Clarify whether `#[repr(C)]` can be used on tuple structs or not.

[nwin]

Until now I had the impression that it would be undefined how this construct would behave. The Rustonomicon implies that this is still the case, as tuple-structs are just like “named tuples”. On the other hand, the reference implies that it should be safe to use #[repr(C)] since tuple structs behave like normal structs with anonymous fields.

I would appreciate if this would be clarified in the docs. If it is still undefined behavior #[repr(C)] on a tuple struct should cause an compiler error.

[eefriedman]

It's well-defined; perhaps the docs aren't as clear as they could be.

[Aatch]

Just to be clear, tuples and structs are handled pretty much identically. There are two properties, named type and named fields that distinguish the 3 options:

	Type Named?	Field Named?
struct Foo { a: i32 }	Yes	Yes
struct Foo(i32)	Yes	No
N/A (Would be "record" or similar)	No	Yes
(i32, i32)	No	No

In theory, there is no reason you couldn't apply #[repr(C)] to tuples, except that there's no source-code definition for them.

[steveklabnik]

/cc @rust-lang/lang, is this well-defined? I'm guessing so, but I'd like to be sure before I document it.

[nikomatsakis]

I see no reason to disallow it. I vote "not a bug".

[nrc]

Agree. Change to a docs bug?

[steveklabnik]

Sounds good, I'll document it.