Merge remote-tracking branch 'origin/main' into fix-1056

# Conflicts:
#	src/tools/queries/aviation.test.mjs

changelog.d/1057-windows-npm-dep0190.md

@@ -0,0 +1,4 @@
+- On Windows, `npm run doctor` and the Pinokio install no longer print Node's
+  DEP0190 warning ("Passing args to a child process with shell option true can
+  lead to security vulnerabilities"): npm.cmd's fixed subcommand is now part of
+  the shell command string instead of an argument array (quanticsoul4772, #1057).

changelog.d/1096-military-hold-track-speed.md

@@ -0,0 +1,3 @@
+- Military aircraft keep their last heading and speed when an adsb.lol row omits
+  `track` or `gs`, instead of turning north and stopping; a grounded row without
+  a speed still reads 0 (Aditya, @Rohilalala, #1096).

changelog.d/1097-overpass-remark-failure-phrases.md

@@ -0,0 +1,5 @@
+- Overpass data whose element tags contain a failure phrase ("timed out",
+  "runtime error", "too many requests" and similar, e.g. a place named "Timed
+  Out Escape Rooms") is returned as data instead of being refused and putting
+  the endpoint on cooldown; upstream failures are read from the remark (Aditya,
+  @Rohilalala, #1097).

changelog.d/1115-aircraft-altitude-band-unknown.md

@@ -0,0 +1,2 @@
+- Exclude aircraft without a reported altitude from requested altitude bands;
+  unfiltered aircraft queries still retain them (rudycelekli, #1115).

changelog.d/1117-google-viewer-lifecycle.md

@@ -0,0 +1,3 @@
+- Release superseded Google Street View open waits and suppress metadata
+  requests whose viewer was unmounted while the service loaded (rudycelekli,
+  #1117).

changelog.d/1171-director-goto-shot.md

@@ -0,0 +1,3 @@
+- Director shot actions land on the requested shot. A seek to a shot's start
+  could round a hair below it and select the end of the previous shot for some
+  shot counts and durations (Aditya, @Rohilalala, #1171).

changelog.d/1172-cctv-missing-numbers.md

@@ -0,0 +1,13 @@
+- Keep missing CCTV numbers unknown instead of reading them as zero. `Number`
+  turns null, blank strings and booleans into 0 or 1, and the camera source
+  list sends an unknown number as JSON null, so a source-pack camera that left
+  out its pose reached the browser as zeros clamped to each field's floor —
+  facing north, pitch −2°, FOV 20°, range 120 m, mount 6 m and ground 0 m —
+  instead of the defaults (id-hash heading, −17°, 74°, 700 m, 24 m and the
+  city's ground); a null or blank position placed the camera at 0,0. The
+  server's `toFiniteNumber` and the browser's `safeNumber` now treat absent,
+  blank and boolean values as missing. The same conversion also stops a
+  Tallinn row with a null heading reading as a confident north-facing camera,
+  a Caltrans camera with a blank elevation losing its 150 m ground prior, and
+  a TfL or Catalonia camera with a null or blank coordinate being kept at 0,0
+  or on the zero meridian (omerbberk, #1172).

changelog.d/1173-antimeridian-ground-floor.md

@@ -0,0 +1,10 @@
+- Flights' ground floor works across the antimeridian. The viewer-distance
+  check behind the 150 km floor radius and its copy in the photoreal mesh-floor
+  sampler subtracted longitudes without wrapping, so two points 21 km apart
+  across 180° measured about 38,000 km and low or grounded aircraft just across
+  the line from the viewer were never floored. A contact's display corridor
+  walked a segment across 180° the long way round, spending its cell budget on
+  ground behind the contact, and a contact on the seam lost its cell hysteresis.
+  SEARCH NEARBY SITES on a view across 180° searched Google Places around
+  longitude 0, on the far side of the globe, with a radius sized from the view's
+  height alone; it now searches the view (Aditya, @Rohilalala, #1173).