Pause on landing + auto-unpause on dispatch (ISS-041)
CI / build + unit + e2e (pull_request) Successful in 7m43s

Replace the ISS-011 auto-slow with pause-on-landing: when the last plane
in flight lands, tickOnce sets sim_state.paused=true WITHOUT touching
timeScale (speed preserved). flyPlane() now clears paused before
persisting so starting a flight resumes a paused sim — inside runAction(),
tick + unpause + dispatch are atomic under the sim lock.

Tests: engine.test.ts auto-slow block rewritten for pause semantics; new
flyPlane test asserts dispatch clears paused with speed untouched. E2E:
auto-slow.spec.ts renamed to pause-on-landing.spec.ts asserting the new
behavior (paused=true AND timeScale still 1000 after landing; /fly clears
paused). Cross-cutting fix: every per-spec setScale helper now POSTs
{timeScale, paused:false} — a scale-only fast-forward after a landing would
otherwise leave the clock frozen. Deliberate mid-test pauses untouched.
This commit is contained in:
2026-09-20 09:48:56 +00:00
parent 4620707abc
commit cbe0cda15d
12 changed files with 105 additions and 64 deletions
+6
View File
@@ -1,5 +1,11 @@
# Changelog
## v0.24.0 — 2026-09-20 09:48
- **Landing now pauses the sim instead of auto-slowing to 1× (ISS-041).** When the LAST plane in flight lands, `tickOnce` (`lib/sim/engine.ts`) sets `sim_state.paused = true` WITHOUT touching `timeScale` — speed is preserved. The trigger condition carries over from ISS-011's auto-slow unchanged: ≥1 landing this tick AND no `'flying'` planes remain (a landing while another plane is still airborne does not pause — there would be nothing to auto-resume it). No idle-enforcement or auto-restore of the pause exists
- **Starting a flight auto-resumes a paused sim (ISS-041).** `flyPlane()` (`lib/game/actions.ts`) writes `sim_state.paused = false` before persisting — inside `runAction()`, so tick + unpause + dispatch are atomic under the sim lock. The cycle: dispatch unpauses → last landing pauses (speed preserved) → next dispatch unpauses again
- Tests: unit — `lib/sim/engine.test.ts`: the auto-slow describe block rewritten for pause semantics (last plane lands → `paused === true` AND stored `timeScale` UNCHANGED, persisted on the DB row; a landing while another plane is still flying → no pause; last landing at scale 1 → pauses anyway; paused tick with a fully grounded fleet → trigger never fires). `lib/game/actions.test.ts`: new test — dispatching on a paused sim clears `paused` and leaves `timeScale` untouched. E2E — `e2e/auto-slow.spec.ts` renamed to `e2e/pause-on-landing.spec.ts` and reworked: after the starter lands, asserts `simState.paused === true` AND `timeScale` still 1000 (NOT reset), then a subsequent `/fly` dispatch clears `paused` on its own post-action snapshot. Cross-cutting fix for the new pause: every per-spec `setScale` helper (`actions`, `hover`, `map-interactions`, `relocation`, `right-pane`) now POSTs `{ timeScale, paused: false }` — a landing pauses the sim, so a scale-only fast-forward after one would leave the clock frozen and hang the next wait; no caller relied on it NOT clearing pause (verified per call site). Deliberate mid-test pauses (`sim.spec.ts` / `panes.spec.ts` freeze tests, marker-click freezes) are untouched
- Verified: `tsc --noEmit` clean; unit suite green locally (291); `npx playwright test --list` = 42 tests in 12 files; e2e not run locally by design — CI verifies against an isolated stack
## v0.23.2 — 2026-09-20 09:15
- **On-board package hover previews the route from the carrier's current position (ISS-040).** Hovering a loaded package in the left pane used to draw the great-circle trace from the package's ORIGIN airport; it now starts at the carrying plane's live lat/lon — identical while grounded at the pickup, correct once the plane has departed with cargo still loaded. New pure builder `hoverTraceGeoJson(from: AirportCoord, toIcao, airportsByIcao)` in `lib/mapData.ts` (sibling of `hoverRouteGeoJson`, same shape/conventions; empty collection when the destination is missing); `components/Map.tsx`'s `applyHoverRef` looks up the carrier via the hover's `planeId` and uses it, falling back to the ICAO trace when no carrier is in state. The `[state]` effect now also re-applies the hover so the trace follows a moving carrier between polls while hovered (idempotent: one small setData + ring filters). Endpoint rings untouched — still ICAO-filtered origin/dest circles
- Tests: unit — `lib/mapData.test.ts` gains a `hoverTraceGeoJson` block (exact endpoints from an arbitrary start coord; empty on missing dest). E2E — `e2e/hover.spec.ts`: new test loads a package at EDDF, dispatches the starter to LEMD via API, waits until its live position is >0.2° away from the hub, pauses the sim, hovers the on-board row and asserts some rendered trace segment starts within EPS of the carrier's live lat/lon AND some ends within EPS of LEMD AND no segment starts near EDDF (per-tile clipping handled via the existing cross-segment endpoint pattern); cleanup resumes + fast-forwards to arrival (auto-delivery) and flies home