Navara lights the map tiles with the real sun and sky, and only the
landscape maps took the night ambient fill, so an OSM street map turned
black at local night where Cesium keeps it lit. Every base map now takes
the night fill; the landscape maps keep their orbital fill as well.
Daylight is unchanged.
This is why the strict acceptance's Navara pixel and first-frame checks
over Austin failed in evening runs: navara-hd, parked-release,
hidden-switch, context-loss, geo-import and scope saw a black frame, and
the soak and failure matrix waited for a first frame that never came.
At night with the fix all nine pass.
Earthquakes, air quality, GNSS Integrity, fire perimeters, Prohibited &
Restricted Airspace, Active Fires, S2 GhostMaps, cyclone advisories and
submarine cables each have one render/display.js on the globe contract;
their render/cesium* and render/navara* files go. Their discs, cells,
polygons and corridors are engine.ground sets, so on Navara they come from
the one ground screen pass with exact colours instead of nine hand-written
classification passes and lit SDK vector layers. Cesium's look stays the
reference:
- Earthquake and air-quality discs have no outline (Cesium drew none on
clamped ellipses); earthquake discs can now be picked on Navara too.
- Perimeters, drone zones and GhostMaps polygons outline their outer
ring only, as Cesium did.
- GNSS cells stay out of the pick pass; cyclone cones, tracks and centres
are culled by a port of Cesium's horizon occluder.
- Submarine cables are one ground corridor set, as Cesium batched them;
landing points, tips and stems are movers and lines.
- FIRMS glows are a sprite set and its cells a ground set. The far-zoom
heat raster, which no primitive draws yet, sits behind one seam,
render/heatField.js, with a Cesium and a Navara backend.
The shared Navara ground helpers that only these layers used go
(navaraGroundCells, the cell and ground geometry, the zone renderers), as
does the ground prototype script.
Coupling: layer renderer lines fall from 1636/6953 to 1011/4832
(Cesium/Navara), Navara ground passes from 7 to 5, Cesium-importing files
from 65 to 53, engine.extensions reads from 26 to 19.
engine.ground draws sets of polygons (with holes), discs, cells and
corridors draped on the surface, with fill and outline colours, order and
picks, written in bulk and committed. Every engine implements it:
- Cesium batches each set into one GroundPrimitive for fills and one
GroundPolylinePrimitive for outlines and pixel corridors, classified
against terrain and 3D tiles; sets that are not pickable stay out of the
pick pass.
- Navara draws them in one screen pass after tone mapping that finds each
pixel's ground from the scene depth and tests the items in its screen
cell; long rings are split into edge bins. Measured at Dubai against
SDK vector layers, the pass matched Cesium's frame time up to 10,000
polygons with a few megabytes instead of 70 to 146, drew exact colours
and outlines, and hid reliably across an engine switch, where vector
layers hidden with update({ show: false }) never hid. Dense corridors
remain its limit (10,000 cost 135 ms a frame).
- The fake engine keeps the same registry and CPU picks.
The contract suite checks it on both engines (Cesium 66/66, Navara 69/69).
Earthquakes and fire perimeters have unwired ground prototypes whose
drawing matches today's Cesium one (0.998 on Cesium, 0.976 on Navara).
docs/GLOBE-PRIMITIVES.md describes the primitives and when to use each.
- Radio: stations, clusters, the selection and its brackets are point
and sprite sets; clusters come from the layer's engine-neutral screen
clusterer with Cesium's EntityCluster rules, and clicks go through the
selection dispatcher. One style module holds the sizes and colours.
- Transit: vehicles are two sprite sets from an up-front atlas, rotated
with Cesium's projected-rotation maths. The selected vehicle's trail is
engine.lines in two lanes and 29 age bands, so trails are no longer
draped over terrain until the contract has ground corridors. The
Cesium contact trail renderer and its re-export go.
- CCTV: markers, wedges, the plane outline and the calibration gizmo
(rings, arrows, handles, drags through engine.input) draw on the
contract. The textured monitor plane and the viewshed volume, which no
primitive expresses yet, sit behind one seam, render/meshes.js, with a
Cesium backend and a Navara backend the Navara mode passes in.
Known gaps against Cesium: transit trails are not draped, and the
surveillance-lens click correction runs only where the engine reports
its style stage (Navara).
Coupling: layer renderer lines fall from 2638/8735 to 1636/6953
(Cesium/Navara), ScreenSpaceEventHandler files from 15 to 10, `viewer`
lines from 2014 to 1901.
