_MaxAbsValueW16(unvoiced_vector, 128);

[kwiberg-webrtc, 2016/12/15 14:11:16]

const?

Also, won't the calculation break if the abs-max element is -2^15? IIRC
WebRtcSpl_MaxAbsValueW16 returns 2^15 - 1 in that case. I think I handle that
case in the CL I linked to earlier.

[ivoc, 2016/12/15 15:12:20]

Wow, good catch. I added an if-statement to increase the prescale by one when
unvoiced_max_abs is equal to 2^15-1, which should prevent this from happening.

+    // The maximum value of the dot product is 2^7 * 2^(2*n) = 2^(2*n + 7), so

[kwiberg-webrtc, 2016/12/15 14:11:16]

"is less than", since unvoiced_max_abs < 2^n.

Also, you never define n. Maybe say that "Pick the smallest n such that 2^n >
unvoiced_max_abs; then the maximum value of the dot product is less than 2^7 *
2^(2*n) = 2^(2*n + 7) [...]"

[ivoc, 2016/12/15 15:12:20]

Done.

+    // to prevent overflows we want 2n + 7 <= 31, which means we should shift by
+    // 2n + 7 - 31 bits, if this value is greater than zero.
+    int16_t unvoiced_prescale =

[kwiberg-webrtc, 2016/12/15 14:11:16]

const int?

[ivoc, 2016/12/15 15:12:20]

Due to the newly added if statement, this can no longer be a const.

+        std::max(0, 2 * WebRtcSpl_GetSizeInBits(unvoiced_max_abs) - 24);
+
     int32_t unvoiced_energy = WebRtcSpl_DotProductWithScale(unvoiced_vector,
                                                             unvoiced_vector,
                                                             128,