Type aliases do not work in #[repr(...)] for C-style enums

[brendanzab]

#[repr(c_int)]
enum ABC {
    A = 0,
    B = 1,
    C = 2,
}

$ rustc --lib repr-test.rs
repr-test.rs:1:7: 1:12 error: unrecognized representation hint
repr-test.rs:1 #[repr(c_int)]
                      ^~~~~
error: aborting due to previous error
task 'rustc' failed at 'explicit failure', /Users/brendan/dev/rust/rust/src/libsyntax/diagnostic.rs:101
task '<main>' failed at 'explicit failure', /Users/brendan/dev/rust/rust/src/librustc/lib.rs:397

[brendanzab]

Apparently you can use #[repr(C)], but this doesn't seem all that flexible.

[huonw]

@bjz, if you are interfacing with C, you want #[repr(C)] rather than c_int, since it's not guaranteed that the size will be an int (i.e. the C type).

[brendanzab]

@huonw Could you give an example?

[huonw]

Discussion on IRC revealed that @bjz is defining an enum wrapper around a series of #define constants, which are used as arguments to functions with type c_int, not just a plain C-enum, so my suggestion of #[repr(C)] doesn't necessarily directly apply.

[steveklabnik]

Triage: I believe we've gotten more restrictive with what can go inside of repr since then, making this an enhancement.

[steveklabnik]

Triage: no change.

[Mark-Simulacrum]

I'm not sure we expect to support type aliases and other "expressions" in repr. In fact, I'm not sure we should.

[Havvy]

IMO, this should be closed, with an RFC dedicated to the idea should somebody want to seriously pursue this. (Although I think it's a bad idea).

[Centril]

Closing per @Havvy's request and motivation and @steveklabnik's note from 2015.