PR013 final development verification

Scope and identity

Date: 2026-10-24, Windows 11, Python 3.13.17, Node 24.21.0, Firefox 158.0.1, geckodriver 0.37.1. Input commit 4ad8fcefdf95ee6987478a9d632d5f33624aa673 plus the recorded dirty worktree. The owner approved the PoC and requested concept creation, review and implementation; Round 9: Stable navigation concept and implementation handover records acceptance. This is development verification, not a release candidate. Historical PoC attempts remain in PR013 position-stability proof of concept.

  • Build/package identity (retained in the engineering checkout)

  • Initial source hashes (retained in the engineering checkout)

  • Final source hashes and test-only difference (retained in the engineering checkout)

  • Exact commands and report timestamps (retained in the engineering checkout)

  • Machine-readable summary (retained in the engineering checkout)

Results and interpretation

111 distinct tests have latest passing outcomes: 31 Node component cases and 80 pytest cases (42 non-browser and 38 Firefox, including two capacity variants). There are no latest failures/errors/skips. This combines complementary selections against the same product bytes, not historical release candidates.

The initial adjacent browser run had two failed missing-target text assertions. Stable navigation placed the newly discovered box partly outside the retained viewport. The ID was present in the DOM. The revised accepted procedure asserts identity, explicitly Reset viewport, verifies unchanged world geometry and then requires visible identity plus detail failure/retry recovery. Both retests pass. Only that test function changed after the initial suites; product bytes and the other test functions did not change. Failed attempts are retained below.

All new case designs have actual executions: Stable navigation geometry ... (STV_TC_SW_043), Complete finite SVG bounds (STV_TC_SW_044), Natural ID ordering for fre... (STV_TC_SW_045), Stable keyboard navigation ... (STV_TC_SY_058), Negative route rendering an... (STV_TC_SY_059), Natural seeded order surviv... (STV_TC_SY_060). Geometry assertions cover shared-test reranking, either anchor role, root collapse/reopen, growing earlier bands and 200 nodes. Firefox checks measure world/screen stability, focus, negative x/y route coverage, endpoint alignment, Reset and initial numeric order. The async harness verifies committed snapshot identity survives rejected work. Existing fresh-centering, alternate-model, paging/picker, inspection and independent-view checks passed as complementary regression gates.

  • Frontend component report (retained in the engineering checkout)

  • Full non-browser report (retained in the engineering checkout)

  • Focused Firefox report (retained in the engineering checkout)

  • Adjacent Firefox first attempt, including failures (retained in the engineering checkout)

  • Placeholder procedure retest (retained in the engineering checkout)

  • Capacity report (retained in the engineering checkout)

  • Adjacent failure diagnostics (retained in the engineering checkout)

  • Placeholder retest log (retained in the engineering checkout)

Capacity, packaging and concept checks

The newly rebuilt fixture contains 30,000 Needs / 90,000 authored edges. Three runs per transport reach exactly 200 visible nodes, reject overflow without committing, exercise uncached/cached details and collapse. Measured maxima were 223 ms/file and 230 ms/HTTP for expansions; these observations are not new performance guarantees. The isolated wheel host imports the installed development package with Node absent from PATH. The source archive excludes local evidence and rebuilds the wheel with identical shipped static assets.

  • Fixture metrics (retained in the engineering checkout)

  • Browser measurements (retained in the engineering checkout)

  • Clean installed consumer identity (retained in the engineering checkout)

  • Consumer build log (retained in the engineering checkout)

  • Source archive member inventory (retained in the engineering checkout)

  • Rebuilt archive checksums (retained in the engineering checkout)

  • Package build log (retained in the engineering checkout)

Warnings-as-errors documentation, full test-coverage traceability and strict bidirectional executable mapping passed. The final graph contains 105 canonical cases and 600 historical/current execution records; all 70 system/software requirements have executable mappings. Standalone and embedded system/software/state diagrams were visually inspected. Separate feedback and geometry arrows are readable; link checks verify 2/9/11 portable Need targets. The initial short-heading warnings and symbolic-link export preflight failure are retained with successful rechecks. An initial root-level Node build invocation also failed before the correct frontend invocation; no product test result is attributed to that preflight.

  • Initial concept build warnings (retained in the engineering checkout)

  • Successful concept build (retained in the engineering checkout)

  • Initial export preflight (retained in the engineering checkout)

  • Successful shared export (retained in the engineering checkout)

  • Published diagram link check (retained in the engineering checkout)

  • Executable mapping check (retained in the engineering checkout)

Canonical case execution records

  • Case-result mapping (retained in the engineering checkout)

No tag, GitHub release, deployment or commit is created. Unrelated security review/hosted gates and later occurrence-mode requirements retain their prior states. The complete next release must be explicitly started and qualified.

Final checks

  • Successful final documentation build (retained in the engineering checkout)

  • Initial aggregation diagnostic (retained in the engineering checkout)

  • Successful corrected graph check (retained in the engineering checkout)

  • Final traceability paths (retained in the engineering checkout)

  • Final requirement-to-case paths (retained in the engineering checkout)

The first final graph check found ambiguous latest timestamps when several runner subtests were modeled as separate executions of one canonical procedure. The corrected unissued record has one aggregate execution per complete new case; every actual subtest selector/result remains in the mapping and runner reports. No per-subtest time or result is invented. Failed/retested placeholder attempts remain separate history. No previously issued execution record was changed.