STV 1.1.0 final qualification

Software baseline: 41e57ed76c2acb521cdeb033f4be7b01fac50115. Final executable test baseline: 3b32d6be34a8a5c3bea1f118e4b24cb919901a28. The owner requested completion of extended qualification before publication, then authorized continuing the release. No publication is asserted by this pre-publication execution record; successful hosted checks remain required.

Observed results

  • 229 automated cases passed: 187 Python and 42 JavaScript, with zero failures, errors or skips, in one complete final selection. The Python wrapper additionally checks two Node reset scenarios; these are not added again to the 229-case total.

  • All 120 current canonical cases have actual execution evidence; seven infrastructure checks are retained separately. The four occurrence-mode cases remain explicitly deferred. Acceptance statuses are preserved.

  • A fresh offline Python 3.10.9 installation of the exact candidate wheel passed dependency checks, installed-package import, the CI runtime command and a separate clean Sphinx host build. Main qualification used Python 3.13.17, Windows 11 and Firefox 158.0.1 / geckodriver 0.37.1.

  • The 30,000-Need / 90,000-link host fixture was reused only after checking exact wheel, runtime/source, dependency, toolchain and fixture equivalence. Current browser capacity measurements ran again; this is not a new host build.

  • The four 200-node / ten-card SVG checks at wide/narrow widths and file/HTTP remain applicable through exact product and fixture byte equivalence. Their original timestamps, files and hashes remain in candidate 2.

  • Real visible-hover export passed separately on file and HTTP, retaining all fully visible source words after the live hover closed. Source/reopened screenshots were inspected; the source and SVG geometry checks remain linked.

  • The owner observed native Save As, folder selection and cancellation, then confirmed readable, usable cards/tooltips and keyboard scrolling at 125% and 150% zoom in Firefox 158.0.1 (64-bit). These are owner observations, distinguished from agent-run browser assertions.

  • Fresh inspection confirms 16 distinct archive identities and complete licensing supplements. Issued 1.0.0 distributions remain unchanged.

The original runner returned nonzero only because its initial documentation build found two newly added evidence pages missing from a toctree. The entries were corrected without modifying product or executable test inputs. The failed log and original result remain intact; final strict documentation and evidence checks are retained separately in final-validation.json. Only scoped verification metadata changes after the full run; criteria and acceptance statuses are unchanged.

Candidate history

Candidate 1 is rejected for SVG export omits visible gr... (STV_PR_016), missing graph nodes in exported SVGs. Candidate 2 fixes that defect and passed its 196-case selection, clean minimum-runtime check and maximum-size supplements. Candidate 3 adds extended procedures but its Node wrapper expected a different default reporter format: 228 passed and one test-wrapper failure. Its actual two Node assertions passed. Explicit TAP output corrects the wrapper in the committed final test baseline; candidate 4 passes all 229 executable cases. Failed attempts are retained.

The source and wheel rebuilt for candidates 2, 3 and 4 are byte-identical. Publish only the final selected assets below; older candidate-1 copies in dist are rejected and remain retained for history.

Exact publication assets

4cfed985248b3f328d46c5ffb57a15b5c4d89c9be0f4296e0c3568bba18549eb  sphinx_traceability_view-1.1.0-py3-none-any.whl
6e964833b68488b722bc761a0116155e589b3928813abeaa8d61a5824bf73785  sphinx_traceability_view-1.1.0.tar.gz

The final GitHub Release must carry these exact distributions and their SHA256SUMS before publication. The isolated workflow validates the tagged source/assets, publishes to PyPI through Trusted Publishing, then deploys the matching documentation. The first live PyPI/Pages run is a publication-time check, not something the local tests can prove.

Evidence

  • Final local gates (retained in the engineering checkout)

  • Canonical selector results (retained in the engineering checkout)

  • Recorded test baseline (retained in the engineering checkout)

  • Capacity reuse proof (retained in the engineering checkout)

  • Fresh runtime commands (retained in the engineering checkout)

  • Historical licensing check (retained in the engineering checkout)

  • Verification assessment (retained in the engineering checkout)

  • Exact checksums (retained in the engineering checkout)

Scope remains Windows/Firefox on file and HTTP. Additional OS/browser versions, mobile/touch optimization and universal vector-editor support are not claimed; occurrence mode remains deferred. Optional cleanup is proposed only. Nothing was deleted; all distributions and evidence remain, and the post-release workspace review remains due after publication.