#[no_mangle] is silently ignored on generic fns

[rpjohnst]

With a definition like this:

#[no_mangle]
fn foo<T>() { /* ... */ }

the compiler does not and cannot un-mangle foo, but it does not give any indication of this. It should probably give an unused_attribute warning.

[huonw]

I think an explicit "generic functions must be mangled" warning/error would be better than just the unused attribute one (this should cover #[symbol_name = "..."] too).

[sfackler]

Ah, this is one of the attributes that are globally whitelisted because the unused attribute lint runs before trans :(

[steveklabnik]

Triage: no change. Still doesn't warn.