Recherche avancée

Sur d’autres sites (8542)

• Fatal Python error : could not acquire lock for at interpreter shutdown, possibly due to daemon threads

  3 octobre 2019, par JavNoor

  Running the following :

  import imageio
  
  
  
  class vidrdf:
  
      def __init__(self, vidfile):
  
          self.vid = imageio.get_reader(vidfile,  'ffmpeg')
  
  
  
  vidfile = 'movie.mov'
  
  rdfobj = vidrdf(vidfile)

  I get :

  Fatal Python error: could not acquire lock for <_io.BufferedReader name=8> at interpreter shutdown, possibly due to daemon threads
  
  
  
  Thread 0x00007fd2dc2a8700 (most recent call first):
  
    File “/path/miniconda3/envs/flower/lib/python3.7/site-packages/imageio_ffmpeg/_parsing.py", line 61 in run
  
    File "miniconda3/envs/flower/lib/python3.7/threading.py", line 926 in _bootstrap_inner
  
    File “/path/miniconda3/envs/flower/lib/python3.7/threading.py", line 890 in _bootstrap
  
  
  
  Current thread 0x00007fd2f0bb9700 (most recent call first):
  
    File “/path/miniconda3/envs/flower/lib/python3.7/subprocess.py", line 1704 in _communicate
  
    File “/path/miniconda3/envs/flower/lib/python3.7/subprocess.py", line 939 in communicate
  
    File “/path/miniconda3/envs/flower/lib/python3.7/site-packages/imageio_ffmpeg/_io.py", line 193 in read_frames
  
    File “/path/miniconda3/envs/flower/lib/python3.7/site-packages/imageio/plugins/ffmpeg.py", line 342 in _close
  
    File “/path/miniconda3/envs/flower/lib/python3.7/site-packages/imageio/core/format.py", line 252 in close
  
    File “/path/miniconda3/envs/flower/lib/python3.7/site-packages/imageio/core/format.py", line 241 in __del__
  
  Aborted (core dumped)

  It doesn’t happen if I try :

  import imageio
  
  
  
  class vidrdf:
  
      def __init__(self, vidfile):
  
          self.vid = imageio.get_reader(vidfile,  'ffmpeg')
  
  
  
  vidfile = 'movie.mov'
  
  vidrdf(vidfile)

  or

  import imageio
  
  
  
  class vidrdf:
  
      def __init__(self, vidfile):
  
          vid = imageio.get_reader(vidfile,  'ffmpeg')
  
  
  
  vidfile = 'movie.mov'
  
  rdfobj = vidrdf(vidfile)

  So this is clearly an issue with copying the object. I’ve searched about daemon threads, but since I’m using imageio directly I can’t figure out why and how to resolve it. I would appreciate any recommendations.

• Ffmpeg does not properly convert videos when run as daemon

  20 janvier 2019, par DGoiko

  I’m using ffmpeg to convert and split videos from a Python process. When I run it from console as a normal user everything works as expected, however, if I run it as a daemon it produces the same amount of files but with silence in almost the whole file.

  This is the line I execute from python :

  def convert_to_ffmpeg(file, ffmpeg_folder, chunk_length):
  
      res = -1
  
      try:
  
          file_name = os.path.join(ffmpeg_folder, ntpath.basename(file) + ".flac")
  
          res = os.system("ffmpeg -i " + file + " -f segment -segment_time "+str(chunk_length)+" " + file_name + " -y  >/dev/null 2>&1")
  
      except Exception as e:
  
          raise MyException(code=210005, detalails=file, cause=e)
  
      if not res == 0:
  
          raise MyException(code=210004, detalails=str(res) + " - " + file)
  
      return file_name

  When I run it from a normal console it works as expected : it splits the video in chunks of "self._chunk_length" seconds.

  However, if I run the program using python-daemon with this code :

  if daemon_run:
  
      pid_file = check_unclean_exit(config['pidfile'], log)
  
  
  
      context = daemon.DaemonContext(
  
          working_directory=approot,
  
          umask=0o002,
  
          pidfile=pid_file,
  
      )
  
      context.signal_map = {
  
          signal.SIGTERM: programa_cleanup,
  
          signal.SIGHUP: 'terminate',
  
          # signal.SIGUSR1: reload_program_config,
  
      }
  
  
  
      with context:
  
          convert_to_ffmpeg("/foo/testfile.mkv", "/foo/ffmpeg", 40) 

  It produces EXACTLY the same chunks, however, most of the content of the chunks is just silence. It splits in the proper point, but doesn’t fill the contents of the audio output properly. Of course, no exception is thrown, as ffmpeg always returns 0

  I did even install a completly fresh Debian 9 into a virtual machine to discard problems with ffmpeg and my old-good computer’s setup.

  Of course this is a simplified example. The real thing is a multithread application which watches a directory for files to convert.

• hevc : fix race condition in max_ra/seq_decode.

  24 mai 2017, par Ronald S. Bultje

  hevc : fix race condition in max_ra/seq_decode.
  These variables are shared between frame threads, but they are updated
  
  post-setup_finished() if a EOB/EOS slice type occurs. Moving the EOB/EOS
  
  slices to the next frame thread instance (by parsing them leading into
  
  the next picture instead of trailing behind the last picture) effectively
  
  prevents this race condition.
  This fixes tsan failures on hevc-conformance-NoOutPrior_A_Qualcomm_1.
  

  • [DH] libavcodec/hevc_parser.c
  • [DH] libavcodec/hevcdec.c