[FFmpeg-devel] [PATCH] avformat/hlsenc: Fix initial setting for start_pts

[Hongcheng Zhong]

----- On Mar 6, 2020, at 12:55 AM, Andreas Rheinhardt andreas.rheinhardt at gmail.com wrote:

> Hongcheng Zhong:
>> ----- On Mar 5, 2020, at 11:38 PM, lq lq at chinaffmpeg.org wrote:
>> 
>>>> 2020年3月5日 下午9:41，Hongcheng Zhong <sj.hc_zhong at sjtu.edu.cn> 写道：
>>>>
>>>> This patch fixes Bug #8469
>>>> If x264 baseline profile is used with other profiles,
>>>> start_pts will be initialized to audio stream's first pts,
>>>> while the duration is calculated based on video stream's pts.
>>>> In this patch the start_pts is initialized with the correct stream's first pts.
>>>>
>>>> Signed-off-by: Hongcheng Zhong <sj.hc_Zhong at sjtu.edu.cn>
>>>> ---
>>>> libavformat/hlsenc.c | 10 +++++++++-
>>>> 1 file changed, 9 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/libavformat/hlsenc.c b/libavformat/hlsenc.c
>>>> index f6dd894..3b2434f 100644
>>>> --- a/libavformat/hlsenc.c
>>>> +++ b/libavformat/hlsenc.c
>>>> @@ -126,6 +126,7 @@ typedef struct VariantStream {
>>>>     int has_video;
>>>>     int has_subtitle;
>>>>     int new_start;
>>>> +    int start_pts_from_audio;
>>>>     double dpp;           // duration per packet
>>>>     int64_t start_pts;
>>>>     int64_t end_pts;
>>>> @@ -2274,9 +2275,15 @@ static int hls_write_packet(AVFormatContext *s, AVPacket
>>>> *pkt)
>>>>
>>>>     if (vs->start_pts == AV_NOPTS_VALUE) {
>>>>         vs->start_pts = pkt->pts;
>>>> +        if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO)
>>>> +            vs->start_pts_from_audio = 1;
>>>> +    }
>>>> +    if (vs->start_pts_from_audio && st->codecpar->codec_type ==
>>>> AVMEDIA_TYPE_VIDEO && vs->start_pts > pkt->pts) {
>>>> +        vs->start_pts = pkt->pts;
>>>> +        vs->start_pts_from_audio = 0;
>>>>     }
>>>>
>>>> -   if (vs->has_video) {
>>>> +    if (vs->has_video) {
>>> What’s happen with this line?
>>>>         can_split = st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO &&
>>>>                     ((pkt->flags & AV_PKT_FLAG_KEY) || (hls->flags & HLS_SPLIT_BY_TIME));
>>>>         is_ref_pkt = (st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) &&
>>>>         (pkt->stream_index == vs->reference_stream_index);
>>>> @@ -2751,6 +2758,7 @@ static int hls_init(AVFormatContext *s)
>>>>         vs->sequence       = hls->start_sequence;
>>>>         vs->start_pts      = AV_NOPTS_VALUE;
>>>>         vs->end_pts      = AV_NOPTS_VALUE;
>>>> +        vs->start_pts_from_audio = 0;
>>>>         vs->current_segment_final_filename_fmt[0] = '\0';
>>>>
>>>>         if (hls->flags & HLS_SPLIT_BY_TIME && hls->flags & HLS_INDEPENDENT_SEGMENTS) {
>>>> --
>>>> 2.7.4
>>>>
>>>> _______________________________________________
>>>> ffmpeg-devel mailing list
>>>> ffmpeg-devel at ffmpeg.org
>>>> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>>>
>>>> To unsubscribe, visit link above, or email
>>>> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>> 
>> Actually, this line is useless because that vs is initialized as all zero.
>> Should I remove this line and submit this patch as v2 again? Thanks for your
>> reply.
> 
> But vs->has_video is initialized to a potentially nonzero value in
> hls_init().
> 
> - Andreas
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".

I write that line to make sure that vs->start_pts_from_audio is initialized as 0.
Since `hls->var_streams = av_mallocz(sizeof(*hls->var_streams) * nb_varstreams);`
in hlsenc.c: 1897, var_streams is allocated with av_mallocz(), that line is not 
necessary.