Root cause: state.runtimes was empty after refresh because the engine
only pushes UnitRuntimeChanged on state transitions — if the engine
is mid-wait-phase, no push occurs and badges show OFFLINE.
Fix: add GET /api/unit/runtimes batch endpoint (returns all known
runtimes as { unit_id: UnitRuntime }) and call it in parallel with
the unit list fetch inside loadUnits(), so runtime badges are correct
immediately after page load.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- engine.rs: replace HashSet<Uuid> with HashMap<Uuid, JoinHandle> in
supervise(); use is_finished() to detect exited tasks so units that
are disabled then re-enabled get a new task on next 10s scan
- control/mod.rs: extract shared monitor_value_as_bool (using the more
complete validator version that includes "yes"); remove duplicate
copies from engine.rs and validator.rs
- runtime.rs: fix get_or_create_notify TOCTOU by using entry API
instead of read-drop-write pattern
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Engine now spawns one async task per enabled unit (supervised every 10s)
- wait_phase uses sleep_until + select! for precise timing; 500ms fault-tick
runs inside each phase so fault/comm is still checked promptly
- WS UnitRuntimeChanged pushed only on state transitions, not every tick
- ControlRuntimeStore gains notify_unit/get_or_create_notify for instant
wake-up when handlers change auto_enabled or fault_locked
- UnitRuntime: remove last_tick_at, current_run/stop/distributor_elapsed_sec;
add display_acc_sec (snapshot at transition, avoids mid-cycle jitter)
- accumulated_run_sec now increments by exact run_time_sec*1000 per cycle
- unit.state_changed events no longer written to DB (too frequent)
- Frontend: show display_acc_sec instead of accumulated_run_sec
- styles: event-card flex-shrink:0 fixes text overlap under flex column
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
caoqianming