[MPlayer-dev-eng] configure force enable all

[Ivan Kalvachev]

On 7/26/05, Reimar Döffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> Hi,
> On Tue, Jul 26, 2005 at 09:21:22PM +0300, Ivan Kalvachev wrote:
> > Look at the attachment and tell me what do you think.
> >
> > The code is not finished yet (actually not sure if it works at all),
> > but you should get the idea.
> 
> Well, not really. Maybe describe in a few words what you want to
> achieve...
> E.g. setting _default_no=auto will not owkr I think, since most of the
> options that are disabled by default simply have no autodetection...
> There is at least a high risk of breaking things I think.

OK, one day passed.
I would like to summarize the prevous flamewars.

The current configure uses the following meaning for --enable--foo:
If option is autodetected by default, --enable will force it.
if option is disable by default, --enable will autodetect it.

configure script autodetects most of the stuff, so --enable usually
mean --force.
Of course if somebody wants to build an minimum featued mplayer (free
of bloat;), he have to use multiple --disable-foo options. With adding
new options this list grows.

I have created 2 new options --disable-all and --enable-all. The first
one disables all optional features. The second one enables
autodetection for all usually disable options.
On the end of second configure flamewar all had agreed that these
*-all options are good idea.

But this have an side effect. If we use --disable-all, we are no more
able to autodetect options as according to the first rule --enable
will force the options.

So I do an cleaning of the options by changing behaviour of --enable.

In the configure script option have 3 states "yes" "auto" "no".
These states now corespond to --force , --enable, --disable. If the
options is set to "auto" then the detection will set it to either
"yes" or "no" state. If there is no detection then it is set to "yes".

The only exeption is --enable-bigendian, that i removed and instead
made --force-bigendian and --force-littleendian. Anyway I believe that
this should go in the first pass options after --target and --host-cc
options.

I haven't changed the mmx/altvect flags (yet).

What is to be done in order the patch to be finished: 
Of course following all options and make sure that detections are
following the new standart. Currently there are functions that
autodetect on " !=no". There are probably also options that doesn't
change 'auto' to 'yes' if no autodetection is available. Also
--enable-all is not tested if it includes stuff that shouldn't be
included (e.g. suboptions, interdependable options etc...).
But there are no fatal problems.

Happy Coding
  Ivan Kalvachev
 iive