[FFmpeg-devel] [PATCH] Refactor OBJS handling in libavcodec/Makefile

[Diego Biurrun]

On Sat, Jul 11, 2009 at 11:20:23AM +0100, M?ns Rullg?rd wrote:
> Diego Biurrun <diego at biurrun.de> writes:
> 
> > On Thu, Jul 09, 2009 at 12:24:04PM +0100, M?ns Rullg?rd wrote:
> >> Diego Biurrun <diego at biurrun.de> writes:
> >> > On Thu, Jul 09, 2009 at 11:15:17AM +0100, M?ns Rullg?rd wrote:
> >> >> Diego Biurrun <diego at biurrun.de> writes:
> >> >> > Notice that an alternative would be to have some decoders depend on
> >> >> > others.  For example, the vc1_vdpau decoder could be made to depend on
> >> >> > the vc1 decoder in configure.  Then it would not need to duplicate the
> >> >> > OBJS list.  However, this is not currently possible in configure.
> >> >> 
> >> >> Why not?  Is vc1_vdpau_decoder_deps=vc1_decoder not enough?
> >> >
> >> > No, if you run
> >> >
> >> >   configure --disable-decoders --enable-decoder=vc1_vdpau
> >> >
> >> > no decoders are enabled.
> >> 
> >> I can fix that.  Give me a moment.
> >
> > Moment granted :)
> 
> The attached patch make the bulk disabling of things weaker, allowing
> a _select on something else to re-enable it.  Explicitly disabled
> things stay disabled, however, making any _select fail as before.
> 
> What do you think?

It's looking good.  IOW

  configure --disable-decoders --enable-decoder=vc1_vdpau

will enable the vc1 decoder, but

  configure --disable-decoders --disable-decoder=vc1 --enable-decoder=vc1_vdpau

will not enable the vc1 decoder.

So if users pass conflicting options, we err on the side of disabling
things.  Which semantics we use exactly is a matter of taste, but I'm
fine with the way your patch handles it.

Diego