This is a little more complicated since I believe this should still be allowed:

function equivalent(): void;
function equivalent(): void {}

I don't think we actually need to allow this. Spicy doesn't need functions to be declared before usage, i.e., no need for prototypes if the implementation can be found somewhere else. The main use case for a body-less prototype are &cxxname functions. Unless I'm missing something ...

I don't think we actually need to allow this

Probably not, I assumed since it was valid that there was more use. Does that mean we can just decline any prototypes without &cxxname then?

FWIW, I think this is still necessary for hooks, and is why my initial implementation wasn't in the initial PR for overloading - I ran the tests, saw failures that looked valid, and stopped. But if we just allow equivalent hooks it seems fine :) Like this from try-across-hooks:

declare public hook void f1();

public hook void f1() &priority=5 { hilti::print("Foo: f1"); }

This can probably done pretty easily after or alongside #1985 - I'd prefer a diagnostic on non-hook prototypes without &cxxname too if that seems reasonable

Probably not, I assumed since it was valid that there was more use. Does that mean we can just decline any prototypes without &cxxname then?

Yes, I think so (assuming there's no other attribute I'm missing right now that would need prototypes too; but as long as the test suite passes we should be fine).

FWIW, I think this is still necessary for hooks,

Yeah, right, because hooks may have no implementation at all.

But if we just allow equivalent hooks it seems fine :)

+1