It looks like #713 reported a bug with new compilers + old stdlib relying on a std::byte that does not exist. The fix in #714 tries to resolve that by checking __cpp_lib_byte. It seems like it does not pull in <cstddef> before doing so, though. I am seeing this manifest as an inconsistent definition of gsl::byte throughout my project. In places where I happened to include <cstddef> before <gsl/gsl_byte>, it becomes an alias for std::byte. In places where I did not include <cstddef>, gsl::byte is a unique type. These two objects then fail to link together when they have gsl::byte on their interface and they don't agree on its implementation.
I've fixed this for now on my side by doing an explicit #include <cstddef> everywhere before including <gsl/gsl_byte>, but I think it would be nice to have this fixed.
It looks like #713 reported a bug with new compilers + old stdlib relying on a
std::bytethat does not exist. The fix in #714 tries to resolve that by checking__cpp_lib_byte. It seems like it does not pull in<cstddef>before doing so, though. I am seeing this manifest as an inconsistent definition ofgsl::bytethroughout my project. In places where I happened to include<cstddef>before<gsl/gsl_byte>, it becomes an alias forstd::byte. In places where I did not include<cstddef>,gsl::byteis a unique type. These two objects then fail to link together when they havegsl::byteon their interface and they don't agree on its implementation.I've fixed this for now on my side by doing an explicit
#include <cstddef>everywhere before including<gsl/gsl_byte>, but I think it would be nice to have this fixed.