Merge branch 'main' into codex/fix-seek-layer-state

changelog.d/1207-voice-entity-labels-english.md

@@ -0,0 +1,3 @@
+- Keep NASA FIRMS fire and ALPR camera labels English in voice tool results
+  (`track_entity`, `get_entity_context`, pointer context); the voice card still
+  shows them in the interface language (daikaginza, #1207).

changelog.d/1226-first-run-touch-dismiss.md

@@ -0,0 +1,4 @@
+- The first-run card can be closed on a phone or tablet. "ESC to dismiss" is
+  now a button that closes the card the same way ESC does; touch screens read
+  DISMISS, and phones no longer hide it, so choosing a mission is not the only
+  way out (KnottyDyes, #1226).

changelog.d/1239-traffic-stale-flow.md

@@ -0,0 +1,3 @@
+- `get_traffic_flow` flags stale cached TomTom tiles and reports the oldest
+  tile fetch time (`stale`, `oldest_tile_at`), with pending freshness metadata
+  released on failed or cancelled tile reads (quanticsoul4772, #1239).

changelog.d/1244-terrain-test-cache-isolation.md

@@ -0,0 +1,4 @@
+- The terrain-heights bounds tests no longer read the checkout's
+  `.gev-cache/terrain-heights.json`. Each test gets a cache file of its own, so
+  `npm test` no longer fails after the dev server has cached terrain near
+  Austin (KnottyDyes, #1244).

changelog.d/1282-polling-issue-fix.md

@@ -0,0 +1,7 @@
+- Pause periodic network updates in hidden tabs; the overdue update on return restarts the polling schedule, so the next poll follows one interval later even when that update fails (asturksever, #1282).
+- Report and guard browser visibility during Transit QA so hidden-page polling pauses cannot masquerade as playback regressions (asturksever, #1282).
+- Require each live city to supply its own feed vehicles in Transit QA; retained vehicles cannot qualify another city (asturksever, #1282).
+- Bound the Transit card-anchor probe by observed poll coverage instead of a fixed 20-second window (asturksever, #1282).
+- Keep visible-page Transit probes focused through the browser testing API (asturksever, #1282).
+- Add a real-browser regression gate for hidden-tab periodic work, explicit refresh, resume and teardown (asturksever, #1282).
+- Restore minimized test windows before screenshots so native compositor pauses cannot stall browser acceptance (asturksever, #1282).

changelog.d/1288-zoomfor-legacy-shape.md

@@ -0,0 +1,4 @@
+- A Street Level provider whose `zoomFor` still answers `{zoom, kind}`, the
+  shape before the core went slim, draws its lines again instead of a silent
+  blank; an overview level or any answer that is not a whole zoom shows the
+  zoom-in hint (asturksever, #1288).

changelog.d/1303-wind-stale.md

@@ -0,0 +1,2 @@
+- `get_wind` flags stale forecasts (`stale`) when the wind proxy serves a
+  cached model run (quanticsoul4772, #1303).

changelog.d/1326-terrain-cache-eviction.md

@@ -0,0 +1,3 @@
+- Evict terrain-height cache entries in fetch-time order without repeatedly
+  sorting the full cache, while keeping refreshed points newest and ordering
+  restored entries by their fetch time (daikaginza, #1326).

changelog.d/1340-search-separated-hemisphere.md

@@ -0,0 +1,3 @@
+- Read a hemisphere letter that is separated from its number, so a search for
+  `40.7128 N 74.0060 W` flies there instead of falling through to the geocoders
+  (Rudy Celekli, @rudycelekli, #1340).

changelog.d/1349-cyclone-classification-codes.md

@@ -0,0 +1,3 @@
+- Label NHC cyclone classifications by their published meanings: `PTC` is a
+  post-tropical cyclone, `PC` a potential tropical cyclone, and subtropical
+  systems arrive as `STS` and `STD` (rudycelekli, #1349).

changelog.d/1350-cyclone-movement-knots.md

@@ -0,0 +1,3 @@
+- Convert NHC cyclone movement speed from miles per hour to knots, so
+  `movement.speedKt` in cyclone query results is no longer about 15% high
+  (rudycelekli, #1350).

changelog.d/1351-ais-static-keeps-motion.md

@@ -0,0 +1,6 @@
+- Keep a tracked vessel's speed, course, heading and fix time when AISStream
+  static data arrives; the static message only merges name, type, destination
+  and IMO (rudycelekli, #1351).
+- Apply changed static metadata immediately, retaining known fields when a
+  later static report omits them, without refreshing the vessel's position
+  (samehkhamis, #1351).

