Brings in 190 commits from main. Integration on the engine contract:
- News and Shipwrecks draw, pick and keep their selection through the
contract and shared map markers, so they follow engine switches.
- Aircraft tracking uses main's segmented trails with contract frame
numbers.
- Decoded weather tiles move into the Cesium adapter and are injected
through the renderer registry; Navara hides those sources on globe
surfaces (docs/KNOWN-ISSUES.md).
- Street-vehicle GPU depth occlusion stays Cesium-only; Navara keeps
the geometric visibility checks (docs/KNOWN-ISSUES.md).
- Transit includes the heading-neutral atlas icons.
A weather feed may also be a frame source. With an optional
getFrameImage({ product, time, size, bbox, signal }) the layers take
whole-extent images, detail windows and the global mosaic from it; with an
optional getFrameTile({ product, time, z, x, y, size, signal }) regional
tiles and their prefetch come from it, null being a transparent tile. Pixels
may be an ImageBitmap, ImageData, canvas or VideoFrame of exactly the
requested size; the layer draws them once and closes them. Cancellation
reaches the source, and a feed without these methods reads the image and
tile routes as before.
AIS Vessels are keyed by the id their live source gives, with the MMSI as an
attribute: an MMSI can be shared, spoofed or reused, so two vessels with one
MMSI are now two vessels. Looking a vessel up or selecting it by MMSI still
finds it.
A share link made while a vessel is selected opens that vessel. A UUID vessel
id travels as 16 bytes in a new thing field (13). A shared thing that is not
reporting now says so in a sentence per kind, with its id (ICAO, NORAD, MMSI
or vessel UUID) on a second line.
A live source may now declare a timeline: it can answer a snapshot at a
given time (source.timeline = { coverage(), stepMs, subscribe? }, read by
readSnapshotTimeline). While such a layer is on it follows the world clock
and registers the coverage. At the live edge it polls exactly as before;
away from it the layer asks for getSnapshot({ ...query, atMs }) at the step
at or before the clock time and shows that snapshot instead of polling.
src/time/snapshotTimeline.js owns the following: a seek, time going
backwards or leaving live replaces every contact with the snapshot at the
new time (play waits on it, for at most 8 s); playing on keeps one
snapshot ahead and gates play on it, applying those snapshots as polls so
contacts move smoothly; outside coverage the layer is empty; back at live
it is cleared and polls at once. One load at a time, and a load the
source drops for a newer one is not a failure.
While following, aircraft are drawn at the clock time with no render
delay, records and absences are judged at the clock, track requests carry
atMs, and vessels move between fixes along the great circle. The world
clock counts jumps (seeks and returns to live) in getState().
A source with no timeline never follows the clock, and every layer stays
live, unchanged.
The radio and push-to-talk helpers shared by both voice providers now pass
English catalog text, and the OpenAI controller words its detail when shown,
like the Gemini session. A token failure without a server message carries
its status as sourceParams, so it is worded again with the number. The docs
say sourceDetail is the message as given, catalog text when it is the app's
own.
The Gemini session translated each status and error message once, when it
happened, so after a language change the voice control kept showing it in
the old language. The state event now also carries the English catalog text
as sourceDetail, and the control words it when shown. The local server's
known token errors stay catalog text until then instead of being translated
on arrival.
The Gemini token route now answers a missing key with code API_KEY_MISSING,
the session carries it to the error state as missingKey, and the voice
control hides "Check microphone permission" for that error, as it already
does for OpenAI. The OpenAI controller sends the same flag, so the shared
control and the controller agree whichever updates last.
The voice error was matched against the translated key message to decide
whether to show the microphone hint, so switching language brought the hint
back and left the message in the old language. The error now carries a
missing-key flag into the status, which words the message at render time.
A controller test covers the language switch, and CURRENT-STATE.md documents
the keyless voice response.
The decision no longer reports not-entity-context or channel-closed, which
its only caller rules out first, nor pointer-scale-not-useful, which the
crop path already checks and logs. A closed channel returns before any
record, as on main. The crop decision is pointer-crop-eligible, since the
crop checks after it decide whether the pointer is on ground, and a sender
called on a closed channel logs channel-lost.
In a dense downtown view from about 400 m up, the occlusion margin of
max(3 m, 2% of the range) came to 14 to 20 m, so a wall, awning, tree or
lower facade standing 10 to 20 m in front of a street was read as the
road itself: the vehicle's dot was hidden by the scene, but its bracket and
callout were still drawn. Checked against scene.pickPosition in one such
view, 18 of 470 vehicles in frame fell in that band.
