WebView 1.1.0 candidate validation

Status: prepared locally; publication paused. The matching candidate manifest records source revision and archive hashes. Existing 1.0.0 assets are unchanged.

Verified source behavior

The final pad/zone refinement completed canonical 3D/main builds, strict TypeScript consumer checks and 713 passing tests, zero failures/skips. The earlier 699-test residency candidate and initial 673-test candidate predate the final geometry fixes. Versioned candidate checks and package inspection must also pass before handoff; their exact outcomes are recorded in the candidate receipt.

Real KERO and approved IRCA review projection checks covered source-colored linked net selection, clear, expansion and immediate saved-state restoration on desktop and a 340 px iframe. IRCA retained 50 modelled components. Earlier stability checks included actual orbit/Pan drags, compact layer/inspection popups, light/dark themes and 390 px embeds. No new warning/error was recorded in the memory pass. This is desktop responsive-browser evidence, not physical phone or installed-client acceptance.

The 100,000-object prepared-source benchmark used three isolated processes per profile and a 512 MiB old-space cap. Its retained-stage reduction was 29.6%; full scope and reproduction. Synthetic repeated-sheet and repeated-assembly stress are distinct from real-project coverage.

Open acceptance items

Preparing archives does not publish, submit to stores, install a client or update the live website. Core API 1.0 and MPL-2.0 remain unchanged.

Additional residency pass (newer than initial candidate archives)

Canonical 3D/main builds and the complete serial suite passed 699 tests, zero failures or skips. New checks cover span quoting, shared KERO mixed-atom AST equivalence, lazy local PCB reads and cancellation, weak camera-history ownership, 2,048 repeated model clones, model-template lifetime, graph admission, GLB accessor expansion and PNG/JPEG dimension admission.

KERO and approved IRCA passed 20 linked net/clear cycles each on desktop and inside a 340 px iframe. IRCA retained all 50 modelled components (395 admitted model nodes, zero model-budget rejections). Full expansion preserved the camera and selection. Narrow light-theme layout had 320 px content width and 320 px scroll width, with no horizontal overflow. This uses a desktop GPU and does not qualify constrained-device limits or physical touch performance.

Desktop console checks were clear; narrow iframe checking reproduced one previously observed unlocalized MutationObserver error without a viewer-control failure. Attribution remains open. The 100,000-object and 4 MiB quoted-payload benchmarks are parser-stage checks, not full browser/GPU memory measurements. Publication remains paused. Initial candidate archives do not contain this newer source and are retained only as prior-source evidence.

Large public-source checks

Pinned Marble (44,059,775 bytes, 1,006 parts) and OLIMEX ESP32-PoE passed online source loading and desktop/340 px iframe layout/3D, full-net/camera, expansion and layer-toggle checks. Model CAD is deliberately absent from these manifests and proxy limitations are disclosed. This pass recorded no console warnings/errors; it does not resolve attribution of the earlier host observer error or qualify physical mobile devices. See detailed measurements and credits. Both publication and installed-client changes remain paused.

The linked Online candidate’s GND stress check reported 48 rejected Marble pad outlines and 47 saved-zone contour omissions; IRCA reported two custom jumper-pad omissions (JP5/JP6). These selected counts are Online findings; the SDK shares the physical pad validator, but its exact selected counts were not measured here. This is a rendering acceptance limitation, not a complete-net geometry pass. See classification and scope. Keep publication paused pending the remaining acceptance work.

Pad/zone follow-up

The latest source passed 713 tests with zero failures/skips and canonical 3D/main builds. Marble’s 139/139 copper contours and IRCA’s two custom jumper pads were checked on desktop and 340 px embeds. Linked selection, expansion, layer toggles and IRCA’s 50 models remain intact. See measured scope and limits. Earlier omission findings above are historical evidence. Online’s final browser comparison is a separate gate. Publication remains paused.