[FFmpeg-devel] [PATCH] avformat/hlsenc: free varstreams after write all varstreams info

[Jan Ekström]

On Sat, Dec 22, 2018 at 10:55 AM Steven Liu <lq at chinaffmpeg.org> wrote:
>
> fix ticket: 7631
>
> Signed-off-by: Steven Liu <lq at chinaffmpeg.org>
> ---
>  libavformat/hlsenc.c | 50 +++++++++++++++++++++++++++++++++-----------------
>  1 file changed, 33 insertions(+), 17 deletions(-)
>
> diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
> index bdd2a113bd..e3cd6f375a 100644
> --- a/libavformat/hlsenc.c
> +++ b/libavformat/hlsenc.c
> @@ -2360,6 +2360,37 @@ static int hls_write_packet(AVFormatContext *s, AVPacket *pkt)
>      return ret;
>  }
>
> +static void hls_varstreams_free(struct AVFormatContext *s)
> +{

I think you only use the AVFormatContext to get the HLSContext, and
you already have that pointer in hls_write_trailer.

Thus I think you could just make this function take in a HLSContext *hls.

Also the function could be something like "hls_free_variant_streams",
but this is just an opinion.

> +    int i = 0;
> +    HLSContext *hls = s->priv_data;
> +    AVFormatContext *vtt_oc = NULL;
> +    VariantStream *vs = NULL;
> +
> +    for (i = 0; i < hls->nb_varstreams; i++) {

As far as I asked on IRC, we're OK with declaring iterators in a for
loop, so it can be "int i = 0;"

> +        vs = &hls->var_streams[i];
> +        vtt_oc = vs->vtt_avf;
> +

Also asked if things utilized within a for loop can have their own
variables declared at the top of the loop.
Thus the AVFormatContext and VariantStream can be declared here.

> +        av_freep(&vs->basename);
> +        av_freep(&vs->base_output_dirname);
> +        av_freep(&vs->fmp4_init_filename);
> +        if (vtt_oc) {
> +            av_freep(&vs->vtt_basename);
> +            av_freep(&vs->vtt_m3u8_name);
> +            avformat_free_context(vtt_oc);
> +        }
> +
> +        hls_free_segments(vs->segments);
> +        hls_free_segments(vs->old_segments);
> +        av_freep(&vs->m3u8_name);
> +        av_freep(&vs->streams);
> +        av_freep(&vs->agroup);
> +        av_freep(&vs->ccgroup);
> +        av_freep(&vs->baseurl);

I also wonder if you should separate actual per-variant stream freeing
into its own function which takes a VariantStream *vs in?

> +    }
> +
> +
> +}

Leave one empty line after the function, and no need for empty lines
at the end of the function inside of it.

F.ex.
{
    for (...) {
    }
}

following_function {
    ...

>  static int hls_write_trailer(struct AVFormatContext *s)
>  {
>      HLSContext *hls = s->priv_data;
> @@ -2451,30 +2482,15 @@ failed:
>              vs->size = avio_tell(vs->vtt_avf->pb) - vs->start_pos;
>              ff_format_io_close(s, &vtt_oc->pb);
>          }
> -        av_freep(&vs->basename);
> -        av_freep(&vs->base_output_dirname);
>          avformat_free_context(oc);
>
>          vs->avf = NULL;
>          hls_window(s, 1, vs);
> -
> -        av_freep(&vs->fmp4_init_filename);
> -        if (vtt_oc) {
> -            av_freep(&vs->vtt_basename);
> -            av_freep(&vs->vtt_m3u8_name);
> -            avformat_free_context(vtt_oc);
> -        }
> -
> -        hls_free_segments(vs->segments);
> -        hls_free_segments(vs->old_segments);
>          av_free(old_filename);
> -        av_freep(&vs->m3u8_name);
> -        av_freep(&vs->streams);
> -        av_freep(&vs->agroup);
> -        av_freep(&vs->ccgroup);
> -        av_freep(&vs->baseurl);
>      }
>
> +    hls_varstreams_free(s);
> +
>      for (i = 0; i < hls->nb_ccstreams; i++) {
>          ClosedCaptionsStream *ccs = &hls->cc_streams[i];
>          av_freep(&ccs->ccgroup);
> --
> 2.15.1
>

Initial quick look.

Jan