Crash on line 229

[chickeneye]

Hi, I'm trying to use this little program to reset the PTS values in a transport stream (I also can't get ffmpeg to work properly with setpts). The code is crashing at this line:
pCodec = avcodec_find_decoder(pVideoStream->codec->codec_id);

I think because pVideoStream is null. I've been trying to track down the issue and it looks like fps is NAN in the for loop above. Not sure why that is. I'm not an expert at video file formats, so I'm a little out of my depth but would love to use this code if I'm able to get it working.

Any thoughts?

[victusfate]

Howdy @chickeneye . We've since moved back to ffmpeg now that the more recent versions properly shift mpegts files, but I'd be happy to help.

As a word of caution, this tool was developed for Video only mpegts files and later adapted to audio/video streams, but it had some problems and was discarded in favor of a pure ffmpeg solution.

If memory serves, the argument we now use while shifting segments is "-initial_offset". We're using a version of FFMpeg2 on our production servers.
Here's what I'm using locally:

ffmpeg version 2.0 Copyright (c) 2000-2013 the FFmpeg developers
built on Nov  6 2013 16:11:29 with Apple LLVM version 5.0 (clang-500.2.79) (based on LLVM 3.3svn)
configuration: --prefix=/usr/local/Cellar/ffmpeglatest/snapshot --enable-shared --enable-pthreads --enable-gpl --enable-version3 --enable-nonfree --enable-hardcoded-tables --enable-avresample --enable-vda --enable-opencl --enable-libx264 --enable-libfaac --enable-libmp3lame --enable-openssl --enable-ffplay --enable-libvpx --enable-libvorbis --cc=cc --host-cflags= --host-ldflags= --enable-libfaac --enable-libxvid

If you prefer the stand alone c++ shifter program, additional debug info would be great for me to assist you. Can you share the media file you're using? If the pVideoStream is NULL that's a signal that the shifter code was unable to identify a video stream from this block of code: https://github.com/victusfate/mpegtsShifter/blob/master/shifter.cc#L208

for (i = 0; i < pInFormatCtx->nb_streams && (videoIndex < 0 || audioIndex < 0); i++) {
    // cout << "checking streams for video, current stream type = " << pInFormatCtx->streams[i]->codec->codec_type;
    // cout << " AVMEDIA_TYPE_VIDEO = " << AVMEDIA_TYPE_VIDEO << " AVMEDIA_TYPE_AUDIO = " <<  AVMEDIA_TYPE_AUDIO << endl;

    int64_t durationInt = pInFormatCtx->duration;
    double durationSeconds = (double)durationInt / AV_TIME_BASE;        
    double fps = av_q2d(pInFormatCtx->streams[i]->avg_frame_rate);
    unsigned int frameCount = 0;
    if (pInFormatCtx->streams[i]->nb_frames > 0) {
        frameCount = pInFormatCtx->streams[i]->nb_frames;
    }
    else {
        frameCount = floor(durationSeconds * fps);
    } 
    if (frameCount <= 0) continue;

    switch (pInFormatCtx->streams[i]->codec->codec_type) {
        case AVMEDIA_TYPE_VIDEO:
            // cout << "Initial for loop VIDEO CODEC, frames " << frameCount << endl;
            videoIndex = i;
            pInFormatCtx->streams[i]->discard = AVDISCARD_NONE;
            pVideoStream = addStream(pOutFormatCtx, pInFormatCtx->streams[i]);
            break;
        case AVMEDIA_TYPE_AUDIO:
            // cout << "Initial for loop AUDIO CODEC, frames " << frameCount << endl;
            audioIndex = i;
            pInFormatCtx->streams[i]->discard = AVDISCARD_NONE;
            pAudioStream = addStream(pOutFormatCtx, pInFormatCtx->streams[i]);
            break;
        default:
            // cout << "Initial for loop AVDISCARD_ALL" << endl;
            pInFormatCtx->streams[i]->discard = AVDISCARD_ALL;
            break;
    }
}

[chickeneye]

Hey, thanks for getting back to me. I'd actually be happy to use ffmpeg as well, but I feel like it's adding packets instead of overwriting the existing PTS. Would you mind sharing your ffmpeg command for offsetting the PTS? Perhaps I'm doing something wrong. I can give that a try and if it seems to work we don't have to go debugging the old stuff!

