Confusing error message when expecting BoxFuture

[tmandry]

Playground

fn foo() -> BoxFuture<'static, i32> {
    async {
        42
    }
}

gives

error[E0308]: mismatched types
 --> src/lib.rs:5:5
  |
4 |   fn foo() -> BoxFuture<'static, i32> {
  |               ----------------------- expected `std::pin::Pin<std::boxed::Box<(dyn core::future::future::Future<Output = i32> + std::marker::Send + 'static)>>` because of return type
5 | /     async {
6 | |         42
7 | |     }
  | |_____^ expected struct `std::pin::Pin`, found opaque type
  |
  = note: expected type `std::pin::Pin<std::boxed::Box<(dyn core::future::future::Future<Output = i32> + std::marker::Send + 'static)>>`
             found type `impl core::future::future::Future`

The expected struct std::pin::Pin in particular is confusing. This can be a pretty intimidating error message for someone new to async/await, when all they forgot was .boxed().

I'm not sure if we can fix this without stabilizing BoxFuture so we can sprinkle some magic compiler dust on its error messages. Maybe there's something we can do, though. Maybe some intelligent use of type aliases in our error messages?

cc @JakeEhrlich

[estebank]

This can be handled without needing to stabilize the struct. It should act the same way that the primitive type cast suggestions do.

[nikomatsakis]

Some observations:

• There is no BoxFuture struct, it's a type-alias
• We could suggest invoking boxed() but that is a method defined only in the futures crate, which feels a bit awkward to me; I'd rather it were part of libstd
• We can suggest Box::pin(async { .. }) as well (playground)

[tmandry]

marking as OnDeck, since BoxFuture gets used quite often

[estebank]

I have a PR that suggests the right code for this:

But if first you try Box::new:

or Pin::new:

we will send you in a wild goose chase:

[estebank]

Filed #69083 for the uncovered cases.