[FFmpeg-user] Robustness on processing transport streams (TS)?

[Carl Eugen Hoyos]

2018-12-27 22:21 GMT+01:00, Uwe Freese <uwe.freese at gmx.de>:

>> A long time back, there was ProjectX to fix transport streams, I
>> believe there is another program now (suggested by the HandBrake
>> user support) but I forgot its name.
>
> ProjectX doesn't work with HD.

HD should have no issues, h264 support is very limited iirc.

> Under Win****, I used "TSDoctor", but I'm now working on
> Linux only and didn't find anything appropriate.

That's probably what I read about.

> But it seems ffmpeg is then enough for this purpose (I'm not expecting
> "very damaged" recordings).
>
>> To "insert silence", FFmpeg needs the "-async 1" option or the
>> corresponding filter chain.
>
> Very good hint, thanks!
>
> Just a small point I'm curious about. Help says:
>
> -async             <float>      ....A... simplified 1 parameter audio
> timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum
> stretch/squeeze in samples per second) (from INT_MIN to INT_MAX) (default 0)
>
> Why does it say "float" and at the end "from INT_MIN to INT_MAX"? It's
> also defined like that in the code in libswresample/options.c, also for
> other parameters.
>
> {"async"                , "simplified 1 parameter audio timestamp
> matching, 0(disabled), 1(filling and trimming), >1(maximum
> stretch/squeeze in samples per second)"
>                                                          ,
> OFFSET(async)          , AV_OPT_TYPE_FLOAT ,{.dbl=0
> }, INT_MIN, INT_MAX   , PARAM },
>
> A bug, or how does it make sense to use INT border values with a float?

I don't know but I believe this values are already too large for real
use cases.

Carl Eugen