changelog.d/1352-cockpit-destination-heading.md

@@ -0,0 +1,3 @@
+- Read a destination heading that rounds to north as 000. 359.6° was shown as
+  360° on the destination bearing while the compass tape already uses 000
+  (Rudy Celekli, @rudycelekli, #1352).

changelog.d/1353-transit-ninety-second-age.md

@@ -0,0 +1,3 @@
+- Count a transit report that is 90 seconds old as one minute. The card kept
+  seconds through 89s, then rounded 90s up to "Reported 2 min ago"
+  (Rudy Celekli, @rudycelekli, #1353).

changelog.d/1354-firms-frp-ten-megawatts.md

@@ -0,0 +1,3 @@
+- Drop the decimal when fire radiative power rounds to 10 MW. A 9.96 MW
+  detection was labeled 10.0 MW because the one-decimal cutoff ran before
+  rounding (Rudy Celekli, @rudycelekli, #1354).

changelog.d/1366-aircraft-models-no-environment-map.md

@@ -0,0 +1,5 @@
+- Aircraft 3D models (Flights, Military Flights, Local ADS-B) skip Cesium's
+  dynamic environment map, a sky render and GPU readback per model. With
+  ~150 models in the scene at Atlanta on Google 3D, GPU time per frame at
+  rest falls from 14.3-14.8 ms to 12.1-12.5 ms; a same-model comparison
+  showed no visible change (quanticsoul4772, #1366).

changelog.d/1367-turned-off-layer-memory.md

@@ -0,0 +1,4 @@
+- Turning a layer off frees the memory its entities held. Cesium kept
+  removed layers' draw commands in reused per-pass arrays, so Data Centers
+  kept 116 MiB after it was switched off, submarine cables 74 MiB and dams
+  30 MiB; they now keep 13, 9 and 9 MiB (quanticsoul4772, #1367).

changelog.d/1378-touch-zoom-aircraft-updates.md

@@ -0,0 +1,5 @@
+- Increase touch-first globe zoom sensitivity to three times the default while
+  retaining Cesium collision and distance limits.
+- Reuse one delayed display timestamp per civil-aircraft fleet update and cache
+  extrapolation coordinate frames until their fix coordinates change. Preserve
+  contact counts, movement cadence, tracked motion and ground corrections.

changelog.d/1379-aircraft-label-candidates.md

@@ -0,0 +1,4 @@
+- Reduce repeated work in aircraft detection candidates by caching unchanged
+  callsign/registration text and reusing resolved fleet models. Metadata updates,
+  live positions and tracked-aircraft/model handoffs remain immediate
+  (samehkhamis, #1379).

changelog.d/1384-hud-summary-moving.md

@@ -0,0 +1,4 @@
+- The HUD's AI place summary drops a reply that arrives after the camera has
+  started moving, so a summary asked for before a flight no longer names the
+  old place mid-flight. Cancelled or partially typed summaries retry even if the place
+  context stays the same (quanticsoul4772, #1384).

changelog.d/1385-route-duration-minute.md

@@ -0,0 +1,3 @@
+- Show a route leg of 59.5 seconds or more as 1 min. The route summary and step
+  cards kept seconds below 60, then rounded 59.6 s up to "60 s"
+  (Rudy Celekli, @rudycelekli, #1385).

changelog.d/1386-route-distance-kilometre.md

@@ -0,0 +1,3 @@
+- Show a route leg of 995 m or more as 1.0 km. The route summary and step
+  cards kept meters below 1000, then rounded 996 m up to "1000 m"
+  (Rudy Celekli, @rudycelekli, #1386).

changelog.d/1387-awareness-distance-kilometre.md

@@ -0,0 +1,3 @@
+- Show a Contacts entry 999.5 m or more away as 1.0 km. The panel kept meters
+  below 1000, then rounded 999.6 m up to "1000 m"
+  (Rudy Celekli, @rudycelekli, #1387).

changelog.d/1388-draw-line-kilometre.md

@@ -0,0 +1,3 @@
+- Show a drawn line of 999.5 m or more as 1.0 km. The draw hint kept metres
+  below 1000, then rounded 999.6 m up to "1000 m"
+  (Rudy Celekli, @rudycelekli, #1388).

changelog.d/1389-cockpit-wind-north.md

@@ -0,0 +1,3 @@
+- Show cockpit wind from 359.5° or more as 0°, not 360°. Rounding 359.6°
+  printed "N · 360°" on the local wind direction
+  (Rudy Celekli, @rudycelekli, #1389).