New `#[must_use]` behaviour for async methods does not apply to `Pin<Box<dyn Future<Output = ()>>>`

[tropicbliss]

Defining traits that contain methods that return such a type:

pub trait Hello {
    fn hi(&self) -> Pin<Box<dyn Future<Output = i32>>> {
        Box::pin(async { 42 })
    }
}

struct Number(i32);

impl Hello for Number {}

async fn say_hello() {
  let number = Number(42);
  number.hi().await;
}

The compiler should probably have warned me about not using the Output of the Future, consistent with the behaviour of async functions in Rust 1.67.0. This will also work without the need for traits but async traits is one of the most common use cases for such a type.

Meta

rustc --version --verbose:

rustc 1.69.0 (84c898d65 2023-04-16)
binary: rustc
commit-hash: 84c898d65adf2f39a5a98507f1fe0ce10a2b8dbc
commit-date: 2023-04-16
host: x86_64-unknown-linux-gnu
release: 1.69.0
LLVM version: 15.0.7

Also tested on:
rustc 1.71.0-nightly (f9a6b7158 2023-05-05)
binary: rustc
commit-hash: f9a6b71580cd53dd4491d9bb6400f7ee841d9c22
commit-date: 2023-05-05
host: x86_64-unknown-linux-gnu
release: 1.71.0-nightly
LLVM version: 16.0.2

Backtrace

[eholk]

We looked at this as the wg-async triage meeting today. We agree this is a bug and that the behavior should be consistent with #[must_use] on async fn and/or -> impl Future.

This is also likely a good first issue for someone wanting to contribute to async.

[thosakwe]

We looked at this as the wg-async triage meeting today. We agree this is a bug and that the behavior should be consistent with #[must_use] on async fn and/or -> impl Future.

This is also likely a good first issue for someone wanting to contribute to async.

Would you say this is also a good issue for a first-time contributor to Rust?

I'm not super familiar with the language, but I want to contribute to open source compilers, so I'm hoping to potentially take on this task.