https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99708
--- Comment #24 from Jakub Jelinek <jakub at gcc dot gnu.org> --- (In reply to Segher Boessenkool from comment #22) > In rs6000_type_string, please just handle the error !type_node case first, > so you don't have to consider it in all other cases separately. Ok, will do. > Do you need to change the __SIZEOF_FLOAT128__ code at all now? I don't > see it? Yes, because the macro is from name about whether __float128 can be used and what it's size. On powerpc*, __float128 is a macro defined to __ieee128 under some conditions, so the condition to define __SIZEOF_FLOAT128__ should be that the macro is defined and __ieee128 can be used.