Each layer now has one render/display.js on the globe contract, the same
factory on both engines; its render/cesium* and render/navara* files go.
Cesium's behaviour is the reference, and the logic that was written twice
is written once:
- AIS vessels: atlas chevrons and engine.lines trails. The display
registers the gev-trail pick claim itself and its trailing visibility
pass now also runs on Navara.
- Satellites: point banks per look, orbit rings as lines turned by their
model matrix, and selection and follow through engine.camera.follow.
The emulated _viewer is gone, and the tracked readout republishes card
text on Navara too.
- Space Missions: dots, reticle, trajectories and tactical orbits as
movers and lines. The layer keeps its own records instead of Entities
and DataSources, and camera work goes through engine.camera. Clicks go
through the selection dispatcher, and hover keeps Cesium's throttle. The
pad zone is its rim only: the contract has no surface fill yet.
- ALPR, bikeshare and OSM Places: one display and, for bikeshare, one
style module each; clicks through engine.input, no
ScreenSpaceEventHandler or EntityCluster. ALPR keeps Navara's horizon
cull, and OSM Places points sample the surface where Cesium clamped them.
Coupling: layer renderer lines fall from 4546/11106 to 2638/8735
(Cesium/Navara), `viewer` lines from 2209 to 2014, ScreenSpaceEventHandler
files from 20 to 15.
In the app the world overlay draws on a canvas of its own above the
globe, so letting every canvas take the pointer sent the suite's clicks
there. Only the engine's canvas takes it now. Without makeCanvas the
pick check's sprite atlas was a plain object no real engine draws, so its
sprites could not be picked; the page now hands the suite real 2D
canvases. --only=prefix,... runs a subset of the checks. With both, Cesium
passes 63 of 63 checks and Navara 66 of 66.
The layers that still had one Cesium and one Navara implementation, or
drew on Cesium only, now draw through engine.movers, engine.lines and
engine.camera on both engines, and application code holds the engine
instead of guessing it from a viewer.
- Aircraft family: flights, military and local ADS-B share one display
(src/layers/aircraft/render/display.js). The Cesium and Navara
renderers are gone. Where they differed, the display follows Cesium:
- attach and visibility tracking, live disableDepthTestDistance
changes, the tracked glyph's distance scale, model ids that change,
local ADS-B markers hidden until shown and its UAT ring as a point;
- the trail head's hidden alpha is 0.45, the contract's
depthFailAlpha scaled by the head's 0.9.
The tracked glyph stays over the fleet, as on Navara. The movers
contract reports readiness per model and reads model axes as the
glTF's own on every engine. Cesium's follow takes any native target
through a proxy entity in the ENU frame and mirrors its readout fields.
- Wildlife Sightings draws badges and the density glow as sprite sets
and its privacy cells as line outlines (no fill yet). It takes clicks
from engine.input and its card anchors to the engine's canvas and
camera, so the layer is in both engines' catalogs again and a switch
keeps the session's layer set.
- Cockpit: cockpitHost takes the engine only. The ridden aircraft is the
engine follow's target, hidden through its sprite; only the Cesium
Viewer's stock double-click is still routed around.
- About forty camera, navigation, share-link, HUD, voice, detection and
layer-lifecycle callers take the engine. isGlobeEngine and
engineFromViewer calls outside src/globe fall from 51 to 14.
The coupling baseline drops accordingly: layer renderer lines from
5239/12331 to 4546/11106 (Cesium/Navara), `viewer` lines from 2424 to
2209, Cesium-importing files from 69 to 66.
Layers are to draw through the globe contract on both engines, so Cesium
gets back the movers and lines removed as unused:
- engine.movers draws sprite sets as a BillboardCollection (atlas cells
cropped once per set), point sets as a PointPrimitiveCollection and
model sets as one Model per instance. depthTest, disableWithinM, order,
scaleByDistance, colorBlend, unlit and minimumPixelSize follow the
contract; a model's matrix is written in place, so instances no longer
share one scratch matrix.
- engine.lines draws each line as a polyline Primitive with width in CSS
px, geodesic arcs, dashes, the tactical and arrow materials,
depthFailAlpha and modelMatrix; a rebuild keeps its show, matrix,
colour and place in order.
