fix: #12247 fully qualified call required to determine trait method type

[bitgaoshu]

under standard core

[lnicola]

r? @flodiebold

[flodiebold]

Maybe I'm misunderstanding the change, but I'm a bit confused. The test code doesn't actually work under rustc. Also, we would actually need to call (or simulate calling) Into::into on the error type, not do a coercion, I think?

[bitgaoshu]

the case is like that the chalk doesn’t have enough knowledge to decide which impl to use. so add error type information. 

[bitgaoshu]

Sorry I don't know what the detail difference between From and Into. And I changed test code to Into, the test code work under rustc. Actually, The code does not coercion, just unify. @flodiebold

[bors]

☔ The latest upstream changes (presumably #14036) made this pull request unmergeable. Please resolve the merge conflicts.

[HKalbasi]

I think this would cause a false positive type mismatch in this code:

fn x() -> Result<(), E2> {
    let a: Result<i32, _> = Ok(2);
    let b = a?;
    let c: Result<i32, E1> = a;
    Ok(())
}

[bitgaoshu]

I use rustc 1.66.1, it's a type mismatch and ra with the patch give a hint.

[HKalbasi]

With the from impl, it works (playground).

[bors]

☔ The latest upstream changes (presumably #14368) made this pull request unmergeable. Please resolve the merge conflicts.

[Veykril]

What is the status of this PR? I am still not sure what this is trying to fix.

[HKalbasi]

I think this is now in an unrecoverable / irrelevant state, as we now desugar ? and there is no Expr::Try anymore. So let's closing.