I just finished helping a colleague track down some strange test failures that appear to stem from enum_name(E value) unintentionally behaving as if E were an empty enum. It took a bit to track down this specific issue because the code that actually used enum_name(E value) wasn't touched in the problematic commit. Some other unrelated shuffling of code appears to have confused the compiler so it started treating the enum in question as a plain forward declaration rather than its full definition, so enum_name(E value) was returning empty strings instead of the correct names.

This got me thinking - should enum_name(E value) work at all if the enum is empty? enum_name<auto V>() has a static_assert to bail if the returned name is empty, so changing enum_name(E value) to bail on empty enums and/or empty names would arguably make the behaviors match and could help catch unintentional changes like this where a previously-defined enum starts being treated as a forward declaration.

I think this question could also be extended to other parts of what magic_enum offers, but enum_name is definitely the place where I've experienced issues the most.

And if enum_name(E value) should work with empty enums, should it return a std::optional<std::string_view> to distinguish between the "no name due to invalid value/empty enum" and "empty name" cases?