[Ffmpeg-devel] Re: ffmpeg build fixes

[Diego Biurrun]

On Thu, Mar 22, 2007 at 11:13:53AM +0100, Stanislav Brabec wrote:
> Diego Biurrun wrote:
> 
> >> ffmpeg.redefined.patch: Fixed lots of repeated compiler warnings.
> >
> >What exactly is this supposed to fix?  I get no warnings related to this
> >code in FFmpeg ...
> 
> I see 976 warnings like these:
> internal.h:252:1: warning: "sprintf" redefined
> In file included from /usr/include/stdio.h:838,
>                  from common.h:36,
>                  from mathematics.c:26:
> /usr/include/bits/stdio2.h:30:1: warning: this is the location of the previous definition

I cannot reproduce it.  Compiling which file?  Which compiler etc?

> And I forgot to write another issue:
> Some distributions package libgsm includes to /usr/include/gsm. It would
> be nice to check this directory, too.

Pass --extra-cflags=/usr/include/gsm to configure.

> >> ffmpeg-system-amr.patch: Changes the code to build AMR support with
> >> system amr libraries from http://www.penguin.cz/~utx/amr
> >
> >This looks interesting, these libs are a pain to use.  The use of
> >autotools is unfortunate, but we'll forgive you for now ;)  Somebody
> >should package this stuff now..
> 
> I added simple spec file there, so rpmbuild -tb can do it for RPM based
> distros.
> 
> I have mailed all people signed in RPMs I found by random search.

Hopefully somebody will pick it up and create a Debian package soon :)

> The amr* code needs fix and I doubt that 3GPP will release new reference
> implementation. So there's also a chance to maintain patches.

It would be better to have a native implementation.  One was started at
last year's Google Summer of Code but unfortunately never finished:

http://svn.mplayerhq.hu/amr/

Spending too much time on the non-free reference implementation is
clearly a waste of time, but a properly packaged library will at least
allow getting rid of the hacks in FFmpeg.

Diego