The language and compiler & linker should reject Zalgo in identifiers, and they should reject confusable script mixes in identifiers, but otherwise they treat all equivalent strings as equivalent. To make it easier on the linker compilers should normalize all symbols to one common form (e.g., NFC).

C does allow for limited unicode in identifiers, though you need to use the \u prefix and write the code out. Compilers like clang let it work like C++ and follow TR31, though this is nonstandard.