Struct RefreshScheduler 

pub struct RefreshScheduler {
    closures: Mutex<HashMap<String, Arc<RegisteredClosures>>>,
    tables: Mutex<ReducerTables>,
    notifiers: Mutex<HashMap<String, Arc<Notify>>>,
}

Fields

closures: Mutex<HashMap<String, Arc<RegisteredClosures>>>

tables: Mutex<ReducerTables>

notifiers: Mutex<HashMap<String, Arc<Notify>>>

Implementations

impl RefreshScheduler

pub fn new() -> Self

pub async fn register( &self, credential_id: impl Into<String>, is_fresh: impl Fn() -> Pin<Box<dyn Future<Output = bool> + Send + 'static>> + Send + Sync + 'static, refresh: impl Fn() -> Pin<Box<dyn Future<Output = Result<(), RefreshErrorKind>> + Send + 'static>> + Send + Sync + 'static, )

Register (or replace) a credential for proactive management. is_fresh must be side-effect-free (may be called on every ensure_fresh/sweep tick) — async so a caller whose freshness check needs a blocking read (e.g. an OS keychain call, which can hang on a slow/unresponsive Secret Service D-Bus daemon on headless Linux) can route it through tokio::task::spawn_blocking instead of stalling the tokio worker thread this scheduler’s own async tasks run on (reagent P1 on #2260). refresh performs the actual refresh and is responsible for persisting the result itself (and for NOT persisting anything on failure) — and for classifying a failure as RefreshErrorKind::Transient (worth retrying) vs ::PermanentAuthFailure (the credential itself is rejected; only a fresh login can fix it — see that type’s own doc comment).

pub fn state(&self, credential_id: &str) -> Option<CredentialState>

The credential’s current coordination state, if registered. Read by muxbus::stale_sign_in to surface NeedsReauth distinctly from a routine in-progress refresh. Not used by ensure_fresh itself, which always dispatches through the reducer regardless.

pub async fn ensure_fresh(&self, credential_id: &str) -> Result<(), String>

Ensure credential_id is fresh, single-flight-guarded. If another caller is already mid-refresh for this id, this waits for that attempt to finish, then re-checks freshness rather than starting a second refresh. Returns Err if the id was never registered.

pub async fn run_sweep_loop(self: Arc<Self>, interval: Duration)

Background sweep — checks every registered credential every interval and proactively refreshes anything stale. Skips entries currently NeedsReauth: retrying a credential already known to require a fresh human login is pointless and just hammers the backing server; ensure_fresh called on-demand still re-checks those (see state’s own doc comment). Spawn once via tokio::spawn; runs until the process exits.

fn notifier_for(&self, id: &str) -> Arc<Notify>

fn wake(&self, id: &str)

fn trace(&self, events: &[Event])

Map diagnostics-only reducer events to tracing — auth.broker.* message prefixes land in muxlog auth’s existing filter regex (\bauth\.\w+|...) with zero changes needed there. Control-flow events (RunFreshnessCheck/RunRefresh/AlreadyInFlight/ Unregistered) are acted on directly by callers and not double-logged here — every ensure_fresh call would otherwise emit noise even when nothing changed.

Trait Implementations

impl Default for RefreshScheduler

fn default() -> Self

Returns the “default value” for a type.