Rustified enums getting both #[repr(u32)] and #[repr(C)]

[faern]

Input C header

enum my_enum {
    VARIANT1,
};

Bindgen Invocation

$ bindgen enum.h --rustified-enum my_enum

Actual Results

With bindgen 0.32.1 I get:

#[repr(u32)]
#[repr(C)]
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
pub enum my_enum {
    VARIANT1 = 0,
}

rustc warning when compiling:

warning[E0566]: conflicting representation hints
 --> enum.rs:3:8
  |
3 | #[repr(u32)]
  |        ^^^
4 | #[repr(C)]
  |        ^

Expected Results

To only have one #[repr(...)]:

#[repr(C)]
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
pub enum my_enum {
    VARIANT1 = 0,
}

[faern]

The same problem can be seen in the generated code in issue #1198. But that issue is about something else, so I felt it was worth reporting on this alone.

[RazrFalcon]

Stumbled on the same bug just now.

[emilio]

Hmm, this is clearly a regression... Shouldn't be hard to fix.

[emilio]

Oh, I lie, this was added intentionally in 43c43ba... Which version of rustc are you using?

[faern]

@emilio rustc 1.25.0-nightly (3f92e8d89 2018-01-14)

EDIT: Latest nightly gives the same warning, but stable does not give any warning when compiling with two representation hints.

[emilio]

Ok, asked in #rustc today. Relevant log: https://botbot.me/mozilla/rustc/2018-01-19/?msg=95934948&page=2

TL;DR: the extra repr(C) is only needed for enums with fields, not for plain enums that bindgen generates.

[emilio]

Will send a patch with a fix.