>>>>> "Abdelrazak" == Abdelrazak Younes <[EMAIL PROTECTED]> writes:

Abdelrazak> Helping the port to Scons is just a side effect. The
Abdelrazak> problem is that config.h is not only used for supposed
Abdelrazak> portability but also to pass some settings that is used
Abdelrazak> somewhere in the source code. For example, could you tell
Abdelrazak> me why every source code should know that lyx will have
Abdelrazak> Aspell?

Can you tell me how it hurt?? We have to check for aspell anyway, and
there is no reason that parsing one more #define should make you loose
sleep... 

Abdelrazak> I think you will achieve the same portability is you limit
Abdelrazak> the include of config.h to C file in src/support/ only.
Abdelrazak> Every use of config.h defined macro should be transfered
Abdelrazak> there (the majority are there already actually).

So, how do we tell whether asserts are enabled or not? How do we know
whether #warnings should be shown (although these should probably be
replaced with plain FIXMES)? How do we know whether NLS is enabled?
How do we...?

I understand your concern about configure taking too much time on
mingw (if I understand correctly). However, 90% of this time is due to
the gettext of libtool scripts, and we are not in a position of
getting rid of them. Therefore, all optimizations you will be able to
do are micro-optimizations.

JMarc

Reply via email to