[libcxx-commits] [libcxx] [libc++] Protect the libc++ implementation from CUDA SDK's `__noinline__` macro (PR #73838)

[via libcxx-commits]

https://github.com/EricWF approved this pull request.

> ```c++
> 
> #define _LIBCPP_NOINLINE _
> ```
OK, I was wrong about this, because we don't know if CUDA has been included yet.

The original solution is very much appropriate for the stated goals, and goes with existing practice. Once we get the formatted pass I say we land it.

It's clear to me that of the proposed solution, the one that keeps `_LIBCPP_NOINLINE` having semantics  seems the best.
I don't understand the objections that suggest we break the macro, even though we have the infra not to.


Let's do the useful thing with this patch and then continue our beef with nvidia elsewhere. We agree they're naughty, so lets add them to the naughty list.



https://github.com/llvm/llvm-project/pull/73838