fix: let-and-return suggests invalid cast (#16135)

[ErinvanderVeen]

Simple PR that handles a special case where a raw pointer was invalidly cast via as _.

First PR, would appreciate feedback.

changelog: [let_and_return]: fix lint suggestion for invalid cast to raw pointer.

[rustbot]

r? @Jarcho

rustbot has assigned @Jarcho.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[Jarcho]

The test you have still works with the extra cast, it's just unnecessary. The type cast is needed to work around deficiencies in type inference:

struct S;
trait T {}
impl T for S {}

fn f(x: *const S) -> *const dyn T {
    Clone::clone(&x) as _
}

[Jarcho]

You aren't including a test for the actual issue. Just copy the function from the issue since using as _ in that case messes up type inference.

View changes since this review

[Jarcho]

This doesn't test the lint. The function body should be:

let x = Clone::clone(&x);
x

[ErinvanderVeen]

You're absolutely right. I'll try to get to this asap.

[ErinvanderVeen]

No, that doesn't work.

struct S;
trait T {}
impl T for S {}

fn f(x: *const S) -> *const dyn T {
    Clone::clone(&x) as _
}

Doesn't actually create an error, so the lint was valid.

[Jarcho]

That not cause a lint warning is why you need to change the test. Right now the test contains what the code should be after rustfix runs, but the test needs to make sure we don't make a broken suggestion. Hence you need to change the function's body to what I said earlier so that it will actually trigger let_and_return.

[ErinvanderVeen]

That, I understand, but I'd love to be able to reproduce the error from the original issue in the test. This test doesn't have that error.

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.