FYI, you can see my issue in full context here:
http://stackoverflow.com/questions/20481249/capturing-segmenting-video-on-ios-and-rejoining-via-hls-results-in-audio-dropout

[victusfate]

sure but it's a little verbose, the only parameters you are interested in are -t for the duration, and -initial_offset for the shift. Here's a segment with a 10 second shift

/opt/ffmpeg -i /tmp/cameo/58527/6fc2fa1a7418bf9d4aa90aa384d0eef2244631e8 -threads 0 -ss 10 -i /tmp/cameo/58527/79e684d793e209ebc9b12a5ad82298cb5e94cb54 -codec:v libx264 -pix_fmt yuv420p -preset veryfast -strict -2 -bsf:v h264_mp4toannexb -flags -global_header -crf 28 -profile:v baseline -x264opts level=3:keyint_min=24:keyint=24:scenecut=0 -b:v 100000 -bt 100000 -bufsize 100000 -maxrate 100000 -r 12 -s 320x180 -map 0:0 -map 1:0 -codec:a aac -strict -2 -b:a 64k -ab 64k -ac 2 -ar 44100 -t 9.958333333333334 -segment_time 10.958333333333334 -f segment -initial_offset 10 -segment_format mpegts -y /tmp/cameo/58527/100K%01d.ts -codec:v libx264 -pix_fmt yuv420p -preset veryfast -strict -2 -bsf:v h264_mp4toannexb -flags -global_header -crf 28 -profile:v baseline -x264opts level=3:keyint_min=24:keyint=24:scenecut=0 -b:v 200000 -bt 200000 -bufsize 200000 -maxrate 200000 -r 12 -s 320x180 -map 0:0 -map 1:0 -codec:a aac -strict -2 -b:a 64k -ab 64k -ac 2 -ar 44100 -t 9.958333333333334 -segment_time 10.958333333333334 -f segment -initial_offset 10 -segment_format mpegts -y /tmp/cameo/58527/200K%01d.ts -codec:v libx264 -pix_fmt yuv420p -preset veryfast -strict -2 -bsf:v h264_mp4toannexb -flags -global_header -crf 28 -profile:v baseline -x264opts level=3:keyint_min=24:keyint=24:scenecut=0 -b:v 364000 -bt 364000 -bufsize 364000 -maxrate 364000 -r 24 -s 320x180 -map 0:0 -map 1:0 -codec:a aac -strict -2 -b:a 64k -ab 64k -ac 2 -ar 44100 -t 9.958333333333334 -segment_time 10.958333333333334 -f segment -initial_offset 10 -segment_format mpegts -y /tmp/cameo/58527/364K%01d.ts -codec:v libx264 -pix_fmt yuv420p -preset veryfast -strict -2 -bsf:v h264_mp4toannexb -flags -global_header -crf 28 -profile:v baseline -x264opts level=3:keyint_min=24:keyint=24:scenecut=0 -b:v 664000 -bt 664000 -bufsize 664000 -maxrate 664000 -r 24 -s 480x270 -map 0:0 -map 1:0 -codec:a aac -strict -2 -b:a 64k -ab 64k -ac 2 -ar 44100 -t 9.958333333333334 -segment_time 10.958333333333334 -f segment -initial_offset 10 -segment_format mpegts -y /tmp/cameo/58527/664K%01d.ts -codec:v libx264 -pix_fmt yuv420p -preset veryfast -strict -2 -bsf:v h264_mp4toannexb -flags -global_header -crf 23 -profile:v baseline -x264opts level=3.1:keyint_min=24:keyint=24:scenecut=0 -b:v 1264000 -bt 1264000 -bufsize 1264000 -maxrate 1264000 -r 24 -s 640x360 -map 0:0 -map 1:0 -codec:a aac -strict -2 -b:a 64k -ab 64k -ac 2 -ar 44100 -t 9.958333333333334 -segment_time 10.958333333333334 -f segment -initial_offset 10 -segment_format mpegts -y /tmp/cameo/58527/1264K%01d.ts

[chickeneye]

Wow, that is verbose! Thanks for sending. Is this to segment an existing TS into smaller ones or to adjust the offsets in a set of pre-segmented TS files? My challenge is the latter. I've got a set of TS files made separately that I'm trying to play back as HLS segments.

[victusfate]

It's one of several, but it generates all the bitrate qualities