SDT = service description table
The information output is a list of three-element maps:
• the program number
• the service provider's name (think TV station name, e.g. "ARD")
• the service's name (think TV channel name, e.g. "arte HD")
The program number corresponds to the track property `program_number`.
See #1990 for the future use case: presenting this information in the
GUI.
Earlier versions of mkvmerge used to detect all tracks in MPEG
transport streams with multiple programs, even though the code wasn't
really implemented & tested for that. However, some tracks (usually
those from the second or a later program) were broken: they might not
contain any data, or only invalid data.
On top of that mkvmerge v12.0.0 contains a fix for #1980 where a track
isn't part of a PMT at all. An unintentional consequence of that fix
was that mkvmerge no longer detected all of the tracks in
multi-program streams. The reason is that in order to detect tracks
not mentioned in a PMT mkvmerge has to do detection by content in the
PES packets. That's only implemented for AAC at the moment. All other
tracks will be blacklisted as soon as they're found.
This wouldn't be a problem if all PMTs of all programs were always
located right at the start of the file with nothing in
between. Unfortunately many files contain track content between
PMTs. So that workflow was:
• mkvmerge finds first PMT, determines types for tracks listed in it
• mkvmerge now considers the PMT to be found
• Continuing scanning the file mkvmerge encounters content for tracks
not listed in the first PMT, attempting type detection by content,
failing for most and blacklisting their PIDs
• Next a second PMT is found, however, the PIDs listed in that PMT may
have already been found and blacklisted before — therefor they won't
be considered anymore
With this fix mkvmerge actively looks for the PMTs for all
programs. Detection by content is only attempted once all PMTs have
been located. That way all tracks will be detected again.
A side effect of either this patch or one of the other ones before is
that the track content is now OK. I don't know exactly why or which
commit actually fixed it.
Fixes parts of #1990.
