How to avoid bugs in nvcc from CUDA Toolkit with new and old unsupported GCC and Clang?

To avoid new requests and bugs to support in cuda toolkit for new versions of gcc (ex 11.1.0) and clang (ex. 12.0.0) please open some parts of nvcc from toolkit to possible patch myself toolkit and compile nvcc with all future changes in gcc and clang.

I’m not clear on the question, but are you asking for a way for nvcc to not give an error when using an unsupported compiler?

If so, then you can use the “nvcc --allow-unsupported-compiler” flag.