[FFmpeg-devel] [PATCH 6/6] Add rotate90 filter.

Aurelien Jacobs aurel
Fri Oct 15 17:54:35 CEST 2010


On Fri, Oct 15, 2010 at 04:54:57PM +0200, Stefano Sabatini wrote:
> On date Friday 2010-10-15 14:36:35 +0200, Aurelien Jacobs encoded:
> > On Fri, Oct 15, 2010 at 01:13:59AM +0200, Stefano Sabatini wrote:
> > > ---
> > >  doc/filters.texi           |    6 ++
> > >  libavfilter/Makefile       |    1 +
> > >  libavfilter/allfilters.c   |    1 +
> > >  libavfilter/vf_transpose.c |  173 ++++++++++++++++++++++++++++++++++++++++++++
> > >  4 files changed, 181 insertions(+), 0 deletions(-)
> > 
> > What about moving this code to its own vf_rotate90.c file (and move
> > shared code in a shared file) ?
> 
> Slightly against this, since it requires more complexity (internal
> function declarations) with no gain,

I feel that it is your current patch which require more complexity
(ifdefery) with no gain. (yes that's a loose argument...)

> also the invert() and transpose()
> functions require the same formats, and I cannot see a safe way to
> expose the query_formats() function.

Huh ? What's the problem with a ff_transpose_query_formats() in a common
transpose.c file used by both filters ?

My point is that we are currently splitting muxers and demuxers which
were written as a single file long ago, and I would like to avoid having
to do the same with libavfilter in a few years...
For now, as long as you only merge 2 filters, it's not much of a
problem. But if you start like this, we will end up with intermingled
mess of ifdef, many filters in the same files, loosely defined
dependencies, code which is always compiled even if it's not used,
etc...

I can't imagine why it would be more problematic to properly split
filters, than to properly split (de)muxers.

Aurel



More information about the ffmpeg-devel mailing list