martell added a comment. > ! In https://reviews.llvm.org/D33852#772290, @davide wrote: > This assumption is not quite right. Also GCC, when targeting Linux, provides > a `-fms-extensions` flag. > There's a bunch of software out there that unfortunately has to target ELF > and build with `-fms-extensions` on, e.g. PS4.
Seems like bad practice but on a second look that is does appear that PS4 and CUDA both use `-fms-extensions` and rely on it for some things. argh CUDA does have a note to move off this in future. Thanks for clarifying davide This patch does LGTM minus the nit above. Adding `selectany` spelling to `-fdeclspec` as well as `-fms-extensions` should cover all use cases. The MachO issue would have been present before `rL300562` so I don't think that should have to be fixed here. It could be covered in a follow up differential as this fixes a regression on my part. ================ Comment at: test/Sema/attr-selectany.c:3 // RUN: %clang_cc1 -triple x86_64-mingw32 -verify %s +// RUN: %clang_cc1 -triple x86_64-unknown-linux -verify -fms-extensions %s ---------------- Based on the current changes you can change `-fms-extensions` to `-fdeclspec` The `-fms-extensions` coverage is covered in `attr-selectany.cpp` below https://reviews.llvm.org/D33852 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits