# Revalidating a slow native mortar aim

The native run `logs_2026-09-19_08-21-23/rereintroduced-hq.log` records the initial ballistic solution at mission time 32.9s (event 44) and the native shooting node at 57.2s (events 53–54). The gun took approximately 24.3 seconds to reach that node; the successor's 10-second authorization lease therefore rejected an otherwise still-fresh stationary delivered observation before any shot. The report was observed around 17.8 seconds, so this failure was distinct from the unchanged 60-second observation expiry.

`ERR_MortarFiring.c` now uses one ballistic solver body for initial aim and later authorization. It still calls the original node's installed `GetAmmoInitialSpeedCoefficients(ResourceName, notnull array<float>)`, `GetVerticalAimLimits(notnull IEntity, out float, out float)` and the native `BallisticTable.GetAimHeightOfProjectileAltitudeFromSource`. Native coefficient order, elevation checks, moving-intercept iterations, terrain, range and safety checks are preserved. Only initial aiming writes native aim/charge output ports; revalidation never changes those outputs or the physical barrel.

Each successful initial aim retains its commanded geometry and a weak reference to the native solver node. An expired 10-second lease can be renewed only after verifying that the same node still addresses the same gun/ammunition and that the gun, delivered-report object, order and shot match the original command. A new solve must use the same charge, remain within 0.05 m of the original gun origin, within 0.1 m of the original impact and aim target in three dimensions, and within 0.05 seconds of the original flight time. The original solution remains the comparison reference across every renewal; repeated small target movements cannot accumulate into permission to fire somewhere the barrel never aimed.

The report's observation time is never changed. Deleted/replaced nodes, superseded reports/orders/shots, expired observations, failed prediction, changed charge/geometry or safety rejection still withhold the shot and use the existing next-HQ-tick cancellation. Physical projectile confirmation and cancellation both clear the retained command/node binding. Player-operated and unrelated artillery retain their native behavior.

The existing firing authorization runs before the native shooting action and again immediately before physical shell transfer. A renewal is not evidence of a shot: only the existing non-null native projectile event counts ammunition fired. Finite shell supply, native loading/alignment, projectile creation and airspace reservation remain unchanged.

`tests/test_mortar_aim_renewal.py` executes the actual equivalence, report freshness, command matching and renewal control-flow bodies with synthetic API inputs. It covers repeated recomputation, unchanged report age, expiry/future time, lost/replaced node inputs, wrong report/order/shot, failed solver, charge/origin/height/flight changes, non-finite flight values and cumulative drift against the original command. The ballistic output is deliberately a fake test boundary; these checks do not demonstrate native aim or firing. Installed source and failing-run hashes are recorded in `mortar-aim-renewal-sources.json`.

Native follow-up must rerun the real stationary-report firing fixture: require `AIM REVALIDATED`, unchanged report source/observation age, and a real projectile event before expiry, with complete raw logs and no engine/script errors. A moving report or changed gun/charge must still be withheld, and later reports must continue to arrive by physical radio delivery. Full salvo completion is subject to report freshness and finite physical supply; the change does not guarantee eight rounds from one expired report.