Keyboard (and mouse) input silently dropped on Windows ARM64

[andreban]

all keyboard and mouse events are silently dropped. The application window opens and renders correctly, but is_key_down, is_key_pressed, and all other input queries always return false/no events.

The same binary compiled for x86_64-pc-windows-msvc works correctly on x86-64 Windows hardware.

---

Root Cause

In src/native/windows.rs, the run() function stores a pointer to WindowsDisplay in the window's user data slot via SetWindowLongPtrA. This pointer is later retrieved by the window procedure (win32_wndproc) to dispatch events. The storage call is guarded by architecture #[cfg] attributes, but aarch64 is not covered:

// src/native/windows.rs — run() function, ~line 945

#[cfg(target_arch = "x86_64")]
SetWindowLongPtrA(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);
#[cfg(target_arch = "x86")]
SetWindowLong(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);

On aarch64, neither cfg block is compiled, so SetWindowLongPtrA is never called. The user data slot remains 0.

The window procedure retrieves the pointer using target_pointer_width = "64" (which correctly covers aarch64), but always reads 0:

// src/native/windows.rs — win32_wndproc()

#[cfg(target_pointer_width = "64")]
{
    display_ptr = GetWindowLongPtrA(hwnd, GWLP_USERDATA)  // always returns 0 on aarch64
}

if display_ptr == 0 {
    return DefWindowProcW(hwnd, umsg, wparam, lparam);  // every event is dropped here
}

Because display_ptr is always 0, win32_wndproc returns early for every window message, including WM_KEYDOWN, WM_KEYUP, WM_MOUSEMOVE, etc. No events ever reach the application.

The same bug exists in the set_fullscreen method (~line 164), which has the identical architecture gap:

#[cfg(target_arch = "x86_64")]
SetWindowLongPtrA(self.wnd, GWL_STYLE, win_style as _);
#[cfg(target_arch = "x86")]
SetWindowLong(self.wnd, GWL_STYLE, win_style as _);
// aarch64 → neither branch compiles → window style is never updated

---

Fix

Add aarch64 to the x86_64 branch in both locations, since ARM64 Windows is a 64-bit platform and uses SetWindowLongPtrA (not the 32-bit SetWindowLong):

run() function:

// Before
#[cfg(target_arch = "x86_64")]
SetWindowLongPtrA(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);
#[cfg(target_arch = "x86")]
SetWindowLong(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);

// After
#[cfg(any(target_arch = "x86_64", target_arch = "aarch64"))]
SetWindowLongPtrA(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);
#[cfg(target_arch = "x86")]
SetWindowLong(wnd, GWLP_USERDATA, &mut display as *mut _ as isize);

set_fullscreen() method:

// Before
#[cfg(target_arch = "x86_64")]
SetWindowLongPtrA(self.wnd, GWL_STYLE, win_style as _);
#[cfg(target_arch = "x86")]
SetWindowLong(self.wnd, GWL_STYLE, win_style as _);

// After
#[cfg(any(target_arch = "x86_64", target_arch = "aarch64"))]
SetWindowLongPtrA(self.wnd, GWL_STYLE, win_style as _);
#[cfg(target_arch = "x86")]
SetWindowLong(self.wnd, GWL_STYLE, win_style as _);

---

Reproduction

1. Any macroquad/miniquad application that reads keyboard or mouse input
2. Compile for aarch64-pc-windows-msvc
3. Run on a Windows ARM64 machine (e.g. Snapdragon X, Surface Pro X)
4. Observe: window opens, renders, but all input is ignored

Not reproducible when the same code is compiled for x86_64-pc-windows-msvc and run on x86-64 Windows.

---

Notes

• The win32_wndproc retrieval already uses target_pointer_width = "64" (correct for aarch64), so only the storage side needs fixing.
• Alternatively, both sites could be simplified to use target_pointer_width consistently to future-proof against other 64-bit Windows targets.

[andreban]

@not-fl3 I have this fixed on a local branch. Happy to send a PR if you agree