[MPlayer-dev-eng] [RFC] configure2 underway

Ponekker Zoltan pontscho at makacs.poliod.hu
Thu Nov 15 14:49:43 CET 2001


Hali

Please look some yes, or no in this config.h :)

Pontscho

On Thu, 15 Nov 2001, pl wrote:

> Hi,
>
> As some (any) of you may be aware, as time goes on, the configure script
> becomes a real pain to extend:
>   - where do I add my stuff ?
>   - several parts to modify in the scripts:
>       before args parsing
>       after
>       default values
>       tests
>       C and makefile #ifdef -llibfoo generation
>   - no common naming style : sometimes _symbol is a yes-no-other value,
>     sometimes it becomes a '#ifdef fooo' sometimes it's a library
>
>
> I've started cleaning things up so as to gather things a bit and make it
> cleanier:
>
>   - preparsing: set standard default values: yes no auto
>   - options parsing: add the --enable-opt and --disable-opt in a row
>     Such options forces a value and disable the auto checks
>     (The l^huser knows what he does)
>
>   - test blocks handle 1 functionnality and look like:
>
>     echo -n "checking opt ..."
>     if test "$_opt" = auto ; then
>        # do C check here or whatever
>
>        $_opt=no
>        testsuccess && $_opt=yes
>     fi
>     if "$_opt" = yes ; then
>       _def_opt='#define HAVE_OPT 1'     # or #define USE_OPT 1 or
> 					# whatever is used for this
> 					# peculiar option
>       _inc_opt=' -I/usr/foo/bar/include '
>       _ld_opt=' -L/usr/foo/lib -lopt '
>     fi
>     echo "$_opt"
>
>   - defines naming style for include or gcc flags
>
>   - You have a progressive test that is run (a la GNU autoconf) and not
>     a "please wait" followed by a burst of 50 lines indicating the
>     result.
>
> Notes:
> ~~~~~~
> I won't commit to CVS till it works (at least for me :) :
>  - it is broken (does not generate correct config files)
>  - it has no longer any common architecture bigger than blocks with
>    standard configure so the diff is really huge
>  - I do _not_ want bugreports right now till it's not working/cleaned
>    the way I want
>  - even if I commit to configure2 they're will be much useless commits
>    since many parts are still "dirty"
>
> It has been a fork of sunday's version so the options added since that
> have not yet been added.
>
> I think it is cleanier and simpler but that's only my view so...
> I attached a snapshot of my file asking for the judgement of the MPlayer
> -dev.
>
> The snapshot just performs tests and create config files (they're not
> usable yet).
>
>
> What I expect now is : a "Go ahead !" or a "Forget that !" reply.
> (No bugreports please)
>
> Of course criticisms are welcome (ahem... ;)
>
> --
> "Janitor-ly" yours,
>   pl
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: config.h
Type: text/x-chdr
Size: 6289 bytes
Desc: 
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20011115/585caa95/attachment.h>


More information about the MPlayer-dev-eng mailing list