It still crashes with the following change. It crashes at
ff_pvq_search_exact_avx () instead. We should wait for your improvement to fix
the crash.
Colin
________________________________
From: ffmpeg-devel <ffmpeg-devel-boun...@ffmpeg.org> on behalf of Rostislav
Pehlivanov <atomnu...@gmail.com>
Sent: November 22, 2017 8:02 PM
To: FFmpeg development discussions and patches
Subject: Re: [FFmpeg-devel] [PATCH] libavcodec/opus: Add NULL pointer check for
incoming data; avoid segfault
On 23 November 2017 at 00:53, Colin NG <colin...@hotmail.com> wrote:
> ---
> Fix for ticket 7674
>
> libavcodec/opus_pvq.c | 16 +++++++++++++---
> 1 file changed, 13 insertions(+), 3 deletions(-)
>
> diff --git a/libavcodec/opus_pvq.c b/libavcodec/opus_pvq.c
> index f98b85d..02ccd69 100644
> --- a/libavcodec/opus_pvq.c
> +++ b/libavcodec/opus_pvq.c
> @@ -504,6 +504,9 @@ static av_always_inline uint32_t
> quant_band_template(CeltPVQ *pvq, CeltFrame *f,
> int longblocks = (B0 == 1);
> uint32_t cm = 0;
>
> + if (!X)
> + return cm;
> +
> if (N == 1) {
> float *x = X;
> for (i = 0; i <= stereo; i++) {
> @@ -795,7 +798,7 @@ static av_always_inline uint32_t
> quant_band_template(CeltPVQ *pvq, CeltFrame *f,
> f->remaining2 -= curr_bits;
> }
>
> - if (q != 0) {
> + if (X && !q) {
> /* Finally do the actual (de)quantization */
> if (quant) {
> cm = celt_alg_quant(rc, X, N, (q < 8) ? q : (8 + (q & 7))
> << ((q >> 3) - 1),
> @@ -902,6 +905,9 @@ static float pvq_band_cost(CeltPVQ *pvq, CeltFrame *f,
> OpusRangeCoder *rc, int b
> float *Y_orig = f->block[1].coeffs + (ff_celt_freq_bands[band] <<
> f->size);
> OPUS_RC_CHECKPOINT_SPAWN(rc);
>
> + if (!X && !Y)
> + return 0.0f;
> +
> memcpy(X, X_orig, band_size*sizeof(float));
> if (Y)
> memcpy(Y, Y_orig, band_size*sizeof(float));
> @@ -911,7 +917,6 @@ static float pvq_band_cost(CeltPVQ *pvq, CeltFrame *f,
> OpusRangeCoder *rc, int b
> int curr_balance = f->remaining / FFMIN(3, f->coded_bands - band);
> b = av_clip_uintp2(FFMIN(f->remaining2 + 1, f->pulses[band] +
> curr_balance), 14);
> }
> -
> if (f->dual_stereo) {
> pvq->encode_band(pvq, f, rc, band, X, NULL, band_size, b / 2,
> f->blocks, NULL,
> f->size, norm1, 0, 1.0f, lowband_scratch, cm[0]);
> @@ -925,7 +930,12 @@ static float pvq_band_cost(CeltPVQ *pvq, CeltFrame
> *f, OpusRangeCoder *rc, int b
>
> for (i = 0; i < band_size; i++) {
> err_x += (X[i] - X_orig[i])*(X[i] - X_orig[i]);
> - err_y += (Y[i] - Y_orig[i])*(Y[i] - Y_orig[i]);
> + }
> +
> + if (Y) {
> + for (i = 0; i < band_size; i++) {
> + err_y += (Y[i] - Y_orig[i])*(Y[i] - Y_orig[i]);
> + }
> }
>
> dist = sqrtf(err_x) + sqrtf(err_y);
> --
> 2.7.4
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
ffmpeg-devel Info Page<http://ffmpeg.org/mailman/listinfo/ffmpeg-devel>
ffmpeg.org
This list is about FFmpeg development discussions and patches; but not for
bug-reports. Please read the Code-of-conduct. To see the collection of prior
postings to ...
>
I actually found out about this bug while working on it a few days ago but
wanted to wait until I got the opus_rc and some opusenc_psy improvements
fully done.
Pushed my fix for mono. I disagree with the rest. There should always be a
channel of audio (X) to analyze.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
ffmpeg-devel Info Page<http://ffmpeg.org/mailman/listinfo/ffmpeg-devel>
ffmpeg.org
This list is about FFmpeg development discussions and patches; but not for
bug-reports. Please read the Code-of-conduct. To see the collection of prior
postings to ...
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