The margin now depends on how steeply the camera looks down on the
contact. A point on the line of sight s metres in front of a vehicle
stands s * sin(elevation) above it, so a surface there hides the vehicle
once that rise passes 2 m. The margin stays at least 3 m and at most 2%
of the range, so grazing and street-level views are unchanged. In the
same views no vehicle with the drawn surface within 3 m in front of it
(its own road) is newly hidden.
From the review of the first change:
- A contact back in view after longer than a retest interval is held back
and retested like a new one, instead of being drawn from an answer
given before it left the view. It gets a fresh record, so a test still
in flight from before cannot end the hold-back either.
- The hold-back for a contact's first answer stretches with the retest
pace when a crowded view slows it.
- A batch whose control point failed no longer counts as a first answer.
- An unknown log-depth flag reads as "no depth" rather than as standard
depth, which would have put every surface in front of every contact.
- A shader that fails to build leaves no shader behind and is not built
again on the same context.
The detection overlay only hid a contact behind the Earth's curve, so a
street vehicle behind or under a building, a 3D tile or a hill still had
its bracket and "VEH-nnnn" callout painted over that geometry, while its
own dot was depth-tested and correctly hidden.
A contact a layer marks `depthTested: true` (only Street Traffic does) is
now tested against the depth the scene drew and left out of the overlay
while a drawn surface stands in front of it by more than max(3 m, 2% of
its range). Other layers are untouched: their markers are drawn over the
scene on purpose.
The test runs on the GPU and never waits for it. Each tested contact is
one point drawn inside an occlusion query; the vertex shader looks up the
frame's packed pick depth at the contact's pixel, decodes it and drops the
point when the surface is nearer than the contact's threshold. Query
results are collected on later frames. Reading depth back was measured
and rejected: `scene.pickPosition` per contact costs 0.2 to 1.4 ms each,
and Chromium answers `getBufferSubData` with a round trip to its GPU
process even behind a signalled sync object (1.5 ms on a quiet machine,
15 to 25 ms on a busy one).
- Contacts are retested every 100 to 200 ms, spread across frames, at
most 256 a frame; with more in view the pace stretches so none starves.
- Hiding is immediate. A hidden contact returns 600 ms after its last
"hidden" answer, so one on a silhouette edge does not blink, and a
contact new to the view is held back until its first answer (400 ms at
most), so orbiting does not flash brackets behind buildings.
- A callout whose contact goes behind something is dropped at once, not
carried to the next solve and not faded out on the facade.
- These times run on painted frames, so a hidden tab does not expire them.
- A control point in every batch must draw; if it does not, the batch is
ignored and after three the probe stands down for ten seconds. With no
pick depth (2D, Columbus view, depth picking off) nothing is hidden.
A WebGL 1 context reads depth back synchronously about 8 times a second.
The renderer internals this depends on are read in one function,
`readSceneDepthSource`, with the engine version they were checked against.
- layerStateTokenLedger.test: the retired-token reuse case now offers the
retired token (02) to the newcomer, not News's 00.
- he.json: the catalog's singular imperative, {count} first in the plural,
Wikidata read as one source, and smoother legend and info text.
- layerManifest: News after Shipwrecks, so the prompt-budget comment stays
next to the entries it explains.
- CURRENT-STATE: Prettier-formatted (table column, blank line before a heading).
Main added the News layer at the same places: both layers are kept in the
layer names, panel order (News stays last of Events), standalone sources,
voice manifest, credits, DATA_SOURCES and the CURRENT-STATE table. News took
share-link token 00 first, so Shipwrecks moves to 01 (ledger, registry,
CURRENT-STATE and the token tests). The catalog now has 40 layers and the
share-link registry 38. Shipwreck strings are added to the new Hebrew
catalog; the other catalogs keep both sides' strings.
Removing one source's wrecks on the page split merged records: a wreck both
sources list is one Wikidata record carrying EMODnet facts, so expiring
Wikidata dropped wrecks EMODnet still had, and expiring EMODnet kept its
facts. The page now keeps every shown wreck, marked stale, until the newest
source passes the limit, then clears them all. Adds a source test for the
reported ages and limit.
