fusion_clock doesn't depend on planning, so planning's models load AFTER it
during -u and the port saw no data. Now detect planning tables via SQL,
defer (no marker) when the ORM isn't loaded, and finish the port from the
deploy odoo-shell step (full registry). Marker now owned by the method.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
is_open + crosses_midnight fields; employee_id optional (open shifts);
company_id computed w/ env.company fallback; drop hard one-per-day UNIQUE
(allow split + open). Overnight math in planned_hours/_check_schedule_times/
scheduled_times. _get_fclk_day_plan resolves multiple posted rows into ONE
work-window so penalties/overtime/absence stay correct. Migration drops the
old constraint defensively. Tests for overnight, window, open shifts.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
post-migrate(19.0.5.0.0) -> fusion.clock.schedule._fclk_port_planning_data:
planning.role -> fusion.clock.role, employee default/allowed roles, and
planning.slot -> fusion.clock.schedule (local date+float, role map, posted
if published, open if unassigned). Guarded (no-op on Community), idempotent
(marker), per-row savepoints. Integration test runs on Enterprise clones.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
