Establish SDL3 platform, input, audio, and supported-target policy

[zoeyrose]

Outcome

Define one supported SDL3 platform layer for Linux and Windows that can serve the connected client without leaking platform details into session or rendering crates.

Scope

• Window creation/lifecycle, DPI/display changes, fullscreen/windowed modes, focus/minimize/suspend, clipboard, cursors, text input/IME, file dialogs where needed, and clean shutdown.
• Keyboard, mouse, controller/gamepad, touch where supported, hotplug, remapping, repeat, chords, and focus routing into semantic actions.
• SDL3 audio device ownership, format conversion/mixing policy, hotplug/loss, category volume/mute, and resource-provider input.
• Native dependency acquisition/pinning, static/dynamic packaging choice, Vulkan/D3D12 prerequisites, and supported fallback/error behavior.
• Deterministic headless substitutes for platform services used in tests.
• Accessibility hooks and input-device-independent focus/navigation expectations.

Acceptance criteria

• A harness repeatedly creates/destroys windows, resizes, changes DPI/fullscreen, hotplugs input/audio, and exits without leaks or stale callbacks.
• Logical coordinates and scale are unambiguous between SDL events, UI, and renderer.
• Input tests prove keyboard/controller parity for semantic actions and text fields.
• Device failures produce actionable UI/state transitions rather than panic or hidden fallback.
• Supported target/driver matrix is documented and enforced in CI/release checks.

Dependencies

Coordinates with renderer’s SDL3/wgpu surface bridge, but owns no GPU rendering. Product control behavior remains in the transferred controls issue.

[github-actions]

🎉 This issue has been resolved in version 1.0.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