PR015 detail acceptance corrections: development verification

Input commit 740c35ce234ec26b16c76a0b77463b273f84863c plus the prior dirty CR014 implementation. Correct detail readability,... (STV_PR_015) retains the acceptance findings; Round 13: Detail acceptance corrections records the scoped technical review. No commit or release is performed. Owner acceptance retest remains pending.

Before/after and criterion review

The real Sphinx/Firefox baseline fails in both file and HTTP modes: dock width 360 px, metadata initially shown, no local toggle handler, and a header-ID URL pointing at a namespaced clone in the view page. The after-fix variants pass: 720 px dock, folded initial metadata with pointer/Space/Enter toggling and ARIA state, independent hover/persistent copies, preserved reuse state, and an actual new tab at the source page/Need element. Ordinary copied-body anchors retain local namespace behavior. At 1050 px, the dock is below the graph and bounded to available width/720 px; the 1120 px breakpoint reserves useful graph space.

The shared content controller executes no host/widget script. Existing safe URL/media policies remain. The adjacent alternate-model procedure explicitly opens metadata before asserting its contents; this preserves the underlying availability check under the requested folded default. Collection age, limit, races, hover timing and original-object navigation remain regression constraints.

All 127 distinct selected checks pass: 72 Python unit/host/package checks, 36 Node components and 19 Firefox variants (eight persistent/metadata plus eleven adjacent). Three host builds appear in both Python and focused reports. The fresh wheel/source distribution passes clean installed-consumer and boundary/ rebuild checks. Capacity and full release qualification are not rerun; earlier results are historical. No version or dependency change is made.

  • Failing baseline (retained in the engineering checkout) and diagnostics (retained in the engineering checkout)

  • Persistent/acceptance result (retained in the engineering checkout)

  • Adjacent rich-content/hover/transport/path result (retained in the engineering checkout)

  • Python/package result (retained in the engineering checkout)

  • Frontend component result (retained in the engineering checkout)

  • Source/artifact identity (retained in the engineering checkout)

  • Exact commands/environment (retained in the engineering checkout)

  • Installed consumer identity (retained in the engineering checkout)

  • Source archive inventory (retained in the engineering checkout)

  • Rendered metadata and independent copies (retained in the engineering checkout)

  • Observed width, fold and link values (retained in the engineering checkout)

  • Draw.io export/link checks (retained in the engineering checkout)

Canonical executions

  • Exact result mapping (retained in the engineering checkout)

Each execution is a complete canonical procedure, aggregating real runner variants. Timestamps are report completion times, not invented per-subtest finish times. The before-fix failure is retained separately; passing executions do not accept unrelated draft case designs. Environment: Windows 11, Python 3.13.17, Firefox 158.0.1 and geckodriver 0.37.1; browser requests outside loopback are blocked.

Final rendered and publication review

The new metadata diagram and its normal HTML embedding were visually inspected; labels, both arrow routes and Need links are readable. The actual project demo was inspected with folded and expanded metadata. Its dock measures 720 px and its header ID points at the original stakeholder document/Need, not demo.html. A refined two-protocol repeat adds explicit arrow-icon and unique-ID assertions; both pass. This repeats existing distinct cases without increasing the 127-case selection. Product code is unchanged across the successful suites.

Warnings-as-errors Sphinx and complete shared evidence validation pass. Strict mapping covers 75 system/software requirements and 158 executable cases with no unknown IDs or unmapped active requirement. The evidence graph includes 115 canonical cases and 805 execution records; deferred/unselected cases retain their prior states. These development gates do not qualify a release.

  • Refined icon/ID report (retained in the engineering checkout)

  • Normal demo, metadata folded (retained in the engineering checkout)

  • Normal demo, metadata expanded (retained in the engineering checkout)

  • Normal demo measurements/URLs (retained in the engineering checkout)

  • Initial folded wide fixture (retained in the engineering checkout)

  • Independent narrow boxes (retained in the engineering checkout)

  • Metadata state diagram (retained in the engineering checkout)

  • Normal HTML diagram embedding (retained in the engineering checkout)

  • Normal documentation build (retained in the engineering checkout)

  • Strict executable mapping (retained in the engineering checkout)

  • Shared evidence validation (retained in the engineering checkout)

  • Requirement paths (retained in the engineering checkout)

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

Pane-frame acceptance follow-up

The owner supplied a screenshot of the invisible side-pane boundary and shadowed first-card edge. The open pane now has a 1 CSS-pixel solid #999 frame on all four sides, inside its existing 720 px width. Its top and 976 px outer height align with the graph’s 975 px client area plus top border. Eight-pixel inset padding and shadow-free cards give the first card a clean top edge. Narrow layouts keep the existing 60vh cap and below-graph position. Hidden panes render no frame.

All ten selected Firefox file/HTTP variants pass: four persistence/navigation, two async-race, two metadata/pane-frame and two viewport-boundary variants. The extended presentation procedure checks all four borders, exact alignment, inset and absence of card shadows. Product changes are CSS-only; no state transition or diagram topology changes. Earlier broad regressions remain historical, not a new full release result. Source and packaged CSS hashes match.

  • Pane-frame regression report (retained in the engineering checkout)

  • Runner log (retained in the engineering checkout)

  • Follow-up source/environment identity (retained in the engineering checkout)

The actual normal demo was visually inspected with two open boxes. The frame makes the clipping boundary visible from top to bottom, aligns with the graph separator and leaves a clean inset around the first box. Normal Sphinx and complete shared evidence validation pass (115 canonical cases, 809 executions).

  • Framed normal demo (retained in the engineering checkout)

  • Actual demo measurements (retained in the engineering checkout)

  • Follow-up documentation build (retained in the engineering checkout)

  • Follow-up evidence validation (retained in the engineering checkout)