Multiple definitions

[TomzBench]

I have a really rough looking header (it's amalgamation).

https://github.com/TomzBench/mongoose-rs/blob/master/mongoose-sys/wrapper.h

Mongoose is a pretty popular single threaded web server for embedded systems written in C. So I went to try and make a binding to it and when I compile it I get a few errors such as seen below. (My repository is pretty minimal to reproduce.)

1007 | pub const IPPORT_RESERVED: u32 = 1024;
     | -------------------------------------- previous definition of the value `IPPORT_RESERVED` here
...
3092 | pub const IPPORT_RESERVED: _bindgen_ty_6 = 1024;
     | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ `IPPORT_RESERVED` redefined here
975  | pub const FP_NAN: u32 = 0;
     | -------------------------- previous definition of the value `FP_NAN` here
...
5166 | pub const FP_NAN: _bindgen_ty_8 = 0;
     | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ `FP_NAN` redefined here

The culprit is defined first as:

pub const FP_NAN: u32 = 0;
pub const FP_INFINTE: u32 = 1;
pub const FP_ZERO: u32 = 2;
pub const FP_NORMAL: u32 = 3;

Then it is redefined again some 5000 lines later as

pub const FP_NAN: __bindgen_ty_8 = 0;
pub const FP_INFINTE: __bindgen_ty_8 = 1;
pub const FP_ZERO: __bindgen_ty_8 = 2;
pub const FP_NORMAL: __bindgen_ty_8 = 3;

It looks like __bindgen_ty_8 is a u32 so they are defined as equal - but still produces error.
(also If I copy out the binding.rs and delete the duplicate's then it does compile ok...)

[emilio]

How are these defined in your system? Your header doesn't include those constants...

This looks like a dupe of #687, where we have both a macro and an anonymous enum with the same name.

Bindgen can't quite now what to generate and generates both. You can instruct bindgen to generate the variant you want as described in that issue.

[TomzBench]

...I thought because those constants were not part of the header, that bindgen added them for some reason. Since the header does not contain FP_* symbols, it is hard to figure out what the header is defining twice. I think the header is indirectly tricking bindgen into defining FP_* twice.

[emilio]

Bindgen generates everything by default, you need to whitelist (which is recursive by default) or blacklist stuff to limit the output.