CMakeDetermineCompilerId: Can not pass -allow-unsupported-compiler flag in CUDA
May be because of this change, !5863 (merged), I could not pass -allow-unsupported-compiler
flag in CMakeDetermineCompilerId for CUDA.
This causes an error in CUDA's host compiler check in some cases. I'm currently using CUDA11.1 on Ubuntu 22.04, so the default GCC version is higher than the supported one and cause this error.
One of the quick workarounds is to use update-alternatives
to change the /usr/bin/gcc
to gcc-10
, but it seems to be not a clever solution.
Is there any way to force use specified options in CMakeDetermineCompilerId
?