https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78478
--- Comment #16 from Joel Sherrill <joel at gcc dot gnu.org> --- Thanks for all the feedback. With this patch, it now builds. Is the style of change to configure.host OK? I need to check how far back this impacts. A user reported it with a released gcc. Since we don't build fortran by default, this might have been broken a while. diff --git a/gcc/config/i386/rtemself.h b/gcc/config/i386/rtemself.h index 9516c45..e8eade2 100644 --- a/gcc/config/i386/rtemself.h +++ b/gcc/config/i386/rtemself.h @@ -28,6 +28,3 @@ along with GCC; see the file COPYING3. If not see builtin_assert ("system=rtems"); \ } \ while (0) - -#undef LONG_DOUBLE_TYPE_SIZE -#define LONG_DOUBLE_TYPE_SIZE (TARGET_80387 ? 80 : 64) diff --git a/libgfortran/configure.host b/libgfortran/configure.host index 0a12a32..e594f31 100644 --- a/libgfortran/configure.host +++ b/libgfortran/configure.host @@ -37,11 +37,16 @@ if test "x${have_feenableexcept}" = "xyes"; then fi # x86 asm should be used instead of glibc, since glibc doesn't support -# the x86 denormal exception. +# the x86 denormal exception. But RTEMS uses newlib and this doesn't apply. case "${host_cpu}" in i?86 | x86_64) - fpu_host='fpu-387' - ieee_support='yes' + case "${host_os}" in + rtems*) ;; + *) + fpu_host='fpu-387' + ieee_support='yes' + ;; + esac ;; esac