[FFmpeg-devel] [PATCH 2/2] avcodec/h264_slice: Never call get_format() for just checking if the format matches

[wm4]

On Sun, 11 Oct 2015 21:16:27 +0200
Michael Niedermayer <michaelni at gmx.at> wrote:

> From: Michael Niedermayer <michael at niedermayer.cc>
> 
> Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
> ---
>  libavcodec/h264_slice.c |    4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c
> index cce97d9..daa3737 100644
> --- a/libavcodec/h264_slice.c
> +++ b/libavcodec/h264_slice.c
> @@ -985,6 +985,10 @@ static enum AVPixelFormat get_pixel_format(H264Context *h, int force_callback)
>      for (i=0; choices[i] != AV_PIX_FMT_NONE; i++)
>          if (non_j_pixfmt(choices[i]) == non_j_pixfmt(h->avctx->pix_fmt) && !force_callback)
>              return choices[i];
> +
> +    if (!force_callback)
> +        return AV_PIX_FMT_NONE;
> +
>      return ff_thread_get_format(h->avctx, choices);
>  }
>  

So if I can see this right, the whole purpose of this is to check
whether the h264 parameters map to a lavc pixfmt, and bail out or
reinitialize if it doesn't. Why can't this be delayed later? What
actually needs it sooner than the first "real" get_format? For dealing
with paramater changes, why can't it check the raw parameters instead?
Doing it this way seems a bit convoluted. (I understand it now that I
thought about it, but normally I'd think it's VERY weird that it
somehow can go on without using the user-decided pixfmt, or that the
user-decided pixfmt sometimes doesn't seem to matter?)