Fix macOS volume and media key posting

[jonstuebe]

Summary

• Post macOS volume and media controls as NSSystemDefined subtype 8 NX events instead of standard keyboard events
• Wire Volume Up/Down/Mute plus Play/Pause/Next/Previous through the shared media-key helper
• Add macOS objc2 AppKit/CoreGraphics bindings needed to create and post NSEvents

Validation

• cargo check -p openlogi-core
• cargo fmt --check

Fixes mouse-button bindings for Volume Up/Down/Mute on macOS, where kVK_Volume* keyboard events are ignored by the system volume handler.

[greptile-apps]

Greptile Summary

This PR replaces the placeholder post_media_key stub with a real NSSystemDefined subtype-8 event pipeline using the objc2 AppKit/CoreGraphics bindings, and routes all six media/volume actions (VolumeUp, VolumeDown, MuteVolume, PlayPause, NextTrack, PrevTrack) through it instead of the broken kVK_Volume* keyboard-event path.

• NX_KEYTYPE_* constants match <IOKit/hidsystem/ev_keymap.h>, and the data1 layout — (keyCode << 16) | (state << 8) — is the standard encoding used by well-known reference implementations.
• post_media_key loops over a (key-down, key-up) pair, creating and posting an NSEvent → CGEvent for each; error branches inside the loop use return rather than break, which can leave a key-down posted without a matching key-up if the second iteration fails.

Confidence Score: 3/5

The approach is correct and the constants are accurate, but the loop's early-return on failure can leave a key-down event posted without a paired key-up, which is worth fixing before merge.

The post_media_key function posts both a key-down and a key-up NSSystemDefined event inside a single for loop. If the key-down iteration succeeds and the key-up iteration then fails, the return in the else branch exits the function entirely, leaving the key-down event in the HID stream with no matching key-up.

crates/openlogi-core/src/binding.rs — specifically the error branches inside the post_media_key loop

Important Files Changed

Filename	Overview
crates/openlogi-core/src/binding.rs	Replaces stub media-key implementation with real NSSystemDefined subtype-8 posting using objc2 AppKit/CoreGraphics bindings; NX_KEYTYPE constants are correct, data1 encoding matches the standard layout, but return inside the down/up loop can leave a key-down posted without a matching key-up if the second iteration hits an error branch.
crates/openlogi-core/Cargo.toml	Adds objc2 ecosystem crates (objc2, objc2-app-kit, objc2-core-graphics, objc2-foundation) under the macOS target; versions are internally consistent and the NSGraphicsContext feature is included (likely needed for NSEvent.CGEvent() via AppKit category).
Cargo.lock	Lock file update reflecting new objc2-app-kit transitive deps (block2, libc, objc2-metal); purely mechanical change consistent with the Cargo.toml additions.

Sequence Diagram

sequenceDiagram
    participant A as Action::execute_macos
    participant M as macos::post_media_key
    participant NS as NSEvent (AppKit)
    participant CG as CGEvent (CoreGraphics)
    participant OS as macOS HID System

    A->>M: "post_media_key(NX_KEYTYPE_*)"
    loop key-down then key-up
        M->>NS: "otherEventWithType(SystemDefined, subtype=8, data1)"
        NS-->>M: "Option<NSEvent>"
        M->>NS: ns_event.CGEvent()
        NS-->>M: "Option<CGEvent>"
        M->>CG: CGEvent::post(HIDEventTap, cg_event)
        CG->>OS: deliver NX_SUBTYPE_AUX_CONTROL_BUTTONS event
    end

Loading

_{Reviews (1): Last reviewed commit: "Fix macOS volume and media key posting" | Re-trigger Greptile}

[greptile-apps]

Early return can leave key-down posted without matching key-up

If the first iteration (key-down) posts successfully and the second iteration (key-up) then hits either NSEvent::otherEventWithType returning None or ns_event.CGEvent() returning None, the return exits the function entirely — leaving a key-down event in the system with no paired key-up. The post_click function in the same module handles this more gracefully with if let Ok. Using break instead of return in the failure branches would avoid posting orphaned key-down events for volume/media while still aborting any remaining work.