How to make port accept cmake cached variable

[vipcxj]

For example, I need set CMAKE_CUDA_HOST_COMPILER to a special g++ version. Of course it works on my direct project. However it is not send to the port my project used. As a result my port always use default g++. And my cuda not accept my default g++. As a result build failed.

[Osyotr]

You're should either pass -DCMAKE_CUDA_HOST_COMPILER=... in VCPKG_CMAKE_CONFIGURE_OPTIONS triplet variable or write a CMake toolchain file that is used by both vcpkg and your project.
https://learn.microsoft.com/en-us/vcpkg/users/triplets#vcpkg_cmake_configure_options
https://learn.microsoft.com/en-us/vcpkg/users/triplets#vcpkg_chainload_toolchain_file