On 27/06/2025 22:49, Marvin Scholz wrote:
> The ret value is already checked earlier, making this condition
> impossible to ever happen.
> 
> Fix CID 1648350
> ---
>  libavcodec/vvc/refs.c | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/libavcodec/vvc/refs.c b/libavcodec/vvc/refs.c
> index 79967b77d3..e52cc0c10d 100644
> --- a/libavcodec/vvc/refs.c
> +++ b/libavcodec/vvc/refs.c
> @@ -310,8 +310,6 @@ int ff_vvc_output_frame(VVCContext *s, VVCFrameContext *fc, AVFrame *out, const
>                  ff_vvc_unref_frame(fc, frame, VVC_FRAME_FLAG_OUTPUT | VVC_FRAME_FLAG_BUMPING);
>              else
>                  ff_vvc_unref_frame(fc, frame, VVC_FRAME_FLAG_OUTPUT);
> -            if (ret < 0)
> -                return ret;
>  
>              av_log(s->avctx, AV_LOG_DEBUG,
>                     "Output frame with POC %d.\n", frame->poc);

I agree there shouldn't be two checks on ret here.  I am not sure this
is the right one to remove however.  Perhaps it is better to remove the
first check and only check ret after having unreferenced the source
frame, as was the behaviour prior to
a8d949bd96364892903bedbbc2eef11b712e5500 ?  I'm not certain, but it
looks to me as though that commit might have introduced a memory leak:
if av_frame_ref fails, then the source frame is never unreferenced and
its data never freed.

-- 
Thanks,
Frank