Enable __float128 for mingw for GCC compatibility and define…


Enable float128 for mingw for GCC compatibility and define SIZEOF_FLOAT128__ on x86

GCC defines FLOAT128 on Power and SIZEOF_FLOAT128 on x86. We're
just following the inconsistency for now so users have some way to test.

Effectively merges this patch as requested by Martell Malone:


rnkMay 26 2017, 10:38 AM
rL304011: Revert r304002 "[DWARF] - Make collectAddressRanges() return section index in…

Event Timeline

@mati865 This should probably be back ported into msys2's mingw-clang now that it has landed with extra support for __SIZEOF_FLOAT128__ replacing the original patch

mati865 added a comment.EditedMay 26 2017, 12:17 PM

Some time ago I both enabled __float128 and defined macro but using it (__float128) cased segfaults. So I reduced it to HasFloat128 and haven't upstreamed it due to broken tests. It's good to see things improved.

@martell You can backport it, I'm totally out of time due to assignments at college and searching for practice (required to graduate).

rnk added a comment.May 26 2017, 12:59 PM

Make sure to also backport rL304013, it was intended to be part of this.

@mati865 I will have a look at the weekend.
@rnk got it :) I also closed the corresponding bug report https://bugs.llvm.org/show_bug.cgi?id=30685