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

[Louis Dionne via libcxx-commits]

ldionne wrote:

^ FWIW I really like that option! I'd be curious to know if that fixes @gribozavr 's issues.

Like I said, I don't have strong push back about working around this in libc++ but I'd want a public issue with CUDA to be referenced here and an acknowledgement that they're willing and working on fixing the issue. If that's the case I see no point in making people's lives hard and we should work around it here or in the cuda wrapper headers.

Otherwise, in case CUDA says something like "yeah we don't really care", then we'd be bending backwards to support something bad that another vendor is doing in complete awareness and I would push back on doing that on the basis that it's not a good precedent to set nor a sane long term direction for libc++.

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