https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119408
--- Comment #20 from chenglulu <chenglulu at loongson dot cn> --- (In reply to Xi Ruoyao from comment #19) > (In reply to Jakub Jelinek from comment #18) > > _Float128 is definitely not for backward compatibility > > Sorry, I mean __float128. > > The problem here is we added __float128 as an alias of _Float128 for > compatibility with existing code (of which the author only considered x86) > and then added q/Q for such a compatibility too. And then we started to > build libquadmath (Qiankang said so off-list) and libgfortran.spec started > to refer it. It's not a good thing IMO, as we are duplicating sinl, cosl, > ... from libm into sinq, cosq, ... from libquadmath, and we are making > fortran programs redundantly depending on libquadmath. > > So IMO perhaps we should gate __float128 and q/Q behind a switch. Sorry, I was confused. What I meant was __float128. Can we remove this support or revert r15-8962 in the form of a bug fix?
