Bindgen gets confused by external inline function definition

[Tazdevil971]

Input C/C++ Header

class Foo {
public:
    void bar();
};

inline void Foo::bar() {
    // Oh-no
}

Bindgen Invocation

Generated by bindgen input.hpp, no flags

/* automatically generated by rust-bindgen 0.59.2 */

#[repr(C)]
#[derive(Debug, Copy, Clone)]
pub struct Foo {
    pub _address: u8,
}
#[test]
fn bindgen_test_layout_Foo() {
    assert_eq!(
        ::std::mem::size_of::<Foo>(),
        1usize,
        concat!("Size of: ", stringify!(Foo))
    );
    assert_eq!(
        ::std::mem::align_of::<Foo>(),
        1usize,
        concat!("Alignment of ", stringify!(Foo))
    );
}
extern "C" {
    #[link_name = "\u{1}_ZN3Foo3barEv"]
    pub fn Foo_bar(this: *mut Foo);
}
impl Foo {
    #[inline]
    pub unsafe fn bar(&mut self) {
        Foo_bar(self)
    }
}

Expected Results

Bindgen shouldn't be generating bindings for bar(). It probably gets confused by the fact that Foo's declaration doesn't specifically say that bar() is inline, but instead it should look forward for a possible inline definition.

Marking the bar() declaration with inline does fix the issue, even if we don't provide an actual body.

The specific header I'm working with is this one, and it makes heavy usage of this to define std dependent helpers.

[pvdrz]

Hmmm I have noticed some finicky details about how inline is handled and I remember seeing a different instance of this using C and not C++.

Not directly related but some of the machinery that will be introduced to fix #1090 might be able to fix this.