The page compared the proxy's fetchedAt with its own clock, so a skewed
browser clock cleared or kept wrecks wrongly; a snapshot without a timestamp
never expired; and one old source cleared the other's fresh wrecks. The proxy
now reports each source's age on its own clock and its seven-day limit; the
page turns them into per-source deadlines when the snapshot arrives (seven
days from arrival when no age is given) and, on a failed refresh, removes only
the sources past theirs. A refused short refresh is logged.
CURRENT-STATE said each source is asked at most once a day; a failed source
is retried after 15 minutes. It now also covers refused short refreshes and
the page clearing expired wrecks, and its two shipwreck paragraphs are
rewrapped. The card's comment no longer says the records carry no names.
- The legend counted each era twice, scanning every record eight times per
row refresh; it now counts once.
- Each year picker listed every year from the oldest dated wreck (well over
a thousand entries); it lists the years that have a wreck, plus the chosen
ends.
- Labels fell back to Portuguese and links to Portuguese Wikipedia while the
app ships 11 locales: labels now fall back through the app's languages
after English, and a wreck with no English article links its Wikidata
item, which lists the article in every language.
DATA_SOURCES and CURRENT-STATE said text naming people is dropped. The filter
drops text with a title or role word (Mr, captain, owner, crew, reported by);
a bare name with none of them still passes. The docs and the code comment
now say so.
Air accidents (#1584) is not merged, so comments, the changelog fragment and
CURRENT-STATE no longer explain the shipwrecks markers, card and picker by
it; the row-range test fixture is the shipwrecks row. CURRENT-STATE now says
the markers are one billboard collection and the period is stored swapped.
A refresh with one source down (or Wikidata answering 200 with no results)
was cached as the day's snapshot, replacing the full set; and once the
snapshot was a day old, every request after a failure fetched both sources
again. Each source now keeps its own last good records, served as stale for
up to seven days, and a source that failed is not asked again for 15 minutes,
even when there is no data yet. A Wikidata answer without results, or with no
wreck in it, counts as a failure.
When the whole feed backs off, no poll counts a miss, so the tracked
aircraft read STALE at full brightness while no newer position came in.
It now takes the same stale fade as a missed update, in the flights and
military layers.
Local kept World News API's rows listed after the server had deleted them.
README names News among the routes refused behind a proxy; a timed-out
request is said to count; the cockpit cached state is near the same place.
Co-Authored-By: Claude Fable 5.2 <[email protected]>
Keep World News API results an hour at most and say caching needs its
permission; refund a read that never reached it; match Local by any pin;
assign a disputed area by a point inside it; untitled GDELT stories are
minor; hotspot counts leave out minor reports.
Co-Authored-By: Claude Fable 5.2 <[email protected]>
The route check allowed 200 km off the great circle between the airports, so a
coast-to-coast flight routed a few hundred kilometres around weather lost its
origin and destination (JFK–SFO over Kansas, 384 km south). The tolerance is
now 200 km or 15% of the route's length, whichever is larger: 623 km for
JFK–SFO, and still 200 km for a short hop, so a wrong leg stays hidden.
The trail no longer joins a day of history into one line. It lifts the pen
across a gap of more than 48 minutes, an implied speed over 1,500 kt, or a
report without altitude (a parked aircraft within 0.1 nm for under 3 h stays
joined), and every history point is kept as its own segment. History never
gets an invented altitude. The live end is the path the aircraft was drawn
along, thinned to its shape, so it no longer cuts corners through turns; a
late report trims what was drawn after it. Old segments are frozen in chunks,
so a 2,000-point trail costs about 0.1 ms a frame.
A tracked aircraft that misses an update shows STALE and is dimmed. It keeps
flying on its last speed and heading for five minutes (at most 1,500 kt),
then stops, dimmed, labelled SIGNAL LOST with the time since its last report,
and is kept for ten minutes after that report so the camera keeps its target.
Aircraft on or near known ground hold at their last report instead of
coasting. Military flights follow the same rules.
- RECORD is «רשומה», a record rather than a recording.
- Time abbreviations end in the geresh again (דק׳, שנ׳, שע׳, ימ׳), and the
counted forms agree with 1.
- Numbers after a Latin name, ordinals and units keep their order in RTL.
- Meaning, agreement and domain terms fixed across the catalog: cyclone
advisories and storm centre, sensor swath, track-up, cinematic, tropical
depression, pilot and sailing vessels, chips, metered billing, radio static,
airspace, NOT AN ALL-CLEAR, and the remaining review items.
