Incorrect suggestion when trying to write to an immutable field in an async function

[jyn514]

Given the following code: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=df6e36b97b73bd3e3ea0fdfe9325b1c9

struct S { foo: usize }
impl S {
    async fn bar(&self) {
        self.foo += 1;
    }
}

The current output is:

error[E0594]: cannot assign to `self.foo`, which is behind a `&` reference
 --> src/lib.rs:4:9
  |
3 |     async fn bar(&self) {
  |                  ----- help: consider changing this to be a mutable reference: `&mut S`
4 |         self.foo += 1;
  |         ^^^^^^^^^^^^^ `self` is a `&` reference, so the data it refers to cannot be written

Ideally the output should look like:

error[E0594]: cannot assign to `self.foo`, which is behind a `&` reference
 --> src/lib.rs:4:9
  |
3 |     async fn bar(&self) {
  |                  ----- help: consider changing this to be a mutable reference: `&mut self`
4 |         self.foo += 1;
  |         ^^^^^^^^^^^^^ `self` is a `&` reference, so the data it refers to cannot be written

&mut S is not a valid argument; it needs to be either &mut self or some_name: &mut S. For some reason this problem only happens when the function is async.

[compiler-errors]

This is the same diagnostic you get when mutably dereferencing let-bindings that are &T: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=b71c99d23b4e58a67afd52a0870bed65
So it seems like this has to do with the fact that ast_lowering emits a let self = self; binding when desugaring.

I wonder if we can get the name of the variable when emitting this diagnostic, and special-case when it == "self".

[jyn514]

Another (possibly related?) bug: self.foo += 1 doesn't have the help suggestion if it's in an async block: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=23ea8e56639ab035349c69050bd74148

struct S { foo: usize }

impl S {
    fn bar(&self) {
        async move {
            self.foo += 1;
        };
    }
}

[compiler-errors]

@jyn514 Same with closures. I would probably classify that as a separate bug, since the original bug has to do specifically with async fn desugaring, and the latter has to do with closure/generator captures.

[eholk]

Thanks for the report and getting a PR up so quickly!

@rustbot label +AsyncAwait-Triaged +AsyncAwait-Polish