cbindgen 0.6.4 moved constants outside namespace

[staktrace]

The constants at https://searchfox.org/mozilla-central/rev/9cb3e241502a2d47e2d5057ca771324a446b6695/gfx/webrender_bindings/webrender_ffi_generated.h#20,22 are inside the mozilla::wr namespace. With cbindgen 0.6.4 they get moved outside the namespace which is undesirable.

[staktrace]

Actually these constants are generated from https://searchfox.org/mozilla-central/rev/9cb3e241502a2d47e2d5057ca771324a446b6695/gfx/webrender_bindings/src/program_cache.rs#16 and are never used on the C++ side so if there's a way to omit them from being generated that would be fine too.

[staktrace]

This appears to be a result of the changes in #170 to bindings.rs, it explicitly moved the namespace opening down to after the constants were written. @IGI-111 was this intentional? None of the tests added seem to have namespaces so I'm wondering if this change was accidental.

[staktrace]

I guess it was intentional, because only constants of type is_primitive_or_ptr_primitive() got moved outside the namespace. But I'm not sure why.

[eqrion]

I don't recall the reason why constants of type is_primitive_or_ptr_primitive() were moved outside namespaces. I know !is_primitive_or_ptr_primitive() were moved to be after all items were output so they have the right definitions.

[eqrion]

@staktrace As a stopgap you could skip emitting constants.

[export]
# don't have a negative version of item-types unfortunately
# item-types = ["Constants", "Globals", "Enums", "Structs", "Unions", "Typedefs", "OpaqueItems", "Functions"]
item-types = ["Globals", "Enums", "Structs", "Unions", "Typedefs", "OpaqueItems", "Functions"]

[IGI-111]

That wasn't an intentional change. I suspect either automerging lines that moved around or my own dumb self borking a conflicting merge.

In any case, there should probably be some tests for the namespacing feature, right now there doesn't seem to be any?