Winit version 0.28.0

[kchibisov]

Patches that left

Some of the patches that might be worth merging for 0.28.0:

• Add Window::is_minimized #2626
• Rework DnD #2615 Will take more design work, so let's not block on this one
• Properly remove window mouse event listeners #2632 (web, so can't review, cc @maroider ?)
• Allow introspection of WindowBuilder default values. #2613 (not sure)
• Fix middle/other mouse buttons on macOS #2605 (bevy)
• Add WindowBuilder::with_active #2585 (awaits macOS)
• Implemented option_as_alt to allow for ignoring option key characters #2576 (macOS patch about Option as Alt)
• wayland: Add support for fractional scaling #2369 (fractional scaling on Wayland)
• [MacOS] Fix deadlock on maximizing window from event callback #2636 (required for egui)
• Bump windows-sys version to 0.45 #2639 (regression on master)
• On Windows, fix window size for maximized, undecorated windows #2584

If I've missed something or you want some bug to be fixed for v0.28.0 let me know.

Be aware that the next major release after the 0.28.0(0.29.0) will contain the new Wayland backend (updated to new wayland-rs) and likely new X11 backend(mixture of Xlib + Xcb, instead of raw xlib).

--
CC of some main downstream consumers(if you're not interested in getting notified about release planning in winit let me know, so I won't even CC you again).

• bevy: @alice-i-cecile, @cart
• egui/eframe: @emilk
• iced: @hecrj
• ggez: @PSteinhaus
• tauri : @amrbashir
• nannou: @mitchmindtree

If you have some issues with winit that you want fixed for 0.28.0 let me know while we're in release preparation stage, so I include them and try to look into.

Also would you prefer if winit had rc period or you're fine with the current policy with patch fixes right away, given that winit don't really have breaking changes. Initially
we decided to not do rc releases, since it's easier to iterate and propagate the update without them, given that majority unlikely to use them and it's better to know that something broke if you push the update.

[alice-i-cecile]

#2605 is the most important of those for us at Bevy :)

[Wumpf]

#2636 is blocking an egui PR for maximizing windows, would love to see that in if possible (:

[MarijnS95]

@rib we have no breaking changes planned anymore for android-activity, right? Just making sure, otherwise we might want to push that out and update winit right before releasing 0.28.

(EDIT: And if not, we should do a nonbreaking patch release for the recent contributions, i.e. to reduce logcat spamming 🙂)

[madsmtm]

I think we should go for the new release now, all the requested changes have been made (or I've written a description on them as to why we'll delay with it).

[kchibisov]

The option as alt PR is still up, and I'll try to finish it today... The Alt PR should be trivial to finish anyway.

[rib]

@rib we have no breaking changes planned anymore for android-activity, right? Just making sure, otherwise we might want to push that out and update winit right before releasing 0.28.

(EDIT: And if not, we should do a nonbreaking patch release for the recent contributions, i.e. to reduce logcat spamming 🙂)

Right, no breaking changes planned currently.

Doing a release for the logging cleanup should be fine. It'll also propagate the new url for the repo.

[rib]

I had a few winit Android issues I started looking at and had wanted to make more progress with (ime + surface_size) but it's fine if those things slip to the next release.

For Android I've been looking more at issues in the JNI crate recently since I want to use it in the Android backend for Winit and but it has had a bunch of ergonomic and safety issues that I think need addressing first - so I've been a bit distracted from winit and android-activity.

[emilk]

I've tested latest winit with eframe and it seems to we working flawlessly, FWIW. I really like the new WindowEvent::TouchpadMagnify event ❤️

[kchibisov]

I'll schedule release for today's evening (UTC+3). If no-one has anything else to add I'd push a release.

[dhardy]

Winit master works with KAS; the one slight concern I note is that EventLoopWindowTargetExtWayland is now under cfg(wayland_platform) which is less easy to reproduce under downstream crates. For Winit 0.27 it's mostly copy-paste:

#[cfg(any(
    target_os = "linux",
    target_os = "dragonfly",
    target_os = "freebsd",
    target_os = "netbsd",
    target_os = "openbsd"
))]
{
    use winit::platform::unix::EventLoopWindowTargetExtUnix;
    if elwt.is_wayland() {
        // Wayland-specific code here
    }
}

Now one needs to read the build.rs, but at least that has a similar cfg there (due to "XXX aliases are not available for the build script itself").

No action needed; better docs may be possible.

[kchibisov]

nah, will wait for #2665...

[MarijnS95]

@kchibisov thanks for releasing! Can you now clean up some of the branch inconsistencies? There's a 0.27 branch that points to the 0.27.1 release whereas 0.27.x points to the latest 0.27.5 release. And the new v0.28.x branch has a v in front of it; replicate that on the other branch?

@madsmtm somehow strangely we seem to have some feature resolver problems on Rust 1.59 getting to objc2 = "=0.3.0-beta.3": https://github.com/ash-rs/ash/actions/runs/4072998223/jobs/7016396013, any idea/suggestions?

[kchibisov]

@MarijnS95 msrv is 1.60 not 1.59 for the new winit.

[kchibisov]

@kchibisov thanks for releasing! Can you now clean up some of the branch inconsistencies? There's a 0.27 branch that points to the 0.27.1 release whereas 0.27.x points to the latest 0.27.5 release. And the new v0.28.x branch has a v in front of it; replicate that on the other branch?

I'll try to look into that, but not sure if I can change much.

[MarijnS95]

@kchibisov right, silly of me to expect cargo to complain about Rust 1.59 being too low for the set MSRV of 1.60 in winit before trying (and failing) to resolve winit's dependencies...