Re: fortran 90

[Steven G. Johnson]

Kate Hedstroem wrote:
> I searched the macro archive and found no mention of f90. Has anyone
> ever talked about using autoconf and/or automake with f90?

Yes, it's been discussed.  Martin Wilck in particular (address@hidden)
has done some recent work on f90 macros:

http://sources.redhat.com/ml/autoconf/2000-06/msg00280.html
http://sources.redhat.com/ml/autoconf-patches/2001-07/msg00027.html

(This was a good start, but I think that eventual support for f90 in
autoconf should try harder to share common code with the f77 support, so
that we don't have essentially duplicate AC_F77_WRAPPERS/AC_F90_WRAPPERS,
etcetera.  Especially since f95 is becoming more widespread...I don't
know quite what to do about all of these Fortran versions.)

The existing autoconf f77 macros should work fine if you force AC_PROG_F77
to find an f90/f95 compiler:

        AC_PROG_F77([f90 xlf90 pgf90 epcf90 f95 fort xlf95 lf95 g95])

(It already searches for these compilers if it doesn't find an f77
compiler.)

The real trick is to support using both f90 and f77 simultaneously, which
is arguably necessary since the languages are so different... (And f95 and
f**...??  Maybe just f77 and f.ge.90 is enough?)

> One major issue with f90 that isn't in f77 is the module feature. At
> compile time, the information needed to use a module is placed in a file
> with extension .mod, .kmo, .M or even in the object file.

I haven't seen any macros of this sort, but that sounds like the sort of
thing that could be detected nicely by a self-contained autoconf macro and
contributed to the macro archive.  (hint, hint)

(Sigh, yet another Fortran feature that seems impossible to portably link
to C or any other language...)

Cordially,
Steven G. Johnson