This is a fix for https://trac.ffmpeg.org/ticket/6106 .
The issue is that when detecting a swapped AC3 marker the data of the frame is swapped. However, in subsequent frames the data swapped is taken from the first frame rather than the current frame. So all frames are considered to be correct even though they might not be. This leads to a high probe score and eventual misidentification. We run a service identifying millions of different wav files and apparently the chance of one 'correct' AC3 header in pcm data is big enough to hit this. But the chance of subsequent frames to be 'correct' is a lot smaller. I tested this with AC3 and EAC3 files from your samples repo and with some of our previously misidentified pcm wav files. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
