Struct TimerHandle 

pub struct TimerHandle { /* private fields */ }

Cheap, clonable submission side of a DelayedTimer — the seam route for deferred (SDLY/ODLY/watchdog-style) hand-offs.

Implementations

impl TimerHandle

pub fn schedule( &self, delay: Duration, priority: CallbackPriority, cb: Callback, )

Schedule deferred work cb to be enqueued on priority after delay. Returns immediately — the callback runs on a pool worker no earlier than delay from now. Port of callbackRequestDelayed (callback.c:410).

pub fn schedule_wake(&self, delay: Duration, cb: Callback) -> Option<WakeKey>

Schedule a non-blocking wakeup cb to run inline on the timer thread after delay — the sleep/interval waker path. cb MUST be trivial and non-blocking (only a waker wake(): an unpark or a tokio task-schedule); it runs on the single timer thread and delays every later deadline until it returns.

This exists so a spawned future that awaits sleep is never blocked by its own wake: running it on the timer thread frees the band from the dual role of “run futures” AND “wake them”. It is what closed the sleep-wake self-deadlock (bug_pattern rtems-exec-sleep-wake-band-deadlock), back when future_exec parked a worker per future for the future’s whole life; that executor is now cooperative and holds a worker only across a single poll, but a wake still costs no worker here.

The returned WakeKey is the caller’s claim on the queued entry, and the caller MUST pass it to cancel_wake when it stops caring about the wakeup. None means the timer had already shut down and cb was dropped unscheduled — there is nothing to cancel.

pub fn cancel_wake(&self, key: WakeKey)

Drop the wake entry filed under key now, instead of leaving it queued until its deadline. Cancelling an entry that has already fired is a no-op, so a caller never has to know which happened first.

Unlike schedule, which is C’s fire-and-forget callbackRequestDelayed, a wake belongs to the sleeper that armed it: the entry holds a clone of the sleeper’s shared cell, so an uncancelled entry keeps that cell — and the OS mutex inside it — alive for the whole remaining delay even though the sleeper is long gone.

pub fn scheduled_count(&self) -> usize

How many entries are queued. For tests and on-target probes: the per-sleep retention this queue used to carry is only visible as a count.

Trait Implementations

impl Clone for TimerHandle

fn clone(&self) -> TimerHandle

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.