On Wed, 17 Nov 1999 11:39:39 -0500 (EST) mcmahill@mtl.mit.edu wrote:

> Problem:
>   - The in tree fortran compiler (/usr/bin/f77) is buggy (see PR's 8574
>     and 8663 for details).
> 
>   - Several of our packages use fortran and contain logic local to the
>     package Makefile which adds the dependency on f2c for systems
>     (pre-1.4) which don't ship f77.
> 
> Proposal:
> 
>   I'd like to create a variable 'USE_FORTRAN' in bsd.pkg.mk which would be
>   set in the package Makefile for those packages which use fortran.  That
>   way the logic pertaining to the fortran compiler, possible dependency on
>   f2c, and selection of the fortran compiler are done in 1 central
>   location.

...

>   The remaining question I have is if we should make f2c be the default
>   for now since f77 is buggy.  Long term, the goal should be to not ship a
>   buggy compiler.  I could perhaps implement f2c being the default by
>   something like:
> 
>   USE_F2C?=	YES
>   .if defined(USE_F2C)
>   .if ${USE_F2C}=="NO"
>   .undef USE_F2C
>   .endif
>   .endif
> 
>   Then users could put USE_F2C=NO in mk.conf to override the f2c default.

I know that it is current style to use several binary variables in
bsd.pkg.mk. But wouldn't it better to change USE_F2C to USE_F_CPL and
set it to "f2c" or "f77" to enforce a specific fortran complier. Yes it's
paranoid to assume we get more than two fortran complier, but who knows.

Or should we reserve USE_ for variable usable in mk.conf and reserve i.e
NEED_ for the package to request a tool class. In the case of fortran
this will change USE_FORTRAN to NEED_FORTRAN and USE_F2C to USE_FORTRAN.
  
BTW for MAKE and CONFIGURE this would also apply, but is it worth the
work to change that now?

Have a nice day
Bernd