macos: modified key issues

[sodiumjoe]

Not totally sure these are all related, but:

alacritty/alacritty#1336
alacritty/alacritty#1358
alacritty/alacritty#1428
#623

I suspect these all started with #518

@francesca64 has given some direction here

[sodiumjoe]

just confirmed these all started with #518 by git bisecting.

[sodiumjoe]

One interesting thing to note is for alacritty/alacritty#1336, this might be working as intended? When I run the multiwindow example, <CMD-`> correctly cycles me through the open windows. I say "correctly," because this is normal macos app behavior.

Apparently <CMD-.> is used for cancel (though I can't find official documentation for that anywhere), and <C-TAB> is Move to the next control when a text field is selected .

Since <C-TAB> and <CMD-.> is sort of application dependent behavior, I guess winit should just pass these key events through to the consuming application.

But for <CMD-`>, I'm not sure, since this is standard app/window behavior. Should winit just handle this? Should it be default behavior with an option to opt out from an application?

[francesca64]

I'm generally in favor of winit conforming to the conventions of the platform, which would entail letting the OS handle <CMD-`>. The only reason winit could previously override it was because our keyboard handling was overtly non-standard.

[sodiumjoe]

@francesca64 So would you consider alacritty/alacritty#1336 a wontfix then? (I think that's a reasonable position.)

[francesca64]

Yes. Even if I weren't against it conceptually, the potential fix is unreasonably convoluted.

[sodiumjoe]

So I found this, which in my testing appears to fix the <C-Tab> issue:

before:

WindowEvent { window_id: WindowId(Id(140332214940576)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 59, state: Pressed, virtual_keycode: Some(LControl), modifiers: ModifiersState { shift: false, ctrl: true, alt: false, logo: false } } } }
WindowEvent { window_id: WindowId(Id(140332214940576)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 48, state: Released, virtual_keycode: Some(Tab), modifiers: ModifiersState { shift: false, ctrl: true, alt: false, logo: false } } } }
WindowEvent { window_id: WindowId(Id(140332214940576)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 59, state: Released, virtual_keycode: Some(LControl), modifiers: ModifiersState { shift: false, ctrl: false, alt: false, logo: false } } } }

after:

WindowEvent { window_id: WindowId(Id(140623425336880)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 59, state: Pressed, virtual_keycode: Some(LControl), modifiers: ModifiersState { shift: false, ctrl: true, alt: false, logo: false } } } }
WindowEvent { window_id: WindowId(Id(140623425336880)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 48, state: Pressed, virtual_keycode: Some(Tab), modifiers: ModifiersState { shift: false, ctrl: true, alt: false, logo: false } } } }
WindowEvent { window_id: WindowId(Id(140623425336880)), event: ReceivedCharacter('\t') }
WindowEvent { window_id: WindowId(Id(140623425336880)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 48, state: Released, virtual_keycode: Some(Tab), modifiers: ModifiersState { shift: false, ctrl: true, alt: false, logo: false } } } }
WindowEvent { window_id: WindowId(Id(140623425336880)), event: KeyboardInput { device_id: DeviceId(DeviceId), input: KeyboardInput { scancode: 59, state: Released, virtual_keycode: Some(LControl), modifiers: ModifiersState { shift: false, ctrl: false, alt: false, logo: false } } } }

But it doesn't fix the <Cmd-.> issue. But I did find this, which appears to be addressing the same issue. I haven't quite figured out how to port this over yet, but I'm working on it.

[sodiumjoe]

#629

[sodiumjoe]

that branch fixes <C-Tab> and <Cmd-{key}> release, but not <Cmd-.>.

[sodiumjoe]

Interestingly, with <Cmd-.>, it's the keydown that's suppressed and the keyup that fires. Unfortunately, the same trick doesn't seem to work:

                if let Some(key_window) = maybe_key_window() {
                    if event_mods(ns_event).logo && ns_event.keyCode() == 47 {
                        println!("keycode: {}", ns_event.keyCode());
                        msg_send![*key_window.window, sendEvent:ns_event];
                    }
                }

the println fires, but the key event does not.

[sodiumjoe]

Ok, added special handling for <Cmd-.> that appears to work.