Asking from the chat composer to open Voice Settings or Voice Preview
closed chat around the focused input, leaving focus on a hidden control.
Focus held in chat now moves to the requested view's tab, as the panel's
own tabs move it; focus anywhere else stays put.
Three 40 ms actions in one round outlast a 50 ms deadline unless each
action re-arms it, so removing that re-arm now fails the test for both
providers.
The conversation is kept to its newest 60 messages, so once full it no
longer grows as a message is sent, and a reader who had scrolled up was
left behind by their own request. The view now spots a send by the newest
user message's id instead of the list's length.
The attach menu promised up to 10 KB per file, but the attachments share
one 10 KB budget, so two 6 KB files were refused. The hint now says the
limit covers them all, in every locale.
With Voice Preview off the voice panel never opens on its own, and chat
Settings hid the voice preferences, so the Preview toggle, inactivity
timeout and model tier could not be reached by pointer. Chat Settings now
show them below the chat options, still in chat's frame and without the
voice tabs.
set_panel_open for voice-settings or voice-preview reported success while
chat's open state kept the voice panel hidden, minimized chat included. A
successful explicit show now closes chat, which keeps the typed session,
its running turn and its messages; reopening chat restores them.
The streamed oversized body was invalid JSON, so the loader returned
nothing whether or not the byte cap applied. Pad a valid catalog with a
usable camera instead, so only the cap can refuse it.
TranStar's Content Usage terms prohibit copying, distribution and
external framing of its images without TxDOT's prior written
permission. The opt-in text only said the pack was off unless
CCTV_HOUSTON_ENABLED=1, without saying what an operator needs before
setting it. Say "enable only if you hold TxDOT's written permission" in
.env.example, the README, DATA_SOURCES.md, CURRENT-STATE and the
changelog, matching the NYC and Florida packs, and fix the column of the
.env.example comment.
The houston-cctv credit now reads "(Houston, when enabled)" and links
the Content Usage terms, like the Florida credit. Where the docs said
frames are "never stored or redistributed", they now say what happens:
frames are relayed on demand, buffered only per request, sent with
no-store and not archived.
TranStar's `dir` field is the side of the freeway a camera sits on, not
the way it faces. TranStar's own map script (translateCameraDirection)
shows it as "Traffic closest to the camera is traveling": South is
eastbound, North westbound, West southbound and anything else
northbound. The live catalog agrees: I-10 cameras are North/South and
I-45 cameras East/West.
Treating `dir` as a high-confidence facing pointed most Houston cameras
across or away from the road they watch, and dropped the "estimated"
marking. Map it to the travel direction instead (S 90, N 270, W 180,
E 0) as a low-confidence prior with the low-confidence pose, and keep
the id-hash fallback for None or unknown values.
npm run changelog consumes changelog.d fragments at release, so a test that reads the fragment would fail with ENOENT afterwards. CURRENT-STATE is permanent.
FAA marks line breaks in TIMESOFUSE with a literal %, which 56 live
restricted-area cards showed verbatim ("SAT - SUN; %OTHER TIMES BY NOTAM").
Read each marker as a '; ' separator, absorbing a semicolon already before
it.
The changelog fragment named share-link token 6 (sea marks); the layer
reserves 8. CURRENT-STATE said voice could toggle and query the layer and
listed it among the analyst record sources, but it is in VOICE_OFF_LAYERS.
Say voice is off for now, and that the layer covers FAA prohibited and
restricted special-use airspace only, not the UAS Facility Map / LAANC grids
or TFRs. A test pins these claims to the token ledger and the voice manifest.
A foreign Origin or Sec-Fetch-Site: cross-site request was answered and
counted against the client's 60-a-minute quota, and could start a refresh.
Run the keyless same-site checks the other map routes use before rate
limiting and cache acquisition; reverse-proxy forwarding headers stay
accepted. List the route with the keyless map routes in SECURITY.md.
The refresh signature left out the polygons, and these rows carry no
timestamp that moves with the geometry, so an amended boundary with the same
attributes kept the old shape and card anchor until the layer was destroyed.
Add a deterministic hash of each zone's rings to the signature.
The source dropped the proxy's stale flag and fetch time, so the layer row
read nominal and freshly updated through an outage, and the proxy served its
last snapshot indefinitely. Carry stale and fetchedAt through the source as
Fire Perimeters does, date a stale snapshot by its fetch, and stop serving a
snapshot more than 24 hours old.
An upstream answer whose zones all fail validation normalized to an empty
list, and paging that stopped at the page limit while the service still
flagged more records was accepted as complete. Both were cached as a fresh
snapshot for an hour and replaced the last good one. Treat them as failed
refreshes so stale-on-error keeps serving the previous snapshot.
A still-tagged feature without exportable geometry already keeps its old
line. A newly tagged one had nowhere to go and was dropped silently; it is
now listed as skipped with a warning, so a refresh never loses a site or
hides a broken one.
