remapping fortran math functions

[Jack Howarth <howarth at bromo dot msbb dot uc dot edu>]

    Over 18 months ago Andrew Pinski proposed an incomplete
patch to address the long double related failures in the
gfortran testsuite on Darwin PPC. I've reposted a version
to gcc-patches with a few requested corrections from the
original reviews...

http://gcc.gnu.org/ml/gcc-patches/2007-05/msg01809.html

Unfortunately Andrew never followed up with the missing
fortran front-end patches. Looking at the fortran front-
end, I am baffled at exactly how we should integrate this.
The patch remaps the C99 long double math functions as
listed in gcc/gcc/builtins.def. However it seems that
for the fortran front-end the f90 math functions are
obtained from gcc/gcc/fortran/mathbuiltins.def.
   The DECL_ASSEMBLER_NAME remapping in  patch
isn't honored and for large_real_kind_2.F90, we
still end up with symbols for cexpl rather than
cexpl$LDBL128. Can any of the fortran developers
explain how the fortran front-end manages to
remap exp() to expl() in the first place (since
it isn't listed in mathbuiltins.def)? For this
patch to work, somehow we have to have the fortran
front-end work with the names used in builtins.def
(which are the ones that get remapped). Thanks in
advance for any help on this.
            Jack