On Thu, 29 Oct 2015, Segher Boessenkool wrote: > There probably should be an effective-target for float128.
Yes, as I said in <https://gcc.gnu.org/ml/gcc-patches/2015-10/msg02991.html> there needs to be standard test boilerplate to: disable a test if __float128 isn't supported at all even with extra options added; add those options on targets that support it provided the right options are added; and turn a run test to a compile test if the options are supported but the hardware support (for VSX in this case) is missing. Then the tests that run for x86 and possibly ia64 targets (some of which may be under gcc.target at present) need to have their target restrictions replaced, being moved out of gcc.target as needed. The tests using feenableexcept would then *also* require some suitable effective-target (or restriction to *-*-*gnu*) to restrict them to glibc targets. -- Joseph S. Myers jos...@codesourcery.com
