[PATCH] D112081: Define __STDC_NO_THREADS__ when targeting windows-msvc (PR48704)

Tue Oct 19 10:56:58 PDT 2021

hans added a comment.

In D112081#3073067 <https://reviews.llvm.org/D112081#3073067>, @aaron.ballman wrote:

> Thanks for this! Should we similarly handle `__STDC_NO_ATOMICS__`, `__STDC_NO_COMPLEX__`, et al at the same time?

I'll look into those, thanks.

> Also, how should we handle `__STDC_VERSION__`? We set that in all C modes, but MSVC only sets it when passed /std:c11, etc explicitly.

In any case we should probably do it separately. I don't really like MSVC's model where it's in a "traditional C with extensions" mode by default, but I clang-cl already also kind of does that because we don't define __STDC__ when targeting Windows. But it looks like unlike MSVC we also don't define it when passing /Za. And, I wonder how /Za and /std: are supposed to interact.

In D112081#3073075 <https://reviews.llvm.org/D112081#3073075>, @aaron.ballman wrote:

> Also, one concern I have for this is what to do if/when the Microsoft CRT gets any of these features. Clang will be unconditionally setting `__STDC_NO_WHATEVER__`, which we can correct for newer versions of Clang, but still means older versions of Clang on those updated systems then report something nonsensical.
>
> It'd be nice if Microsoft had a solution similar to stdc-predef.h used by glibc, so the CRT could define information that we could use to determine what macros to predefine rather than having to guess.

I think the best we can do is make updates in future versions of Clang.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D112081/new/

https://reviews.llvm.org/D112081