Hi,

I recently started using some of the sorting algorithms that you so nicely provide in cpp-sort. These work perfectly fine in a CPU-only CMake build. However, when a CUDA target links to cpp-sort, the MSVC flags /permissive- and /Zc:preprocessor (see here) get propagated to the nvcc compilation command without getting wrapped inside -Xcompiler. This makes my GPU library fail compilation.

I wanted to inquire as to why these flags are necessary and if we can make them optional in some way, for example, via a generator expression (perhaps COMPILE_LANGUAGE).

Thank you