The look's sun follows the page clock, so the frame these checks compare was as bright as the hour they ran at: in the acceptance run the half-strength bloom changed the frame by 0.29 against 3.5 earlier, and at dusk the full bloom changes nothing (luma 13).
Measured over the second half the heap still grew about 20 KiB per switch in the acceptance run (117.8 to 123.8 MiB from switch 250 to 500), so the narrower window did not separate warm-up from growth; the original check stays.
Under the soak workload V8's compiled code and the layers' caches grow for about 250 switches and then the heap is flat (121.4, 121.7, 120.4, 119.1 MiB after 250 to 400 switches), so a line fitted from switch 20 measured the warm-up. Over the second half the threshold stays 10 KiB per switch; the run before the layer effect fix measures 19 KiB there.
Layers re-create their ground effects on every attach. Navara's removePass leaves the EffectPass undisposed, so each switch kept its EffectMaterial in three's program and material caches, together with the uniforms' DataTextures and their Float32Array data (earthquakes: about 47 KB of buffers and one WebGL program per switch). engine.LayerEffectDesc now disposes the raw pass on delete, as the app's own effects do.
From the whole-globe view the camera could end up facing empty space with
only stars on screen. Cesium's free look (Shift+drag) turns the camera in
place, and from 18,000 km the planet's edge is only about 15 degrees from
nadir, so a few drags turned it out of frame. The rolled pose was then
saved into the share hash, so a reload or shared link opened on stars too.
- Allow free look only below 1,000 km. Orbit, tilt and zoom are unchanged.
- Measure how much of the frame shows the planet's surface by casting a
32x18 grid of view rays at the WGS84 ellipsoid. While the scope mask is
on (the default) only the rays inside its keyhole count, since the mask
hides or dims the rest of the frame.
- When a camera move ends above 1,000 km with no ray on the surface, turn
straight down at the same position and heading. Limb glow and a sliver
up to about one grid cell thick can count as out of view. The end of a
move comes from Cesium's moveEnd and from a check of its own that
samples the pose every 250 ms after moveStart and acts once it has held
for half a second: at some of these poses the camera drifts by a
rounding error every frame, so Cesium never sees it stop and never
raises moveEnd. The zoom-out guard from #1522 keeps wheel and pinch
zoom-outs in view; this catches the other ways a move can end looking
away, such as a saved or shared view that did, or a
drag that Cesium turns into an in-place look below its 7,500 km
trackball height (one that starts on the stars, or a tilt drag with the
screen centre off the planet). Views with more surface, tracked or
locked cameras and the 2D and Columbus modes are left alone.
- Voice tilt and in-place turns of a free 3D globe camera above 1,000 km
stop when about a tenth of the frame still shows the planet (or about as
much as it showed when it started, two grid cells of sampling noise
allowed), never below about 1% of the visible frame; a view that already
shows less only keeps or gains what it has. A turn that cannot go half a
degree further without crossing that limit is refused with a reason
instead of reported as done, and a turn it accepts turns at least that
half degree. A turn that keeps or gains surface is not refused, even
after the window has been resized, except that near the limit, where the
planet's edge crosses the sample grid, sampling wobble can occasionally
refuse one (most often above about 15,000 km and on phone or ultrawide
frames); nothing then turns away from the planet. A turn that starts
where the last voice turn left the camera keeps that turn's limit, or a
higher one if the planet now fills more, so repeating the request does
not take any more; once anything else has moved or turned the camera, a
turn is measured from what is on screen. A request made while the camera
is tracked is judged by the pose voice's release of the follow camera
will leave, before that release, so a refused turn keeps the tracking.
The frame measure and the look-step limit live in src/globeInView.js,
shared by the viewer and the voice camera verbs.
Asking the canvas again during the switch made the reuse depend on getContext at that moment: with new contexts refused (qa-strict-failure-matrix, cold Navara to Cesium) the rollback to the parked Navara threw and left no globe.
The aircraft scene binding re-tracked with origin share-restore, which stayed on the entity: a later trackedEntityChanged for it (cockpit exit, refocus) never stamped navigation, and the aircraft lost its selection origin. Flights, military and satellites now set a transient gevRestoring mark around follow.start, and the aircraft binding resumes with the origin the aircraft was selected by.
qa-strict-prod-smoke opens the keyless vite preview on each engine, switches once and fails on any page error. qa-strict-cold-chunks reports Cesium.js at a cold Navara start as a known gap instead of a failure.
This reverts ae13a56f. In the production build every 'cesium' import is the global that Cesium.js defines; with the script loaded only when a Cesium globe started, modules evaluated at boot (cesiumConstants, annotationEngine) threw ReferenceError and neither engine started. A cold Navara start loading Cesium.js stays a known gap.
quanticsoul4772