Document behavior of #[cfg] in macros

[RalfJung]

The macro reference should explain how #[cfg] and #[cfg_attr(...)] behave inside macros.

Does this macro check the allow_clippy flag of the crate the macro is defined in, or of the crate the macro is used in?

macro_rules! field_check {
    ($type:path, $field:tt) => {
        // Make sure the field actually exists. This line ensures that a
        // compile-time error is generated if $field is accessed through a
        // Deref impl.
        #[cfg_attr(allow_clippy, allow(clippy::unneeded_field_pattern))]
        let $type { $field: _, .. };
    };
}

[ehuss]

I assume you are interested in more than just allow/deny diagnostic behavior. I mention that because I believe (some?) diagnostics have special behavior in external macros where they are disabled. The reference has historically shied away from documenting diagnostics, so that particular behavior is a bit of a gray area, but probably should be mentioned.

There are a few other behaviors of cfg/cfg_attr that I'd like to see documented (#565, #103). In particular, I think the order of transformations ("phases of translation" in C++ parlance) would be useful to specify to make it clear when cfg processing is done in relation to other phases.

[RalfJung]

I assume you are interested in more than just allow/deny diagnostic behavior.

Yes, I wonder about cfg in general (including cfg!). I didn't even know diagnostics have some special magic hacks.^^