tess::experimental::maintenance::RegisteredScheduler< Backend > Class Template Reference

#include <tess/experimental/registered_maintenance.h>

Public Member Functions
	RegisteredScheduler (std::size_t capacity)
	RegisteredScheduler (std::size_t registry_capacity, std::size_t backend_capacity)
	RegisteredScheduler (const RegisteredScheduler &)=delete
auto	operator= (const RegisteredScheduler &) -> RegisteredScheduler &=delete
	RegisteredScheduler (RegisteredScheduler &&)=delete
auto	operator= (RegisteredScheduler &&) -> RegisteredScheduler &=delete
auto	register_task (MaintenanceTask &task) -> std::optional< MaintenanceHandle >
void	seal ()
auto	valid (MaintenanceHandle handle) const -> bool
auto	try_schedule (MaintenanceHandle handle) -> std::optional< ScheduleResult >
auto	schedule (MaintenanceHandle handle) -> ScheduleResult
auto	run_some (MaintenanceBudget budget) -> DrainResult
auto	flush () -> DrainResult
auto	try_release (MaintenanceHandle handle) -> ReleaseResult
void	release (MaintenanceHandle handle)
auto	metrics () const noexcept -> MaintenanceMetrics

Detailed Description

template<typename Backend>
class tess::experimental::maintenance::RegisteredScheduler< Backend >

Fixed-registration handle and result contract over an experimental backend.

Register caller-owned tasks during setup, then call seal() exactly once. A handle belongs to one scheduler owner epoch and remains valid until checked or unchecked release, or scheduler destruction. Tasks are non-owning, non-movable, and must remain alive while registered.

try_schedule() and try_release() are the checked operations for expected stale-handle uncertainty. Their unchecked counterparts fail fast on wrong-owner, stale, or lifecycle misuse in every build. Release never cancels work. After sealing it requires a separate Idle observation made after the last successful schedule; that observation means only that the scheduler has no reachable execution. It does not prove that external dirty state is clean or that world/residency mutation is quiescent.

Callback exceptions propagate verbatim. The throwing invocation and its synchronous call-local reentrant chain are consumed, while independently retained accepted work remains reachable; callers retain authoritative dirty state and explicitly decide whether retry is safe. A callback may schedule through its own registered scheduler, but nested identity, scheduling, drain, or lifecycle operations on a different registered scheduler fail fast, independent of backend type. The no-throw metrics() snapshot remains callable across owners. FIFO and coalescing remain comparison machinery, and wrapping them grants no new execution authority.

Member Function Documentation

flush()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::flush ( ) -> DrainResult

inlinenodiscard

Runs reachable work with the backend's unbounded flush policy.

metrics()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::metrics ( ) const -> MaintenanceMetrics

inlinenodiscardnoexcept

Thread-safe read-only snapshot; permitted from any callback owner.

register_task()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::register_task ( MaintenanceTask & task ) -> std::optional< MaintenanceHandle >

inlinenodiscard

Registers a task during setup; empty reports fixed capacity exhaustion.

release()

template<typename Backend>

void tess::experimental::maintenance::RegisteredScheduler< Backend >::release ( MaintenanceHandle handle )

inline

Releases a known-live registration or fails fast on lifecycle misuse.

run_some()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::run_some ( MaintenanceBudget budget ) -> DrainResult

inlinenodiscard

Runs reachable work within a shared unit budget.

schedule()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::schedule ( MaintenanceHandle handle ) -> ScheduleResult

inlinenodiscard

Schedules a known-valid handle or fails fast on identity misuse.

seal()

template<typename Backend>

void tess::experimental::maintenance::RegisteredScheduler< Backend >::seal ( )

inline

Publishes the fixed registry and ends setup.

try_release()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::try_release ( MaintenanceHandle handle ) -> ReleaseResult

inlinenodiscard

Checked release. Post-seal release requires a fresh positive Idle result. It never cancels pending work and fixed capacity is not reused post-seal.

try_schedule()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::try_schedule ( MaintenanceHandle handle ) -> std::optional< ScheduleResult >

inlinenodiscard

Checked scheduling; empty means stale, foreign, or otherwise invalid.

valid()

template<typename Backend>

auto tess::experimental::maintenance::RegisteredScheduler< Backend >::valid ( MaintenanceHandle handle ) const -> bool

inlinenodiscard

Returns whether a handle currently names a live local registration.

---

The documentation for this class was generated from the following file:

• include/tess/experimental/registered_maintenance.h