Recherche avancée

Sur d’autres sites (13701)

• FFMPEG memory leak on FLV video frame decoding

  10 août 2014, par Michael IV

  I am decoding FLV video on Windows using FFMPEG latest dev version(20140810) .Monitoring memory consumption of my program process I found the memory footprint constantly increasing.I do packet deallocation and also tried to delete and then reallocate the AVFrame anew on each decode.But it doesn’t help.I read on some threads people pointing out there is an internal memory leak in H264 decoder but I have seen no official confirmation of it nor any solution.
  Here is how I decode a frame :

  AVPacket packet;
  
  av_read_frame(_ifmt_ctx, &packet);
  
  
  
  if (packet.stream_index == _in_video_stream->index) 
  
  {
  
      int isGotVideoFrame = 0;
  
  
  
      // Decode video frame
  
      ret = avcodec_decode_video2(_dec_in_video_ctx, _src_video_frame,
  
                     &isGotVideoFrame, &packet);
  
  
  
      if (1 == isGotVideoFrame) 
  
      {
  
          sws_scale(_sws_ctx, (const uint8_t * const*) _src_video_frame->data,
  
                    _src_video_frame->linesize, 0,_inVideoHeight, 
  
                    _dst_video_frame->data, _dst_video_frame->linesize);
  
  
  
          uint8_t* dest = new uint8_t[_numBytes];
  
          memcpy(dest, _dst_video_frame->data[0], _numBytes);
  
  
  
          av_free_packet(&packet);
  
          _frames_cache.push_back(dest);
  
      }
  
  
  
      av_frame_unref(_src_video_frame);
  
      av_frame_free(&_src_video_frame);
  
      _src_video_frame = av_frame_alloc();
  
  }

  Then in another place on each frame I delete ’dest’ from the vector :

      uint8_t * fr = _frames_cache.front();
  
      _frames_cache.erase(_frames_cache.begin());
  
      delete [] fr ;

• c++ ffmpeg "Starting new cluster" error

  9 novembre 2016, par R2-D2

  I want to encode a live stream into webm, but ffmpeg gets stuck in a live lock after 5 seconds stating

  [webm @ 0x1d81940] Starting new cluster at offset 0 bytes, pts 5040dts 5040

  I tried increasing the related AVFormatContext params

  av_opt_set_int(oc->priv_data, "chunk_duration", INT_MAX, 0);
  
  av_opt_set_int(oc->priv_data, "cluster_time_limit", INT_MAX, 0);
  
  av_opt_set_int(oc->priv_data, "cluster_size_limit", INT_MAX, 0);

  which avoids the error for about 30 seconds, but then again ffmpeg hangs

  [webm @ 0xbc9940] Starting new cluster due to timestamp
  
  [webm @ 0xbc9940] Starting new cluster at offset 0 bytes, pts 32800dts 32800

  The error can be reproduced with the official example doc/examples/muxing.c just by writing into a buffer instead of a file like this

  oc = avformat_alloc_context();
  
  oc->oformat = av_guess_format("webm", NULL, NULL);
  
  oc->oformat->flags |= AVFMT_NOFILE;

  and for actual writing

  uint8_t *output_buf;
  
  avio_open_dyn_buf(&oc->pb);
  
  
  
  avformat_write_header(oc, &opt);
  
  /* or */
  
  av_interleaved_write_frame(fmt_ctx, pkt);
  
  
  
  avio_close_dyn_buf(oc->pb, &output_buf);
  
  av_free(output_buf);

  How can I encode webm into a buffer ?
  (And why does it work for files ?)

• avformat/dashenc : Added support for Low-latency HLS(Experimental)

  12 décembre 2018, par kjeyapal@akamai.com

  avformat/dashenc : Added support for Low-latency HLS(Experimental)
  Apple doesn't have an official spec for LHLS. Meanwhile hls.js player folks are
  
  trying to standardize a open LHLS spec. The draft spec is available in https://github.com/video-dev/hlsjs-rfcs/blob/lhls-spec/proposals/0001-lhls.md
  
  This option will also try to comply with the above open spec, till Apple's spec officially supports it.
  
  Applicable only when @varstreaming and @varhls_playlist options are enabled.
  

  • [DH] doc/muxers.texi
  • [DH] libavformat/dashenc.c