CR_013 classified evidence rows: development verification

Scope and identity

Development increment from 0cecafb4836ab252d9e30e4553ab901cbd8660b7 plus recorded worktree; Windows 11, Python 3.13.17, Node 24.21.0, Firefox 158.0.1 and geckodriver 0.37.1. Round 10: CR_013 classified evidence rows records scoped acceptance and the owner’s fallback choice. STV_CR_013: Vertical target levels for artifacts and test records retains the request/diagram. No release, tag, commit or publication is performed.

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

  • Source/package identities and phase notes (retained in the engineering checkout)

  • Outcome summary (retained in the engineering checkout)

Results

134 distinct tests passed, with no failures/errors/skips: 34 frontend Node tests, 59 full non-browser tests, one additional 30,000-record classification test, two new published-row variants, 36 existing browser variants and two capacity variants. Complementary development selections use unchanged runtime bytes; these are not results from earlier release candidates. The large classification case and import-only test cleanup followed the initial Python/browser suites; its actual additional run is retained separately. Every current runtime Python/JS/CSS member was compared byte-for-byte with the tested development wheel.

  • Frontend report (retained in the engineering checkout)

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

  • 30,000-record classifier report (retained in the engineering checkout)

  • New published-row file/HTTP report (retained in the engineering checkout)

  • Existing Firefox regression report (retained in the engineering checkout)

  • Capacity file/HTTP report (retained in the engineering checkout)

Criterion assessment

The new Sphinx fixture uses renamed host types/tags/relations and three starting views (software, case, execution). Each reaches all nine category rows, multiple cases/executions, a shared cross-level case and three fallback records. Assertions cover same-row top alignment, immutable classification during navigation, anchor/peer stability, independent views, resolved control ordering, visible fallback labels, unchanged world coordinates on Reset and screen-space edge ports. Model/graph cases cover invalid configurations, strict mode, absent or conflicting tags, orphan/mixed/hidden owners, outgoing owner mapping, cycles, record-order determinism and input immutability. The separate 30,000-record test resolves 15,000 cases and executions with their actual owner links.

  • File geometry observations (retained in the engineering checkout)

  • HTTP geometry observations (retained in the engineering checkout)

  • File rendering (retained in the engineering checkout)

  • HTTP rendering (retained in the engineering checkout)

All existing legacy-role, PR013 navigation/arrow/order, picker/paging, hover/focus, missing-target, detail transport and alternate-model browser cases passed. The rebuilt 30,000-Need/90,000-edge fixture runs three repetitions per transport, reaches exactly 200 visible nodes, rejects overflow, loads/caches details and collapses successfully. Maximum measured expansion was 214 ms/file and 204 ms/HTTP; observations do not introduce new performance guarantees. The clean source-free wheel host has no Node on PATH. Source archive checks rebuild the wheel and compare shipped assets; package members match current runtime source. No dependency or version pin changed.

  • Browser capacity measurements (retained in the engineering checkout)

  • Large fixture build metrics (retained in the engineering checkout)

  • Installed consumer identity (retained in the engineering checkout)

  • Consumer environment (retained in the engineering checkout)

  • Source archive inventory (retained in the engineering checkout)

  • Source archive checksums (retained in the engineering checkout)

  • Development package build (retained in the engineering checkout)

Concept and documentation checks

The initial concept build exposed an ADR link outside its toctree. The index was corrected and warnings-as-errors validation passed before acceptance. Standalone and embedded system/software/state diagrams were visually inspected; classification labels refine existing flows without new topology or unreadable routing. Final warnings-as-errors documentation and complete traceability/evidence validation passed after recording the four canonical executions (109 cases and 604 historical/current executions). Strict executable mapping covers all 71 system/software requirements, with no unknown IDs or unmapped active requirement. Unrelated security/hosted gates and later occurrence-mode requirements retain their previous states.

  • Initial concept navigation diagnostic (retained in the engineering checkout)

  • Corrected concept build (retained in the engineering checkout)

  • Concept test-coverage validation (retained in the engineering checkout)

  • Draw.io export log (retained in the engineering checkout)

  • Executable mapping (retained in the engineering checkout)

  • Final documentation build (retained in the engineering checkout)

  • Final traceability check (retained in the engineering checkout)

Canonical executions

Validate classified placeme... (STV_TC_SW_046), Resolve stable classificati... (STV_TC_SW_047), Align classified rows throu... (STV_TC_SW_048) and Published nine-row evidence... (STV_TC_SY_061) each have one actual aggregate execution of the complete procedure. Runner subtests, including relevant legacy regression in the full frontend selection, remain individually identified in the mapping. Report completion timestamps are disclosed; no individual finish times are invented.

  • Canonical procedure/subtest mapping (retained in the engineering checkout)

  • Final requirement trace paths (retained in the engineering checkout)

  • Final test coverage paths (retained in the engineering checkout)

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