Bump variable sizes in response to fuzzer bug
The fuzzers detected a possible overflow in the multiplication of sum and gainQ10. Since gainQ10 cannot be larger than 2048000 (see WebRtcIsac_kQGain2Levels) and sum cannot be larger than 2^16, a int64 is large enough to hold the result. Bug: chromium:904909 Change-Id: Icb12821d4006aaaaf70a5735d2abd2b96f7a2f0e Reviewed-on: https://webrtc-review.googlesource.com/c/111921 Reviewed-by: Ivo Creusen <ivoc@webrtc.org> Commit-Queue: Jonas Olsson <jonasolsson@webrtc.org> Cr-Commit-Position: refs/heads/master@{#25787}
This commit is contained in:
@ -96,7 +96,7 @@ static void FindInvArSpec(const int16_t* ARCoefQ12,
|
||||
const int32_t gainQ10,
|
||||
int32_t* CurveQ16) {
|
||||
int32_t CorrQ11[AR_ORDER + 1];
|
||||
int32_t sum, tmpGain;
|
||||
int64_t sum, tmpGain;
|
||||
int32_t diffQ16[FRAMESAMPLES / 8];
|
||||
const int16_t* CS_ptrQ9;
|
||||
int k, n;
|
||||
|
Reference in New Issue
Block a user