#[repr(C)] silently accepted on ADT enums

[durka]

This code compiles without error or relevant warning:

#[repr(C)]
enum Foo { A { x: i32 }, B(i32), C }

#[repr(i32)]
enum Bar { Bar(i32), Baz(u32) }

But obviously the representation of the enum is still going to be undefined, and not integral or C-compatible (trivially checked by mem::size_of in this case).

There should be an error (probably infeasible) or at least a lint that is on by default.

@Aatch says there is already an issue about this, but I can't find it, so feel free to close as a duplicate if you can find it :)
cc @niconii

[nodakai]

Take a look at #25651

[durka]

Okay, so #[repr] on an enum controls the size of the discriminant, not the whole enum. At least #[repr(C)] should warn IMO because #[repr(C)] means "give this thing a defined layout", but that doesn't happen in this case.

[hanna-kruppe]

rust-lang/rfcs#2195 makes this do something useful so if/when that's accepted this issue can be closed I guess.