Rustdoc: maybe don't display #[repr(C)] sometimes

[CAD97]

If a struct has some public fields and some non-public fields, // some fields omitted is displayed at the end of the declaration shown by rustdoc. Rustdoc also shows attributes on the declaration, such as #[repr(C)].

This is actively misleading for #[repr(C)] types. If a private type is before any public types, then the struct as displayed suggests that the public fields are at the prefix of the struct, and thus have defined offsets smaller than where they actually are.

Example:

#[repr(C)]
#[derive(Debug, Eq, PartialEq, Hash)]
pub struct ThinData<Head, SliceItem> {
    length: usize,
    pub head: Head,
    pub slice: [SliceItem],
}

There are two obvious potential solutions:

• Put // some fields omitted in the correct place(s) among public fields for #[repr(C)] types, or
• Don't display #[repr(C)] for types with some fields omitted.

[dtolnay]

Workaround: #[cfg_attr(not(doc), repr(C))] can be used to manually hide the attribute from documentation.

I agree rustdoc is incorrect by showing repr(C) on structs with a nonpublic field.