used_underscore can't handle macros and derive properly at the same time

[Manishearth]

I removed the test in Manishearth@4a32445 because we need a stronger macro check to handle stuff like https://github.com/Manishearth/rust-clippy/issues/488#issuecomment-166032639

Might need to write a better macro check. Looks like hygiene works differently for MacroAttributes. Perhaps we should just check for those?

cc @devonhollowood

[devonhollowood]

I can take a look at this tonight. While we are improving the macro checking, we might want to lint in cases like this:

macro_rules! increment {
    ($_x:expr) => {$_x + 1}
}

(currently we don't lint that)

[Manishearth]

We don't want to lint in that case. The exact example is obvious, but it's possible that with a more complex macro that the unused variable was due to the exact set of arguments passed which may have deliberately made it unused.

[devonhollowood]

I wrote something that I think would work, but I'm not sure how best to test this. I don't want to test using #[derive(RustcEncodable)] because it doesn't seem worth adding the rustc-serialize dependency for this. Just doing #[derive(Clone)] doesn't set off the lint, even if I remove the new check.

[Manishearth]

(Ah, sorry, didn't see this message)

Eh, rustc-serialize is stable (since the derive functionality is part of the compiler in this case), so I'm okay with a dev-dependency on it. We already do this for regex for a different macro test.