Revise and generalize the macros-unlinted tests.

Review feedback asked for the test to be generalized to include macros
2.0; that generalization is dyn-2015-idents-in-decl-macros-unlinted.rs

As a drive-by, I also decided to revise the test to make it clear
*why* we cannot generally lint these cases. (I already had similar
demonstrations in dyn-2015-edition-keyword-ident-lint.rs, but it does
not hurt to try to emphasize matters.)

I also added some commentary on the cases where we could choose to
make the lint smarter, namely the situations where a macro is
*definitely* using `dyn` as an identifier (because it is using it as a
path component).

src/test/ui/dyn-keyword/dyn-2015-idents-in-decl-macros-unlinted.rs

@@ -0,0 +1,51 @@
+// compile-pass
+
+// Under the 2015 edition with the keyword_idents lint, `dyn` is
+// not entirely acceptable as an identifier.
+//
+// We currently do not attempt to detect or fix uses of `dyn` as an
+// identifier under a macro, including under the declarative `macro`
+// forms from macros 1.2 and macros 2.0.
+
+#![feature(decl_macro)]
+#![allow(non_camel_case_types)]
+#![deny(keyword_idents)]
+
+mod outer_mod {
+    pub mod r#dyn {
+        pub struct r#dyn;
+    }
+}
+
+// Here we are illustrating that the current lint does not flag the
+// occurrences of `dyn` in this macro definition; however, it
+// certainly *could* (and it would be nice if it did), since these
+// occurrences are not compatible with the 2018 edition's
+// interpretation of `dyn` as a keyword.
+macro defn_has_dyn_idents() { ::outer_mod::dyn::dyn }
+
+struct X;
+trait Trait { fn hello(&self) { }}
+impl Trait for X { }
+
+macro tt_trait($arg:tt) { & $arg Trait }
+macro id_trait($id:ident) { & $id Trait }
+
+fn main() {
+    defn_has_dyn_idents!();
+
+    // Here we are illustrating that the current lint does not flag
+    // the occurrences of `dyn` in these macro invocations. It
+    // definitely should *not* flag the one in `tt_trait`, since that
+    // is expanding in a valid fashion to `&dyn Trait`.
+    //
+    // It is arguable whether it would be valid to flag the occurrence
+    // in `id_trait`, since that macro specifies that it takes an
+    // `ident` as its input.
+    fn f_tt(x: &X) -> tt_trait!(dyn) { x }
+    fn f_id(x: &X) -> id_trait!(dyn) { x }
+
+    let x = X;
+    f_tt(&x).hello();
+    f_id(&x).hello();
+}

src/test/ui/dyn-keyword/dyn-2015-idents-in-macros-unlinted.rs

@@ -15,10 +15,42 @@ mod outer_mod {
     }
 }
 
+// Here we are illustrating that the current lint does not flag the
+// occurrences of `dyn` in this macro definition; however, it
+// certainly *could* (and it would be nice if it did), since these
+// occurrences are not compatible with the 2018 edition's
+// interpretation of `dyn` as a keyword.
 macro_rules! defn_has_dyn_idents {
-    ($arg:ident) => { ::outer_mod::dyn::dyn }
+    () => { ::outer_mod::dyn::dyn }
+}
+
+struct X;
+trait Trait { fn hello(&self) { }}
+impl Trait for X { }
+
+macro_rules! tt_trait {
+    ($arg:tt) => { & $arg Trait }
+}
+
+macro_rules! id_trait {
+    ($id:ident) => { & $id Trait }
 }
 
 fn main() {
-    defn_has_dyn_idents!(dyn);
+    defn_has_dyn_idents!();
+
+    // Here we are illustrating that the current lint does not flag
+    // the occurrences of `dyn` in these macro invocations. It
+    // definitely should *not* flag the one in `tt_trait`, since that
+    // is expanding in a valid fashion to `&dyn Trait`.
+    //
+    // It is arguable whether it would be valid to flag the occurrence
+    // in `id_trait`, since that macro specifies that it takes an
+    // `ident` as its input.
+    fn f_tt(x: &X) -> tt_trait!(dyn) { x }
+    fn f_id(x: &X) -> id_trait!(dyn) { x }
+
+    let x = X;
+    f_tt(&x).hello();
+    f_id(&x).hello();
 }