[FFmpeg-devel] [PATCH v3 3/6] avformat/s337m: Make available as subdemuxer

[Gaullier Nicolas]

>Ah, so it's time and not phase? Renaming the #defines to make that clearer would be nice. Other than that, this isn't a huge issue
I assume the start of the packet to be synched with the start of "a" video frame, I think there is no use case where pcm and video are not synced, at least when dolby E is concerned.
So, DOLBY_E_PHASE_MIN/MAX are phase and well named I think.
Concerning S337M_PHASE_PROBE_MIN that I renamed S337M_PROBE_GUARDBAND_MIN_BYTES, this is still a phase, a kind of safe guard phase, to make even more unlikely to have a wrong detection. But the question is to determine if this phase should be specified in milliseconds or just in raw bytes (or another man could say in number of samples whether there are 16 or 24 bits). There may be other approaches, but if we consider strictly the risk of statistical match, it is related to the number of raw bytes, this is why I choosed raw bytes at the end and thus the name S337M_PROBE_GUARDBAND_MIN_BYTES looks good to me.
Furthermore, at the end, this S337M_PROBE_GUARDBAND_MIN_BYTES is zero for now (so whatever the unit), and in my experience, this is generally required as there are some programs which indeed have a zero-phase Dolby E (and BTW it appears some softwares doesn't like it). Taking the worst case scenario (16 bits), the sync code takes 4 bytes, so the raw probability is only 1/2^32 (and even less taking into account that usually broadcast programs starts with a little bit of silence). So anyway, maybe I took too much care on this, but at least here this is explicit in the code, I think.

Nicolas