Re: Accessing C++ function with Mixed Upper/Lower case name from F90 subroutine

[Toon Moene <toon at moene dot org>]

[ Copied to the GNU Fortran mailing list - probably a better forum for
  this question ]

Harald Servat wrote:

> Hello Glenn,

> >   I have a problem trying to prepare a large simulation package, using
> > mixed F90 and C++ routines, to compile and build under GCC (g++, and
> > gfortran).  The problem is I have a C++ function: Gauss() which is
> > called from an F90 routine:
> >   x=gauss()
> >
> > This package was made able to  build under the Intel fortran compiler by
> > adding to the fortran file the Intel compile directive:
> >
> > !DEC$ ATTRIBUTES ALIAS:'Gauss' :: gauss
> >
> > I attempted to build this program using GCC  where for F90 code the
> > gfortran compiler option -fno-underscoring was used.
> >
> > Presently my fortran routine won't build. It gets the error:
> >
> > 485: undefined reference to `gauss'
> >
> > So, my question is, is there a way in GCC to enable the linker to find
> > the C++  'Gauss' method?
>
>   I don't know if the linker supports for this directly. But you can try
> with the function attribute 'alias("target")' which is documented in
> [1]. It creates a new name for the same routine, so you can call it
> through the two names (the original, and the aliased).
>   In your case, you should define in your C++ file an alias to "Gauss"
> which is called "gauss".
>
>   Besides, IIRC, ICC also supports this function attribute, so you
> should be able to apply it transparently.
>
> [1] http://gcc.gnu.org/onlinedocs/gcc/Function-Attributes.html

The Standard (as of Fortran 2003, supported by GNU Fortran and ifort) 
way to do this is to add the following to your Fortran code:

$ cat a.f90
interface gauss
function gauss() bind(c, name="Gauss")
use, intrinsic :: iso_c_binding
real(c_float) :: gauss
end function gauss
end interface gauss
x = gauss()
end
$ gfortran -c a.f90
# The command 'nm' will show you that the right external name for the
# "gauss" routine has been encoded in the object file:
$ nm a.o
                 U Gauss
0000000000000000 T MAIN__
                 U _gfortran_set_options
0000000000000000 r options.0.1534

Hope this helps,

--
Toon Moene - e-mail: toon@moene.org - phone: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
At home: http://moene.org/~toon/
Progress of GNU Fortran: http://gcc.gnu.org/gcc-4.5/changes.html