[FFmpeg-devel] [PATCH] avcodec/h264: Fix regression caused by removial of default_ref_list

[Hendrik Leppkes]

On Sun, Jan 3, 2016 at 7:03 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> From: Michael Niedermayer <michael at niedermayer.cc>
>
> This fixes a regression of the sample from Ticket 2371
>
> Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> ---
>  libavcodec/h264.h      |    1 +
>  libavcodec/h264_refs.c |   11 +++++++++--
>  2 files changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/libavcodec/h264.h b/libavcodec/h264.h
> index 5d9aecd..a5fc3a0 100644
> --- a/libavcodec/h264.h
> +++ b/libavcodec/h264.h
> @@ -669,6 +669,7 @@ typedef struct H264Context {
>       */
>      int max_pic_num;
>
> +    H264Ref default_ref[2];
>      H264Picture *short_ref[32];
>      H264Picture *long_ref[32];
>      H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size?
> diff --git a/libavcodec/h264_refs.c b/libavcodec/h264_refs.c
> index 52fedc1..f42d6a2 100644
> --- a/libavcodec/h264_refs.c
> +++ b/libavcodec/h264_refs.c
> @@ -208,6 +208,8 @@ static void h264_initialise_ref_list(H264Context *h, H264SliceContext *sl)
>              }
>          }
>      }
> +    for (i = 0; i < sl->list_count; i++)
> +        h->default_ref[i] = sl->ref_list[i][0];
>  }
>
>  static void print_short_term(H264Context *h);
> @@ -351,10 +353,14 @@ int ff_h264_decode_ref_pic_list_reordering(H264Context *h, H264SliceContext *sl)
>              if (   !sl->ref_list[list][index].parent
>                  || (!FIELD_PICTURE(h) && (sl->ref_list[list][index].reference&3) != 3)) {
>                  int i;
> -                av_log(h->avctx, AV_LOG_ERROR, "Missing reference picture\n");
> +                av_log(h->avctx, AV_LOG_ERROR, "Missing reference picture, default is %d\n", h->default_ref[list].poc);
>                  for (i = 0; i < FF_ARRAY_ELEMS(h->last_pocs); i++)
>                      h->last_pocs[i] = INT_MIN;
> -                return -1;
> +                if (h->default_ref[list].parent
> +                    && !(!FIELD_PICTURE(h) && (h->default_ref[list].reference&3) != 3))
> +                    sl->ref_list[list][index] = h->default_ref[list];
> +                else
> +                    return -1;
>              }
>              av_assert0(av_buffer_get_ref_count(sl->ref_list[list][index].parent->f->buf[0]) > 0);
>          }
> @@ -524,6 +530,7 @@ void ff_h264_remove_all_refs(H264Context *h)
>      }
>      h->short_ref_count = 0;
>
> +    memset(h->default_ref, 0, sizeof(h->default_ref));
>      for (i = 0; i < h->nb_slice_ctx; i++) {
>          H264SliceContext *sl = &h->slice_ctx[i];
>          sl->list_count = sl->ref_count[0] = sl->ref_count[1] = 0;
> --

Patch works for me. But I don't have a big opinion on which to use.
This one makes future merges easier, buty  ou are the most likely to
work on the H264 decoder in the future, so if noone else has any
opinions, i would just let you push whichever of the two you prefer.

Thanks

- Hendrik