- The picker returns the pickId a layer wrote last, skips sets that are
not pickable and instances behind the horizon, and falls through to
what lies under them. Surface samples can exclude mover sets again.
movers and lines move from the optional capabilities into the required
contract shape, and the contract suite always runs their checks.
scripts/report-engine-coupling.mjs counts the engine-specific code left
outside src/globe/cesium and src/globe/navara: layer renderer lines per
engine, Cesium imports, `viewer` mentions, engine-name checks, raw Navara
reads, Navara ground passes, Cesium SDK loads and ScreenSpaceEventHandler
users. engineCoupling.test.mjs fails npm test when a count grows past
scripts/engine-coupling-baseline.json, so each step towards drawing
through the globe contract can only lower them.
scripts/qa-contract-suite.mjs runs src/globe/testing/contractSuite.mjs in
headless Chrome on the Cesium and Navara of a keyless dev server, which
until now only ran on the fake engine. The page imports the suite from the
dev server, starts at a fixed view, waits for the camera to rest and lets
the pointer through the panels. Cesium passes 57 of 57 checks, Navara 64
of 67; docs/CURRENT-STATE.md names the three that fail.
Main's camera, voice, Cockpit and layer work now runs on the globe
contract where this branch draws through it:
- The voice tilt to an angle, its pivot about the place under the
crosshair and the frame limit from orbit run through the camera rig;
src/globeInView.js measures a camera pose with src/geo instead of a
Cesium camera. The free-look height limit and the globe-in-view
recovery stay Cesium-only for now.
- Cockpit takes main's own-layer entry rule. Double-clicking the
followed track goes through cockpitHost.routeDoubleClick, so it works
on both engines.
- Space Missions' failed-entry rollback and GhostMaps' age filter run
in the engine-neutral layer code. On Cesium an expired report is
dropped in place; on Navara its product is redrawn.
- Wildlife Sightings joins the catalog Cesium-only: Navara leaves it
out, and the shell builds its card only for the real layer, so a
Navara page no longer fails to start on its stand-in.
- One instruction sentence is shorter, which keeps the Realtime prompt
within its 55,500-character budget with the Sentinel-2 basemap.
createTrafficSource({ roadModes }) lets a host offer a subset of the TomTom,
OSM and Hybrid road modes (all three by default). Without OSM the layer never
draws simulated roads, and a single mode shows no road chip.
While flow is live the Street Traffic row offers Lines (every road with flow
as a steady green/amber/red line, from the heat-line renderer) and Dots (the
moving dots), saved in layer state and share links. Defaults: Lines off,
Dots on.
A throw from the crop capture now has a test that it closes its decision with an error outcome. Direct calls to the crop and image senders pin the channel-closed and pointer-scale-not-useful skips. The docs say a send decision requests an image, and when a pointer snapshot counts as already requested.
Review fixes on top of #1578:
- Exiting Space Missions cancels the layer's own camera flights again (entry,
roster preview, mission focus); only a flight navigation started after
releaseCameraOwnership() runs on.
- A failed first index load says so: "Space Missions unavailable: the
mission index could not load. Try again later." on both the Context and
Data Layers routes, instead of a generic transition toast. Translated.
- Regression tests for both; docs and changelog describe the real feedback.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Third review pass on #835. No runtime change.
- Controller test: viewport decisions reach the controller's debug log.
- Policy test: a second request for a snapshot still being captured is
logged pointer-already-requested and is not captured twice.
- CURRENT-STATE: pointer-not-fresh also covers a missing snapshot,
already-requested covers an earlier attempt that threw, and the image
step runs when any tool output of the response was sent.
- Changelog: describe what is recorded rather than a fixed miscount.
Second review pass on #835's diagnostics. Behaviour is unchanged.
- Report viewport_context.result as soon as the send returns, so a throw
in the retained-frame bookkeeping cannot log a real send as "error".
- List the decision reasons and when each field is present in
CURRENT-STATE, note that the rate limit can split a pair, and say which
codes are defensive.
- Changelog: the decision is logged when an entity-context answer reaches
the image step, not for every answer.
Review fixes on top of #1601:
- Only redirect a provider the params switch on to a keyless source, so a
filter-only setParams (360°, SINCE) no longer moves an explicitly chosen
keyless Mapillary to KartaView.
- Keyless QA gate: assert the new KartaView default and that the Mapillary
chip explains its key instead of switching (it used to expect KEY REQUIRED).
- Tests: Graph-proxy keyReady, filter-only updates; shared test helper.
- README and CURRENT-STATE describe the keyless default.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review follow-up for #835. What is captured, sent and kept is unchanged.
