Missing visit_const/fold_const when visit_ty/fold_ty exists and consts are relevant.

[eddyb]

For visiting, I've identified these two examples:

rust/src/librustc/traits/structural_impls.rs

Lines 256 to 274 in 5aa8f19

	fn visit_ty(&mut self, t: Ty<'tcx>) -> bool {
	match t.kind {
	ty::Bound(debruijn, bound_ty) if debruijn == self.binder_index => {
	self.types.insert(
	bound_ty.var.as_u32(),
	match bound_ty.kind {
	ty::BoundTyKind::Param(name) => name,
	ty::BoundTyKind::Anon => {
	Symbol::intern(&format!("^{}", bound_ty.var.as_u32()))
	}
	},
	);
	}
	_ => (),
	};
	t.super_visit_with(self)
	}

rust/src/librustc/ty/fold.rs

Lines 976 to 990 in 5aa8f19

	fn visit_ty(&mut self, t: Ty<'tcx>) -> bool {
	// if we are only looking for "constrained" region, we have to
	// ignore the inputs to a projection, as they may not appear
	// in the normalized form
	if self.just_constrained {
	match t.kind {
	ty::Projection(..) | ty::Opaque(..) => {
	return false;
	}
	_ => {}
	}
	}
	t.super_visit_with(self)
	}

(the latter would need to handle ty::ConstKind::Unevaluated the same way ty::Projection is handled, as "unevaluated" consts are effectively projections)

For folding, I've listed one missing fold_const and two which are missing some recursion, in #70125 (comment).

cc @varkor @yodaldevoid