Is [repr(C)] propagated to child members?

Question

I have a struct:

struct A {
    names: Vec<String>,
}

And another struct:

#[repr(C)]
struct B {
    pub a: A,
}

Does this make A also have the repr(C) annotation? Would it be correct to pass such structure to C code ?

Answer 1

Your answers are:

1. No
2. No

#[repr(C)] doesn't make structures "safe to FFI", it guarantees that the in-memory representation will be as is and the compiler will not try to optimize them for space/efficiency under the cover. This is accidentally what is also needed for safe FFI.

Answer 2

It certainly isn't safe to pass your structure to C code. There are some caveats in the documentation , including:

• DSTs, tuples, and tagged unions are not a concept in C and as such are never FFI safe

  You have no guarantee about whether any of these are in the Vec . Also, since Vec is opaque.

• If the type would have any drop flags, they will still be added"

  Again, Vec may have some.

And in any case, for it to make any sense at the receiving C code, you need to declare your struct in C as well; and since Vec<String> is opaque, there isn't a reasonable way to do so.

I think the reason for having #[repr(C)] is to reproduce data to match an existing C API you want to interoperate with; not to somehow directly expose internal Rust types to C.