libsyntax visitor does not visit the paths in macros

[nikomatsakis]

It's a minor point, but the macro struct includes a path:

rust/src/libsyntax/ast.rs

Lines 1223 to 1228 in 551244f

	#[derive(Clone, RustcEncodable, RustcDecodable, Debug)]
	pub struct Mac_ {
	pub path: Path,
	pub delim: MacDelimiter,
	pub tts: ThinTokenStream,
	}

However, the walk_mac helper just ignores this completely:

rust/src/libsyntax/visit.rs

Lines 650 to 652 in 551244f

	pub fn walk_mac<'a, V: Visitor<'a>>(_: &mut V, _: &Mac) {
	// Empty!
	}

As a result — well, a partial result — the early lint checker just ignores this path altogether (in particular, visit_path is never invoked, and hence check_path and friends are not invoked):

rust/src/librustc/lint/context.rs

Lines 1119 to 1121 in 551244f

	fn visit_mac(&mut self, mac: &'a ast::Mac) {
	run_lints!(self, check_mac, mac);
	}

This was a contributing factor to #53686.

I say that this is a "partial result" because the early lint checker never calls walk_mac -- but, really, it ought to. And, if it did, and walk_mac did what it was supposed to do and walked all the subparts, then everything would be fine.

Adding a hacky fix for now.

[nikomatsakis]

cc @Manishearth @oli-obk — Do you agree that the lint visitor ought to change here?

cc @petrochenkov — do the paths in macros have a NodeId like other paths? That was one thing I wans't sure of when I tried to push this through.

[petrochenkov]

do the paths in macros have a NodeId like other paths?

No, node ids are assigned during expansion to "final" nodes that will stay in AST after expansion.

[Manishearth]

Yeah, the lint visitor should change.