Missing self_in_typedefs help message

[leonardo-m]

This code compiles with recent Nightly:

#![feature(self_in_typedefs)]
enum Tree { Empty, Node(Box<Self>, i32, Box<Self>) }
fn main() {}

If I remove the feature, on Nightly:

enum Tree { Empty, Node(Box<Self>, i32, Box<Self>) }
fn main() {}

It gives error messages but no suggestion to add the self_in_typedefs feature:

error[E0411]: cannot find type `Self` in this scope
 --> ...\test.rs:1:29
  |
1 | enum Tree { Empty, Node(Box<Self>, i32, Box<Self>) }
  |                             ^^^^ `Self` is only available in traits and impls

error[E0411]: cannot find type `Self` in this scope
 --> ...\test.rs:1:45
  |
1 | enum Tree { Empty, Node(Box<Self>, i32, Box<Self>) }
  |                                             ^^^^ `Self` is only available in traits and impls

error: aborting due to 2 previous errors

For some other features Rustc gives such help:

enum Tree { Empty, Node(Box<Tree>, i32, Box<Tree>) }
fn main() {
    let t = Tree::Node(box Tree::Empty, 0, box Tree::Empty);
}

error[E0658]: box expression syntax is experimental; you can call `Box::new` instead. (see issue #49733)
 --> ...\test.rs:3:24
  |
3 |     let t = Tree::Node(box Tree::Empty, 0, box Tree::Empty);
  |                        ^^^^^^^^^^^^^^^
  |
  = help: add #![feature(box_syntax)] to the crate attributes to enable

[Centril]

cc @alexreg and #49303

[estebank]

The help needs to be added here

rust/src/librustc_resolve/lib.rs

Lines 2998 to 3007 in 31789a6

	if is_self_type(path, ns) {
	__diagnostic_used!(E0411);
	err.code(DiagnosticId::Error("E0411".into()));
	let available_in = if this.session.features_untracked().self_in_typedefs {
	"impls, traits, and type definitions"
	} else {
	"traits and impls"
	};
	err.span_label(span, format!("`Self` is only available in {}", available_in));
	return (err, Vec::new());

You can see how such a help text was added in the past.

You can also use ::rustc::session::config::nightly_options::is_nightly_build() to only suggest it in nightly builds.

[alexreg]

Thanks for the above guidance, @estebank. And sorry I forgot to include this with the initial PR.

Let me add that this is slightly tricky, due to the fact we do not know the context of the Self usage at the point the error is emitted. We have to work it out again. I'll see if I can come up with something.

[alexreg]

@estebank I noticed there's also tcx.sess.opts.unstable_features.is_nightly_build(). I this any different?

[estebank]

Nope, if that is available use that one. They should both be the same value.

[alexreg]

Okay sure. We don't have a tcx in this case though, I later realised, so I'll use your original approach! Ready to merge?