Virtual Keycode is incorrect on macOS non-qwerty layouts

[Toqozz]

The virtual_keycode field seems to be incorrect when using a non-qwerty layout. It seems to always return values corresponding to the qwerty (or perhaps physical?) layout.

I'm running the following program:

extern crate winit;

fn main() {
    let mut events_loop = winit::EventsLoop::new();
    let window = winit::Window::new(&events_loop).unwrap();

    events_loop.run_forever(|event| {
        println!("{:?}", event);
        match event {
            winit::Event::WindowEvent {
                event: winit::WindowEvent::CloseRequested, ..
            } => winit::ControlFlow::Break,
            _ => winit::ControlFlow::Continue,
        }
    });
}

Pressing a dvorak P here (qwerty 'R') results in the following lines:

WindowEvent { window_id: WindowId(Id(140217433672064)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 15, state: Pressed, virtual_keycode: Some(R), modifiers: ModifiersState { shift: false, ctrl: false, alt: false, logo: false } } } }

WindowEvent { window_id: WindowId(Id(140217433672064)), event: ReceivedCharacter('p') }

As you can see, we receive virtual_keycode: Some(R). Also note that right after the window receives the P character. (Shouldn't these always be the same?)

The issue doesn't seem to occur on Linux/X11.

winit version: 0.18
MacBook Pro (2017)
macOS version: 10.14 (Mojave)

[francesca64]

Thanks for reporting this!

We're currently using NSEvent keyCode to generate VirtualKeyCode: https://github.com/tomaka/winit/blob/139686ddce870614f12eb3a0ed33a9adfee1e47a/src/platform/macos/view.rs#L415

This is wrong, since that's really just a scancode: https://stackoverflow.com/a/18413469/5435443

We likely inherited this oversight from GLFW, since it's done the same way there (and in SDL). It doesn't look like AppKit has a concept of layout-aware virtual key codes. WebKit has a good explanation and solution here: windowsKeyCodeForKeyEvent, windowsKeyCodeForCharCode

If you're interested in contributing, this would likely be an easy PR. Otherwise, it would be a week or so before I'd get a chance to implement this myself.

[Toqozz]

I will definitely have a look, thank you!

[francesca64]

@chrisduerr are you aware of any Alacritty bugs that look like they're caused by this? (or any people who might want to test #755?)

[chrisduerr]

The issue that comes to mind is alacritty/alacritty#458. I've linked the PR there for anyone interested in testing.