http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55601
--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-12-10 08:41:36 UTC --- Hmm, seemingly, the check ignores --disable-libquadmath-support by only testing for C's __float support, cf. libgfortran/acinclude.m4's LIBGFOR_CHECK_FLOAT128. gfortran itself is checked via libgfortran/acinclude.m4's LIBGFOR_WORKING_GFORTRAN. One probably should augment the C test by a Fortran "REAL(16)" test. Additionally, one could directly honour --disable-libquadmath-support, which I think is also passed to libgfortran's configure. Regarding the --disable-libquadmath combined with --enable-libquadmath-support: Some warning would be useful, but the question is how to trigger it.