- LOOK and OFF-NADIR are translated; {value} KM stays Latin as an instrument
readout, like KM/S.
- rtl.css shares the Arabic and Hebrew letter-spacing reset and drops the
duplicated Arabic rules; he.md and its glossary follow the catalog.
The catalog source 'NASA FIRMS · LIVE' still led the row; getStats() now
names the source 'NASA FIRMS' for such a snapshot. get_active_fires says
"in the 24 hours to" its end and reports window_end. The end is checked
to be a printable time, cleared under KEY REQUIRED, the test moves next
to the layer, and the changelog and CURRENT-STATE describe it.
scripts/qa/browser.mjs launches headless Chrome with QA_ANGLE and the
shared Chrome lookup; scripts/qa/ready.mjs adds frame, tile and transition
readiness and bounded polling; scripts/qa/geometry.mjs hands page
evaluations the engine-neutral geometry without dev-only source URLs.
CODE-BOUNDARIES gains the rule for new QA scripts.
The orbital view, sensor zoom, ground focus and satellite switching run on
the contract: camera.setFovy sets the vertical lens on both engines and the
fake, including sub-degree angles on Navara through the guarded core
bridge, with a contract check and mutants. Cesium's imagery look-ahead
stays inside its adapter. KNOWN-ISSUES keeps the two parts Navara still
lacks: imagery prefetch for the orbital view and finest-detail meshes for
the ship bridge waterline.
The Bhote Koshi event flies with engine.camera instead of the viewer's
camera, tools listens to camera.followChanged instead of the viewer's
tracked entity, and the celestial ring uses the engine's host, frames,
pose and render requests. The embedded story media no longer imports
Cesium, which removes one of the three import exceptions.
src/i18n/, src/geo/ and the engine-neutral part of src/globe/ are owned by
directory entries, and aggregate groups use a directory wherever their
list covered it, so a new shared module no longer needs a manifest edit.
Small groups keep explicit lists, now checked as a snapshot: the check
fails on a reached module that is not declared and on a declared one no
longer reached, and prints both per group. --write keeps directory
entries as they are.
The manifest goes from 4,737 entries to 3,494; src/i18n accounts for 87
of them instead of 666.
The probe exposes settled() and the first-frame outcome; scripts/qa/ready.mjs
waits for drawn frames with the camera at rest and for a pick at the click
position, with a deadline and its reason in the error. The strict scenarios
use these instead of fixed pauses and loading-screen polling, and the
hidden-switch scenario checks how the loading screen was lifted instead of
how long it took.
Each switch scenario now records a draw ledger per engine and runs a
contract monitor; both write to the scenario's evidence file only, with an
empty allowlist until browser runs show what differs on purpose.
Meshes now go through the same overlay as sprites and lines, so their
colours are the ones asked for without undoing the view's exposure. The
contract suite no longer carries the mesh colour check as a known
failure on Navara.
Name routeId among the canonical analyst IDs in CURRENT-STATE, say what the
changelog entry changes for users, and drop the comment that repeated the
canonical-field doc.
The contract suite is split into one module per area and grows from 87
to 114 checks, each named after the member it exercises: a gesture during
a flight does not cut it, a render request inside a frame draws the next
one, a pure rotation raises changed, ground corridors are as wide in CSS
pixels at device scale 2, imagery replans on resize, drapes draw only on
their surface, sampleTerrainMostDetailed keeps order and passes on a
rejection, terrainRevision changes on a new terrain and on fallback,
follows raise followChanged on retarget, and a mesh, a ground fill and an
imagery pixel draw the colour asked for. In the browser Cesium passes all
113 it runs at device scale 1 and 2; Navara passes 110 of 111 and reports
its single-colour mesh, drawn before tone mapping, as a known failure.
scripts/contract-mutants.test.mjs breaks one promise of the fake engine
at a time, 30 mutants in all, and names the check that catches each; the
list of contract members the suite never calls is down to five.
maps.errors reports tile and terrain errors where an engine has them
(the tileErrors capability: Cesium yes, Navara no), and the map
controller falls back to another map only where it does; movers.list()
rows tell whether a set is shown, as the other families do; FollowHandle
.name and PickResult.primitive are marked deprecated. qa-contract-suite
takes --dpr and the ANGLE backend from QA_ANGLE.