Re: Fortran name-mangling scheme fails on HP-UX 11.11

[Steven G. Johnson]

On Fri, 29 Jun 2001, John W. Eaton wrote:
> On 28-Jun-2001, Steven G. Johnson <address@hidden> wrote:
> | Well, then you have the opposite problem: figuring out whether there are
> | any special flags or libraries that you need to pass to the Fortran
> | linker in order to link to C programs.  (Generally, the answer seems to be
> | no, but can we count on that always being true?)
> 
> No, if you include C++ compilers in with C compilers.

Ah, good point, I forgot about C++ and its many wonders.

Since the name-mangling detection wrappers use the C compiler, in
principle they don't have to worry about that and could use the Fortran
linker.  But, in any case, it is important to have a working
AC_F77_LIBRARY_LDFLAGS for other link tests and for the end-users (like
Octave).

> I think the macros in Autoconf were originally derived from the ones I
> wrote for Octave, and eventually I would like to drop my macros and go
> with the ones from Autoconf.  Please don't change things to make that
> a difficult or impossible thing to do.

Convenience for existing programs is definitely a goal of mine as well (I
use cross-language linking extensively), so please let me know if you
encounter any shortcomings with the autoconf macros.

(The AC_F77_LIBRARY_LDFLAGS macro was indeed derived from Octave.  The
Fortran name-mangling detection was derived from macros I wrote for FFTW.
The dummy MAIN__ routine detection is new (Octave uses compiler-specific
#ifdefs at the moment).  I'm not sure where AC_PROG_F77 and friends comes
from.)

Steven