- Log nothing for tool results other than get_entity_context, so other
tools stop spending the shared debug-log budget.
- Give every decision and outcome the same contextId and image kind, so
overlapping requests pair up.
- Log why a crop decision did not send (stale, off ground, off screen,
expired, pose changed, already requested) instead of leaving a lone
"pointer-ground-crop" that reads as sent.
- Split capture-invalidated into session-reset, superseded, channel-lost
and frame-stale.
- Close the pair with "error" when a step throws, then rethrow.
- Document the events in CURRENT-STATE and reword the changelog fragment.
- Keep the open card across a widened reload only when its index still
holds the same report; the server may answer a pinned version with newer
bytes.
- Narrowing to a window the loaded data already covers clears a failed
wider reload, so the row no longer stays in error.
- Label the every-age option "All" (contextual) so the six-option switch
fits the row in every locale; tooltip and info line keep "All markers".
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Review fixes on top of #1528:
- Narrowing the age window only removes markers; a product loaded with a
narrower window is reloaded when the window widens again. Fixes markers
missing after 7d -> 1d -> 7d and avoids reparsing every KMZ on narrowing.
- A wider window over the same edition keeps the open report card.
- Date-only event dates cover their whole UTC day, so 1d still shows
yesterday's reports.
- The age options are one labelled "Event age" radio switch (row
`segmented` control, as FIRMS confidence) instead of toggle chips mixed
with the product chips.
- Un-nest the per-product age test; QA compares data-source counts with the
live defaults instead of a hard-coded 2.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Publish classificationName next to the raw NHC code; unknown codes stay
unexpanded.
- Publish staleCacheFetchedAt only while the proxy serves its cache.
- Touch the shared store on clear only after this layer published, so a
feed failure no longer evicts another layer's dangling selection.
- Drop the unreachable no-storm branch and the unused return value.
- Tests: split the heading fixture into due north and stationary, cover
the unavailable and failure paths with the real store, and check that
clear/disable leave another layer's selection alone; fold the fake-store
test into the real-store one.
Wildlife Sightings (#1591) published share token 9, so News takes the
next free token, 00. The token tests read the ledger before both where
they were written against it, and the README counts both layers.
A slow re-read of a refused key read as unavailable until it landed; it
now stays refused. The first-read wait is 10 s, for a margin under the
browser's 15 s. A failed Conflicts read no longer stops Local's refresh,
which the server reads on its own. Tests cover GDELT's wait, the refused
key and a caller abort past the timeout, and fail rather than hang; the
docs say what a section serves meanwhile.
A country's first World News API read, or Liveuamap's, could take up to
20 s while the browser waits 15, so the panel showed the browser's own
"signal timed out". A read with nothing to serve yet is now waited for
at most 12 s, as GDELT's already was: past that the section answers
unavailable while the read goes on, and the next refresh is served from
it. A client timeout reads "News feed unavailable".
Navara now draws at CSS pixels without MSAA, like Cesium; Navara HD keeps the device pixel ratio and MSAA 4. Only one Navara variant is ever kept. A parked engine is released after three minutes and its WebGL context is lost. Three leaks that kept a released Navara alive are fixed, the canvas context drops its unused antialias buffer, cloud shadows use two 256 px cascades, and the tile cache is budgeted from the screen buffers.
Main's wildlife-sightings layer took share-link token 9, so shipwrecks
takes the next free token, 00. Locale, docs, format-scope and voice
manifest conflicts keep both sides. Ledger and registry tests now model
the ledger before shipwrecks took 00, and layer counts include both
new layers.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Keyless GTFS-Realtime feed published by ZDiTM Szczecin under CC0 1.0.
Speeds are m/s (checked against vehicle movement), the file refreshes about
every 20 s, and the feed carries no route ids, so vehicles take the default
bus mode.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
From 400 km, a tilt to -5° that a zoom out x2.5 caught 70 frames in eased down from the camera's old pitch at the new range. That put the camera below the place's horizon: the first frame's crosshair hit the ground 186 km from London, and London stayed hidden for about 11 frames. The re-solve also left the pivot pitch at 0°, so a zoom in before the next frame computed an infinite clearance floor, was refused, and stayed at 1,000 km.
The re-solved pivot pitch now stops at the place's horizon limit, or where the pivot already was. A camera that the zoom leaves past that limit steps down to it with the zoom and holds it, and the zoom still reports that angle. A zoom in straight after brings the range back to 400 km.
German Khokhlov