[MPlayer-cvslog] r35939 - trunk/configure

[Reimar Döffinger]

On Sun, Mar 10, 2013 at 03:10:38PM +0200, Ivan Kalvachev wrote:
> On 3/10/13, reimar <subversion at mplayerhq.hu> wrote:
> > Author: reimar
> > Date: Sun Mar 10 13:30:54 2013
> > New Revision: 35939
> >
> > Log:
> > Allow compiling without linking to a system-specific xvmc backend library.
> >
> > Modified:
> >    trunk/configure
> >
> > Modified: trunk/configure
> > ==============================================================================
> > --- trunk/configure	Sun Mar 10 13:17:19 2013	(r35938)
> > +++ trunk/configure	Sun Mar 10 13:30:54 2013	(r35939)
> > @@ -4591,7 +4591,8 @@ EOF
> >  fi
> >  if test "$_xvmc" = yes ; then
> >    def_xvmc='#define CONFIG_XVMC 1'
> > -  libs_mplayer="$libs_mplayer -lXvMC -l$_xvmclib"
> > +  libs_mplayer="$libs_mplayer -lXvMC"
> > +  test -n "$_xvmclib" && libs_mplayer="$libs_mplayer -l$_xvmclib"
> >    vomodules="xvmc $vomodules"
> >    res_comment="using $_xvmclib"
> >  else
> 
> I think this is wrong.
> libXvMC doesn't implement all essential XvMC api functions, many of
> them were left for the specific videocard implementation
> library/driver one. (in other words... -lXvMC only, would break the
> build).

This patch certainly doesn't break anything more than before.
It only changes the case where we would end up with
"-lXvMC -l" to just "-lXvMC".
Which in addition can only happen if used with --enable-xvmc
Maybe it's not actually useful, but I can't see how it would be wrong.

> I think that the proper solution is to put XvMCW before XvMCNVIDIA or
> even always link to it. What do you think?

I think that is a completely unrelated issue, but it seems like a good idea.