Type inference analysis confused by use in scope

[CAD97]

(Sorry for the not-super-minimized rerpo, I ran into this while following the gfx-hal intro.)

Cargo.toml

[package]
name = "gfx-hal-tutorials"
version = "0.0.0"
edition = "2018"
publish = false

[dependencies]
winit = "0.22.0"

main.rs (use at root scope)

use winit::dpi::{LogicalSize, PhysicalSize};
const WINDOW_SIZE: [u32; 2] = [512, 512];

fn main() {
    let event_loop = winit::event_loop::EventLoop::new();

    let (logical_window_size, physical_window_size) = {
        let dpi = event_loop.primary_monitor().scale_factor();
        let logical: LogicalSize<u32> = WINDOW_SIZE.into();
        let physical: PhysicalSize<u32> = logical.to_physical(dpi);

        (logical, physical)
    };
}

main.rs (scoped use)

const WINDOW_SIZE: [u32; 2] = [512, 512];

fn main() {
    use winit::dpi::{LogicalSize, PhysicalSize};
    let event_loop = winit::event_loop::EventLoop::new();

    let (logical_window_size, physical_window_size) = {
        let dpi = event_loop.primary_monitor().scale_factor();
        let logical: LogicalSize<u32> = WINDOW_SIZE.into();
        let physical: PhysicalSize<u32> = logical.to_physical(dpi);

        (logical, physical)
    };
}

This second form infers the wrong type for logical_window_size, and no type for physical_window_size.

[edwin0cheng]

Local imports is still not being supported yet. (#1165)

[CAD97]

An interesting side-effect is r-a pulling what appears to be a completely random type that satisfies [u32; 2] as Into<$0> (maybe?) for logical_window_size here.

[flodiebold]

I think that may be related to a bug where Chalk nondeterministically returns a definite result when it shouldn't. Does it still happen in latest RA?

[lnicola]

They show up as unknown for me.

[flodiebold]

Let's close this since we have #1165 for local imports and the weird inference is fixed.