[FFmpeg-devel] [PATCH] lavfi/atempo: avoid asendcmd assertion failure

[Pavel Koshevoy]

On Tue, Mar 12, 2024 at 9:01 PM Pavel Koshevoy <pkoshevoy at gmail.com> wrote:

> From: Rajiv Harlalka <rajivharlalka009 at gmail.com>
>
> Check for zeros equal to the total samples early, because in
> that case we would already be leaving the first few frames out.
>
> Fixes trac ticket #10692
> ---
>  libavfilter/af_atempo.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/libavfilter/af_atempo.c b/libavfilter/af_atempo.c
> index 4621b67b03..654b080e89 100644
> --- a/libavfilter/af_atempo.c
> +++ b/libavfilter/af_atempo.c
> @@ -531,21 +531,21 @@ static int yae_load_frag(ATempoContext *atempo,
>      dst = frag->data;
>
>      start = atempo->position[0] - atempo->size;
> -    zeros = 0;
>
> -    if (frag->position[0] < start) {
> -        // what we don't have we substitute with zeros:
> -        zeros = FFMIN(start - frag->position[0], (int64_t)nsamples);
> -        av_assert0(zeros != nsamples);
> -
> -        memset(dst, 0, zeros * atempo->stride);
> -        dst += zeros * atempo->stride;
> -    }
> +    // what we don't have we substitute with zeros:
> +    zeros =
> +      frag->position[0] < start ?
> +      FFMIN(start - frag->position[0], (int64_t)nsamples) : 0;
>
>      if (zeros == nsamples) {
>          return 0;
>      }
>
> +    if (frag->position[0] < start) {
> +        memset(dst, 0, zeros * atempo->stride);
> +        dst += zeros * atempo->stride;
> +    }
> +
>      // get the remaining data from the ring buffer:
>      na = (atempo->head < atempo->tail ?
>            atempo->tail - atempo->head :
> --
> 2.35.3
>
>
This is a reformatting of an earlier patch from Rajiv --
https://ffmpeg.org/pipermail/ffmpeg-devel/2024-March/322946.html
I've tested it locally -- looks good to me, feel free to apply and push.

Thank you,
    Pavel.