Error While Decoding Mb 40 9 Bytestream 9

FFmpeg stops outputting frames unexpectedly and indefinitely on \'bad\' h264 over UDP stream, it happens quite rarely on a wireless link. I\'ve managed to reproduce by recording (pcap) and manipulating a stream by discarding every second packet, which causes the video to be played very badly.

At 01:38 ffplay stops displaying frames (restarting ffplay plays the stream again). It is as if a specific packet/frame causes ffmpeg to enter a specific state from which it cannot recover.How to reproduce:Play the attached pcap file on an active local network adapter with the IP 192.168.19.121.% ffplay -i udp://192.168.19.121:40002Video starts at 0:27, issue occurs at 01:38.Tried on latest git, 2.7 and 2.5.2 (Windows). Yes, definitely same machine. What I had to do, since I\'m using a wireless link, is to make sure the physical network adapter is connected to a network switch or similar (so it is actually \'working\'), and configure that IP on the physical network connection/interface.Another thing, the video will start 27 seconds into the pcap file. You can use for example to make sure you see packets with destination 192.168.19.121:40002.I highly appreciate your efforts to reproduce this issue, very impressive indeed since it\'s not that trivial.

Console: h264 @ 0afaa0 error while decoding MB 59 10, bytestream (-9) h264 @ 5e7420 left block unavailable for. h264 @ 0x91d7820 concealing 133 DC, 133 AC, 133 MV errors Previous Topic Next Topic › Classic List: Threaded ♦ ♦.

Why?I mean: If you can explain how I can reproduce your issue, please explain here so that other potentially interested people could also test. If you cannot explain, how would talking on irc help?Well, I am able to reproduce using the pcap file, and ffplay (Windows). I\'m just trying to be creative and help you reproduce, after that it will be easy to update this ticket with instructions.Have you tried to physically connect the Ethernet interface to a hub/switch/router? For me it was a requirement. Can you see 192.168.19.121:40002 packets with Wireshark or alike?Thanks!

Replying to:Just for the record, this stream/file is in very bad quality (every other UDP packet, alternately) just in order to reproduce this issue which seems to happen even on a brief bad stream of sort.Sorry, I do not understand what you are trying to report: The file you provide contains 390 broken frames and all can be transcoded: If you cut 400k (10%) from the end, 361 frames are decoded, if you cut 500k from the beginning, 358 frames are decoded indicating that the whole file can be read. With a file as broken as this one, you cannot expect that you get a decodable stream no matter where you cut it. This could be different if the stream were recognisable but this is not the case. Sorry, I do not understand what you are trying to report: The file you provide contains 390 broken frames and all can be transcoded: If you cut 400k (10%) from the end, 361 frames are decoded, if you cut 500k from the beginning, 358 frames are decoded indicating that the whole file can be read. With a file as broken as this one, you cannot expect that you get a decodable stream no matter where you cut it.

This could be different if the stream were recognisable but this is not the case.Let me explain from the beginning. I have a \'black box\' generating live h264 over UDP stream, but once in a while, ffmpeg seems to enter a situation where it unexpectedly stops outputting frames, without recovery. The only solution in that point is to restart playback/ffmpeg, and then, the stream is immediately decoded normally. I\'m not an expert in this, but it seems the decoder enters a dead-lock-like situation within itself.By recording the network stream of that UDP/h264 stream, omitting every second packet, this was the only way I could reproduce the described \'decoder halt\' condition.

FFmpeg stops outputting frames unexpectedly and indefinitely on \'bad\' h264 over UDP stream, it happens quite rarely on a wireless link. I\'ve managed to reproduce by recording (pcap) and manipulating a stream by discarding every second packet, which causes the video to be played very badly.

At 01:38 ffplay stops displaying frames (restarting ffplay plays the stream again). It is as if a specific packet/frame causes ffmpeg to enter a specific state from which it cannot recover.How to reproduce:Play the attached pcap file on an active local network adapter with the IP 192.168.19.121.% ffplay -i udp://192.168.19.121:40002Video starts at 0:27, issue occurs at 01:38.Tried on latest git, 2.7 and 2.5.2 (Windows). Yes, definitely same machine. What I had to do, since I\'m using a wireless link, is to make sure the physical network adapter is connected to a network switch or similar (so it is actually \'working\'), and configure that IP on the physical network connection/interface.Another thing, the video will start 27 seconds into the pcap file. You can use for example to make sure you see packets with destination 192.168.19.121:40002.I highly appreciate your efforts to reproduce this issue, very impressive indeed since it\'s not that trivial.

Console: h264 @ 0afaa0 error while decoding MB 59 10, bytestream (-9) h264 @ 5e7420 left block unavailable for. h264 @ 0x91d7820 concealing 133 DC, 133 AC, 133 MV errors Previous Topic Next Topic › Classic List: Threaded ♦ ♦.

Why?I mean: If you can explain how I can reproduce your issue, please explain here so that other potentially interested people could also test. If you cannot explain, how would talking on irc help?Well, I am able to reproduce using the pcap file, and ffplay (Windows). I\'m just trying to be creative and help you reproduce, after that it will be easy to update this ticket with instructions.Have you tried to physically connect the Ethernet interface to a hub/switch/router? For me it was a requirement. Can you see 192.168.19.121:40002 packets with Wireshark or alike?Thanks!

Replying to:Just for the record, this stream/file is in very bad quality (every other UDP packet, alternately) just in order to reproduce this issue which seems to happen even on a brief bad stream of sort.Sorry, I do not understand what you are trying to report: The file you provide contains 390 broken frames and all can be transcoded: If you cut 400k (10%) from the end, 361 frames are decoded, if you cut 500k from the beginning, 358 frames are decoded indicating that the whole file can be read. With a file as broken as this one, you cannot expect that you get a decodable stream no matter where you cut it. This could be different if the stream were recognisable but this is not the case. Sorry, I do not understand what you are trying to report: The file you provide contains 390 broken frames and all can be transcoded: If you cut 400k (10%) from the end, 361 frames are decoded, if you cut 500k from the beginning, 358 frames are decoded indicating that the whole file can be read. With a file as broken as this one, you cannot expect that you get a decodable stream no matter where you cut it.

This could be different if the stream were recognisable but this is not the case.Let me explain from the beginning. I have a \'black box\' generating live h264 over UDP stream, but once in a while, ffmpeg seems to enter a situation where it unexpectedly stops outputting frames, without recovery. The only solution in that point is to restart playback/ffmpeg, and then, the stream is immediately decoded normally. I\'m not an expert in this, but it seems the decoder enters a dead-lock-like situation within itself.By recording the network stream of that UDP/h264 stream, omitting every second packet, this was the only way I could reproduce the described \'decoder halt\' condition.