test: harden full-core-loop against Load/Fly precondition races

2026-09-19 13:36:27 +00:00
parent 687e4f2169
commit ae1cb93a4f
2 changed files with 28 additions and 6 deletions
@@ -13,6 +13,7 @@
 - e2e (`right-pane.spec.ts`): new test — fly the longest hub leg on a full tank, load a package back to the hub at the far end, click the row "Refuel" and assert via the endpoint's own response snapshot that cash drops by exactly L×price and fuel rises by exactly L (L from the mirrored closed form); also asserts the market button label is "Refuel to full". `e2e/api.spec.ts` path list unchanged
 - Verified: `tsc --noEmit` clean; unit suite green locally; e2e not run locally by design — CI verifies against an isolated stack
 - **Fix (CI run #40): forFlight refuel leaves an exactly-feasible tank, and the new e2e's premise corrected.** The right-pane e2e picked the dataset's absolute farthest airport from EDDF — UNNT at ~4803 km, whose leg needs ~6955 L > the 6300 L tank (max range from empty is ~4436 km) — so the full-tank fly rightly 409'd; `farthestToHub` becomes `farthestFlyableToHub`, picking the farthest airport whose leg still fits a full tank (UWUU, ~3195 km). Server hardening in the same area: forFlight now sets fuel to EXACTLY `requiredFuelFor(...)` instead of `fuel + liters` — that sum is not guaranteed ≥ R in IEEE754 (`f + fl(R − f)` can round a hair below), and flyPlane's strict comparison would then 409 on an otherwise-feasible leg; cost math unchanged (still the added liters × price). New unit regression test: forFlight on an empty-ish tank → immediate `flyPlane` to the package destination succeeds
+- **E2e flake hardening (runs #36/#41): full-core-loop Load/Fly precondition races** — test mechanics only, no app code changes. Two independent races in `e2e/actions.spec.ts`: (1) the Fly click fires its POST asynchronously (`runAction` is fire-and-forget from the test's view), so the immediate `getState` could race ahead of it and still see status 'available' — now an `expect.poll` waits until the server shows 'flying' with the expected destIcao before proceeding; (2) `flyHomeViaApi` made its repositioning decision from a snapshot taken BEFORE waiting for a leftover in-flight leg to land — while flying, locationIcao stays at the departure airport, so a plane that landed away from home was still believed to be home and never repositioned (run #41 retry: starter stranded at EDFH) — state is now re-read after landing; plus an explicit server-side precondition poll (starter grounded at the package origin) before clicking Load

 ## v0.15.1 — 2026-09-19 11:04
 - **Right pane widened, "On board" wrapping fixed (ISS-004).** The 3-pane shell grid (`app/globals.css` `.app-shell`) moves from `300px / 1fr / 320px` to `280px / 1fr / 360px` — the map column absorbs the delta (MapLibre sizes itself from its container, no JS change). Three wrap offenders in the right pane's "On board" blocks are fixed: