[FFmpeg-devel] [PATCH 03/19] adpcm: fix return value for small overreads.

[Paul B Mahol]

On 7/29/12, Nicolas George <nicolas.george at normalesup.org> wrote:
>
> Signed-off-by: Nicolas George <nicolas.george at normalesup.org>
> ---
>  libavcodec/adpcm.c |    5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
>
> Comment for this one and the next few:
>
> I believe it is better to know what decoders rely on
> FF_INPUT_BUFFER_PADDING_SIZE rather than catch everything in a common test,
> but if people prefer the opposite I can drop these cases.
>
>
> diff --git a/libavcodec/adpcm.c b/libavcodec/adpcm.c
> index b86168a..6706b07 100644
> --- a/libavcodec/adpcm.c
> +++ b/libavcodec/adpcm.c
> @@ -1265,7 +1265,10 @@ static int adpcm_decode_frame(AVCodecContext *avctx,
> void *data,
>      *got_frame_ptr   = 1;
>      *(AVFrame *)data = c->frame;
>
> -    return bytestream2_tell(&gb);
> +    ret = bytestream2_tell(&gb);
> +    if (ret > avpkt->size && ret <= avpkt->size +
> FF_INPUT_BUFFER_PADDING_SIZE)
> +        ret = avpkt->size; /* small overreads are acceptable */
> +    return ret;
>  }
>

Isn't point of bytestream2 to not overread?
So I'm against applying this hack.

Instead all cases where unchecked bytestream2 functions are used
(get_be16u and similar)
should have bytestream2_get_bytes_left.. before them.
>
> --
> 1.7.10.4
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>