Type inference for closure gets confused without trivial help

[frankmcsherry]

The following falls over at the moment, but works with the commented version of the closure:

fn main() {
    let f = |&x| x as u64;         // <-- doesn't work
    // let f = |&x: &_| x as u64;  // <-- does work
    test_me0(&f);
    test_me1(&f);
}

fn test_me0<F: Fn(&u32) -> u64>(func: &F) -> u64 { func(&0) }
fn test_me1<F: Fn(&u32) -> u64>(func: &F) -> u64 { func(&1) }

@eddyb's thinking is that a &x pattern may not result in a &T type hint, same for &mut x. His thinking was that way back when &x would also match &mut which is no longer possible, but attempts to get playbot to verify this for us in public did not go as planned.

In any case, it was there determined that &x: _ and &x: &_ should be identical, and that this is not expected. :)

[steveklabnik]

#27828 is basically the same issue, and has a bunch more examples, so I'm going to close this one. I've left a note over there that this example matters too, thanks!

[frankmcsherry]

You've referenced this issue, not another one. Meant #12679 maybe?

[steveklabnik]

Yes, sorry :( copy-paste error.