[FFmpeg-devel] ffmpeg logging on stderr

Måns Rullgård mans
Thu Mar 10 15:35:17 CET 2011

Stefano Sabatini <stefano.sabatini-lala at poste.it> writes:

> On date Thursday 2011-03-10 01:15:40 +0000, M?ns Rullg?rd encoded:
>> Stefano Sabatini <stefano.sabatini-lala at poste.it> writes:
> [...]
>> > We could print it just before the transcode() function (or the ff*
>> > corresponding function) is called, but since the parsing is done
>> > before you could have the app printing an error and exit, in this case
>> > you don't have the version message in the error log (which is annoying
>> > in case you are reading a bug-report).
>> >
>> > Possible solutions:
>> > * implement pre-parsing of the commandline (overkill?)
>> > * remove the version header, and ask the users to always report ff*
>> >   -version when reporting bugs / asking for help (and you can bet that
>> >   they mostly won't)
>> We should at least make it go away with -v 0 (or maybe -1).  The amount
>> of console spam spewed forth by ffmpeg these days isn't even funny.  All
>> the useless metadata dumps didn't make it better.  In 99% of cases I
>> don't care whatsoever about what muxer some file claims to have been
>> created with etc.  If I need to know that specific bit of information, I
>> know how to find it.
> Yes I agree. I have some ideas, for example:
> always use av_log for logging in ffmpeg / other tools, advantages:
> fancy coloring, the users don't have to care about the distinction
> between application and library logging (why should they?), and we
> could drop -v. Currently we're mixing av_log with fprintf(stderr) with
> no criteria.
> Then we could preparse the -loglevel option *or* a -q option (easily
> done with few lines of code, shared between all ff* tools), which
> enables skipping to print the header.
> What do you think?

Sounds reasonable.

M?ns Rullg?rd
mans at mansr.com

More information about the ffmpeg-devel mailing list