Recherche avancée

Sur d’autres sites (7643)

• avformat/rpl : Fix check for negative values

  18 novembre 2024, par Michael Niedermayer

  avformat/rpl : Fix check for negative values
  Fixes : signed integer overflow : 10 * -1923267925333400000 cannot be represented in type 'int64_t' (aka 'long')
  
  Fixes : 378891963/clusterfuzz-testcase-minimized-fuzzer_loadfile_direct-5714338935013376
  
  Found-by : ossfuzz
  
  Reported-by : Kacper Michajlow <kasper93@gmail.com>
  
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavformat/rpl.c

• avcodec/proresdec : Consider negative bits left

  19 juillet 2024, par Michael Niedermayer

  avcodec/proresdec : Consider negative bits left
  Fixes : 70036/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PRORES_fuzzer-6298797647396864
  
  Fixes : shift exponent 40 is too large for 32-bit type 'uint32_t' (aka 'unsigned int')
  Found-by : continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
  
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavcodec/proresdec.c

• Negative pts timestamps in RTSP packets using FFMPEG/C++ while remuxing H264 format into MPEG-TS

  17 juillet 2024, par RPAnimation

  I am muxing an H264 stream from RTSP camera to a MPEG-TS container (no re-encoding, just putting the bytestream to a container). At some point I figured out how to deal with pts and dts fields in AVPacket by combining information I found online. It works, but I am wondering whether my approach is correct :

  


  Main loop :

  


  int cnt = 0;
AVPacket *packet = av_packet_alloc();
int64_t last_mux_dts = AV_NOPTS_VALUE;
int dropped = 0;
while ((av_read_frame(in_fmt_ctx, packet) >= 0) && cnt < 1000)
{
////// Current approach
    if (packet->stream_index != video_stream_index)
    {
        continue;
    }

    if (packet->pts <= 0)
    {
        dropped++;
        continue;
    }

    if (packet->dts <= last_mux_dts)
    {
        packet->dts = packet->pts = last_mux_dts + 1;
    }

    last_mux_dts = packet->dts;

////// Previous approach
//  in_stream = in_fmt_ctx->streams[pkt->stream_index];
//  out_stream = out_fmt_ctx->streams[pkt->stream_index];
//  av_packet_rescale_ts(pkt, in_stream->time_base, out_stream->time_base)
//////
    packet->pos = -1;

    if (av_interleaved_write_frame(out_fmt_ctx, packet) < 0)
    {
        fprintf(stderr, "Error muxing packet\n");
        break;
    }
    cnt++;
}


  


  Previously I tried setting permutations of following options to my AVFormatContext :

  


  AVDictionary *options = NULL;
av_dict_set(&options, "use_wallclock_as_timestamps", "1", 0);
av_dict_set(&options, "max_interleave_delta", "0", 0);
av_dict_set(&options, "seek_timestamp", "1", 0);
av_dict_set(&options, "avoid_negative_ts", "make_zero", 0);
av_dict_set(&options, "fflags", "+genpts", 0);


  


  and using av_packet_rescale_ts after receiving a packet (like in this example from FFMPEG repo).

  


  But every time the output of such approach was a .ts file that was about 20h in length while only recording for 5s. It was still playable though.

  


  Is it a right practice to modify pts/dts fields manually or is there a built in functionality in libav to deal with them in such a use-case ?