In this line it is not checked whether the key was pressed or released, so egui::Event::Text gets added twice. Once on press, once on release.
I noticed this while using egui::TextEdit.

I think changing this line to

if pressed && !self.modifier_state.intersects(ModifiersState::CONTROL | ModifiersState::SUPER)

fixes the issue

winit 0.28.x introduced some breaking changes that broke egui_winit_platform:

error: Please select a feature to build for unix: `x11`, `wayland`
  --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/platform_impl/linux/mod.rs:10:1
   |
10 | compile_error!("Please select a feature to build for unix: `x11`, `wayland`");
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

error[E0392]: parameter `T` is never used
  --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/event_loop.rs:35:22
   |
35 | pub struct EventLoop<T: 'static> {
   |                      ^ unused parameter
   |
   = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`

error[E0392]: parameter `T` is never used
  --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/event_loop.rs:46:34
   |
46 | pub struct EventLoopWindowTarget<T: 'static> {
   |                                  ^ unused parameter
   |
   = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`

error[E0392]: parameter `T` is never used
   --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/event_loop.rs:350:27
    |
350 | pub struct EventLoopProxy<T: 'static> {
    |                           ^ unused parameter
    |
    = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`

error[E0392]: parameter `T` is never used
   --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/platform_impl/linux/mod.rs:635:20
    |
635 | pub enum EventLoop<T: 'static> {
    |                    ^ unused parameter
    |
    = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`

error[E0392]: parameter `T` is never used
   --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/platform_impl/linux/mod.rs:642:25
    |
642 | pub enum EventLoopProxy<T: 'static> {
    |                         ^ unused parameter
    |
    = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`

error[E0392]: parameter `T` is never used
   --> /home/<redacted>/.cargo/registry/src/index.crates.io-6f17d22bba15001f/winit-0.27.5/src/platform_impl/linux/mod.rs:769:32
    |
769 | pub enum EventLoopWindowTarget<T> {
    |                                ^ unused parameter
    |
    = help: consider removing `T`, referring to it in a field, or using a marker such as `PhantomData`
    = help: if you intended `T` to be a const parameter, use `const T: usize` instead

For more information about this error, try `rustc --explain E0392`.
error: could not compile `winit` due to 7 previous errors
warning: build failed, waiting for other jobs to finish...

... Not sure where to begin to fix this! Maybe updating the dep on winit would suffice?

I updated winit to 1.29.2, and, because there doesn't seem to be support for it already, decided to quickly update the platform code.

Fortunately, it's all pretty smooth! Although there are a bunch of breaking changes, everything seems to work just fine except one thing: winit::event::WindowEvent::ScaleFactorChanged

The current implementation uses new_inner_size, but this field has been replaced by inner_size_writer in the new version of winit. The new inner_size_writer doesn't allow you to get the suggested inner_size (i.e. new_inner_size) for some reason... So we have to have access to the Window (for Window::inner_size()) Maybe I haven't looked hard enough, but I don't see any way to access the window from anywhere inside Platform (the optional argument in end_frame doesn't work)

My question is: is passing a window reference to handle_event the only way to do this for now?

Note: I will make a pull request when this gets resolved (or should I do it now?)

Unlike the egui_glium platform, this platform doesn't seem to support zoom input. This code from the glium backend seems directly portable to this platform.

BTW. Other dependencies has new version.

Hello!

In 9fd9278, egui_winit_platform was upgraded to use winit 0.25. However, this has not been published to crates.io yet. Would be it be possible to cut a 0.8.0 release that uses winit 0.25?

Same as hasenbanck/egui_wgpu_backend#15 😉

Egui already tells us if it wants keyboard or mouse input, but it would be nice if the integration could sort out for us which winit events that includes. See this prior art for imgui. This is useful to be able to do other stuff with events that egui doesn't care about. If this is something that you would like to see, I could try making a pull request.

Trying to use egui_winit_platform on Android (and I guess the same might be true on iOS) we can only create a winit Window once the application receives a Resumed lifecycle event, even though, ideally, the Platform could be constructed once when initializing the app.

This is awkward because it leads to passing in dummy scale_factor and physical_width/height Descriptor values while we don't yet know anything about the Window we'll be rendering with.

If Platform is initialized with these dummy values then it's also necessary to explicitly update RawInput pixels_per_point and screen_rect to make up for this, which works against the way in which handle_event is designed to handle updating this state automatically in general.

The biggest difficulty though is that the private Platform scale_factor is practically inaccessible and the only (awkward) way this state can be corrected later is by passing a fake ScaleFactorChanged changed event to `handle_events.

All in all I currently do the following to update the Platform state that can't always be determined until the app is Resumed and we get a Window...

                let mut size = window.inner_size();
                let scale_factor = window.scale_factor() as f32;
                let raw_input = platform.raw_input_mut();
                raw_input.pixels_per_point = Some(scale_factor);
                raw_input.screen_rect = Some(egui::Rect::from_min_size(
                    egui::Pos2::default(),
                    egui::vec2(
                        size.width as f32,
                        size.height as f32,
                    ) / scale_factor,
                ));
                platform.handle_event::<Event>(&winit::event::Event::WindowEvent {
                    window_id: window.id(),
                    event: winit::event::WindowEvent::ScaleFactorChanged {
                        scale_factor: scale_factor as f64,
                        new_inner_size: &mut size }
                });

which is uncomfortably dependent on implementation details of egui_winit_platform.

I wonder if it could be best to remove the physical_width/height and scale_factor arguments from Platform::new() considering that this state can't be assumed to be available early on for all platforms.

Maybe there could a set_window API that could be used for associating the Platform with a winit Window which would be an opportunity to initialize both the physical width/height and scale factor which can then be correctly maintained by handle_event.

Perhaps though it's also possible to remove the separate scale_factor state internally and just refer to RawInput::pixels_per_point which seems to be the same scale factor?