Invalid code generated when a struct contains fields called `type` and `type_`

[sgrif]

Header to reproduce:

typedef struct foo {
  int type;
  int type_;
};

The resulting code will be:

#[repr(C)]
#[derive(Debug, Copy)]
pub struct foo {
    pub type_: ::std::os::raw::c_int,
    pub type_: ::std::os::raw::c_int,
}

Which is of course, invalid. It would probably make sense to at least give an error rather than generating invalid code here.

[emilio]

Argh, nasty. Thanks for reporting. We should be able to just mangle the reserved names further if a type has a duplicated name.

[divagant-martian]

I tried solving this by mangling too names that look like they would be the result of being mangled. So type_ would be mangled to type__. I don't see how to solve this checking duplicates, since duplicates only matter inside scopes (duplicates in function arguments, for example).

A consequence of doing this is that names that don't need to be mangled because they are not duplicates would be changed anyway. Another consequence is that parts of the bitfield special handling wouldn't be needed now.
The bitfield-method-same-name.hpp

struct Foo {
    char type_ : 3;
    char type();
    void set_type_(char c);
    void set_type(char c);
};

has differences in this fashion

- pub fn type__bindgen_bitfield(&self) -> ::std::os::raw::c_char {
+ pub fn type__(&self) -> ::std::os::raw::c_char {

Would appreciate ideas on how to tackle this

[pvdrz]

I think the change in the tests are fine as the _bindgen_bitfield suffix is added when there's a name collision between a method that already exists in the headers and the setter method that's being generated by `bindgen.