A scenario whose body ran past about 19 s idled until 45 s after its
start, and twice one never exited and hit the runner's 300 s timeout
after it had passed: something of Chrome outlived Browser.close and kept
its stdout and stderr open.
- scripts/qa/browser.mjs closes Chrome within 5 s, kills its process
group and closes its pipes; the harness then finishes and exits with
the scenario's code. QA_EXIT_DIAGNOSTICS=1 keeps the old ending and
prints the open handles and Chrome's process tree 5 s after finish.
- Every strict page starts with its clock at the 18:30 UTC nearest to
the real time, so Navara's sun is up over the test cities whenever the
acceptance runs; a scenario can ask for its own instant or the real
clock (Wildlife does, for its freshness fixtures).
- The runner records a scenario that passed but did not exit as hung,
kills the whole process group on a timeout, runs a failed scenario
once more and reports one that passes then as flaky, and appends each
run to --history=FILE; scripts/report-acceptance-flake.mjs reads it.
Layer display tests no longer re-check Cesium primitives in every layer:
they run on the fake engine, and one Cesium translation check per ground
kind lives in the Cesium engine's ground tests. Their harnesses are shared,
and the fake-engine, Cesium-engine and flights tests that repeated the
contract suite are gone, 340 tests in all.
The press-travel measurement is shared by the selection dispatcher and
CCTV, which loses its ScreenSpaceEventHandler shim and world-click module.
ALPR, installations and bikeshare share one selection keeper and selected
marker write. Layers with their own click rules keep their handlers.
Navara's pulse is a class like the ground pass, the ground and imagery
passes share the surface uniforms, and the colour, log-depth and occlusion
GLSL lives in one chunk module, each caller keeping its depth bias. The
ground suite runs inside the contract suite (--only=ground:), and fixture
and script names say what they test.
Behaviour is unchanged on both engines.
- Navara 0.2.1 never frees pick batch ids, and the mover banks rebuilt
their instanced meshes on each attach, leaking about 260 KiB of
WebAssembly memory per switch. A per-view pool hands released ids back
out; the long soak now asserts native growth of at most 32 KiB per switch.
- Navara 0.2.1 sometimes drops the first show:false update on a new
vector layer, so a zone turned off right after a switch back stayed
drawn. Polygon displays hide by deleting their layers and rebuild them
when shown.
- Camera verbs drive both engines through the globe contract; the Cesium
viewer rig and the tests of states it can no longer reach are gone.
- Flights and military tracking share src/layers/aircraft/tracking.js.
- The geo import scenario reads movers only where the engine has them.
- Add the Navara section to docs/CURRENT-STATE.md and a changelog fragment.
- src/geo drops the ported debug argument checks and the tests that
bitExact already covers; package boundaries accept a directory entry,
so src/geo/ is listed once per group.
- Cesium no longer carries movers and lines nobody draws with: they are
optional engine capabilities. The fake engine and the Cesium camera use
the shared cameraMath; unused contract tables, clock.ticked, the
'finest' surface detail and Navara click statistics are gone.
- ALPR keeps one overlay core and one presentation model for both
engines, bikeshare one station bookkeeping, the Navara ground effects
one definition helper; image atlases, view centres, mover banks and
colours are shared.
- The legacy switch route, the clock offset and rate controls and other
unused exports are removed.
- The acceptance runs the Navara draw, geo import and scope checks;
scenarios use the harness switch and page helpers; the main chunk
check keeps the size budget only and runs once in CI.
The long soak's heap slope counted the per-aircraft dead-reckoning history (POSITION_HISTORY_LIMIT fixes for 10 000 aircraft, about 14 MiB over the first 2.5 min of polls) as switch growth: a run with the same workload and no switches grows as much and then levels off. With --assert the soak now waits until the flights feed has been read more times than the history keeps before it switches; over 500 switches the heap then grows 1.07 KiB per switch.
qa-strict-context-loss records, only when no non-blank frame came, how many frames the running and the watched engine drew in 3 s and how much of the last one differs from its first pixel.
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.
Every scenario runs through scripts/qa-strict-harness.mjs: each page, frame and worker gets an ordered block list before it runs (paid hosts and routes first, then only the preview, declared fixtures, keyless probes and cached free tile hosts), switches with real input on the picker, and records its assertions per layer and engine. The runner reads the catalog from the live page, fails any layer without an assertion on either engine, can build and serve a keyless checkout itself, and adds rapid, hidden-tab, context-loss, failure-injection, cold-chunk and soak scenarios. The main chunk gets a 1% size budget.
German Khokhlov