Module idle 

“Seconds since the last input” probing.

user_idle wraps the windowing system’s native idle counter — XScreenSaver on X11, IOKit HIDIdleTime on macOS, the Win32 last-input timer on Windows. On GNOME/Wayland none of those exist: Wayland deliberately doesn’t expose a global idle counter, so UserIdle::get_time() returns Status not OK and every idle-driven feature (idle-reset, typing-defer, screen-time accounting) silently treats the user as permanently active (#190).

GNOME’s compositor publishes the same counter on the session bus as org.gnome.Mutter.IdleMonitor.GetIdletime (milliseconds since last input). When the primary user_idle probe fails we fall back to it. Like super::session_lock, we shell out (gdbus) rather than link a D-Bus crate — one integer read isn’t worth 30+ transitive dependencies — and reuse the shared 2 s probe timeout so a wedged session bus can’t stall the scheduler tick.

Modules

mutter 🔒

Functions

combine_idle 🔒

Resolve the primary reading against the platform fallback: the primary value wins; on primary failure the injected fallback is consulted and, if it too has nothing, the primary error is surfaced. The fallback is a closure so this decision is pure and unit-testable on every OS without touching a windowing system — only the two FFI sources it’s wired to in idle_secs stay platform-bound.

fallback_idle_secs 🔒

The platform fallback when user_idle can’t read the counter. Only GNOME/Wayland has one (Mutter); everywhere else there’s nothing more to try, so the primary error stands.

idle_secs

Seconds since the last keyboard/mouse input, or a display-error string if no idle source is available this call.

parse_mutter_idletime_ms 🔒

Parse the gdbus call … GetIdletime reply — (uint64 12345,) — into idle milliseconds. Pure so the Wayland fallback’s parsing is testable without a live session bus; the gdbus spawn stays in mutter.