fix(docx-compare): remove accepted ghost section breaks

[stevenobiajulu]

Summary

• removes a ghost paragraph-level section break left when accept-all discards a w:sectPrChange snapshot whose container has no live section properties
• preserves body-level final section properties and paragraph sections with live formatting
• leaves the exact formatting-fidelity gate and thresholds unchanged
• adds a strictly validated OpenSpec delta plus independently invented XML and built-DOCX regressions

Dependency of UseJunior/legal-context#918. This PR does not close that issue.

Exact revision

• base (main): 31b1d8fd2e9f0f285cc6167906ef6e2c9f220699
• head: 3b9f80095c753bcb4d99771b5adabf906fd0eebb

Evidence

• invented pre-fix-equivalent invariant: formatting score 0.6666666666666666, with two divergent section units caused by the empty paragraph-owned section node
• invented fixed regression: accept 1.0, reject 1.0
• private verified real comparison: succeeded through the unchanged publication gate; accept 1.0, reject 1.0; zero run, paragraph, table, or section divergences and zero unaligned paragraphs
• private source text, filenames, paths, and artifacts are not included here or in the commit/package

Verification

• focused comparison tests: 69 passed
• npm run build
• npm run lint:workspaces (six pre-existing unused-disable warnings, zero errors)
• full workspace tests; sandbox-only IPC/LibreOffice invocation failures were rerun under normal managed approval and passed (docx-core: 1,314 passed/1 skipped; render verifier: 56 passed)
• npm run check:spec-coverage
• npm run check:conformance-citations
• npm run check:conformance-doc
• openspec validate remove-accepted-ghost-section-break --strict
• package audit: 12 tarballs / 1,207 entries, zero private sentinel hits

Review status

Supplemental same-model self-review found and fixed a predicate hardening issue by requiring the change node itself to be the sole direct element child. No independent-family review was available under the owner constraint; this self-review does not satisfy that independent gate.

READY for review. No merge, automerge, deployment, or publication is requested or authorized.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
site	Ready	Preview	Sep 14, 2026 6:32pm UTC

[stevenobiajulu]

Supplemental implementation self-review adjudication (same model; not an independent-family review):

• Finding — predicate precision: child-count alone was not explicit enough about which child authorizes removal. Adjudicated/fixed: the implementation now requires the w:sectPrChange node itself to be the sole direct element child of a paragraph-owned w:sectPr.
• Finding — preservation controls: an over-broad empty-container cleanup could remove final or live section properties. Adjudicated/fixed: dedicated regression coverage proves live paragraph section formatting and the body final section container survive acceptance.
• Finding — synthetic evidence level: raw XML alone would not exercise package construction and the publication gate. Adjudicated/fixed: an independently invented built-DOCX comparison now runs end to end. Its pre-fix-equivalent invariant scores 0.6666666666666666; the fixed accept and reject projections both score 1.0.
• Private verification: the verified real pair succeeds through the unchanged exact publication gate at accept 1.0 / reject 1.0, with zero formatting divergences and zero unaligned paragraphs. No source text, filenames, paths, or artifacts are disclosed.
• Leakage: package audit covered 12 tarballs and 1,207 entries with zero private sentinel hits.

Local checks are complete. CI is currently running. Independent-family review is unavailable under the owner constraint; this same-model review is supplemental and does not satisfy that gate. No merge/automerge/deploy/publication is authorized.

[usejunior-llm-gate]

LLM gate (advisory)

1 finding to review (advisory, does not block) - 4 pass, 1 warn, 0 error, 11 skipped, 16 total

Findings (1)

Rule	Verdict	Detail
Re-derived facts vs canonical sources	WARN	The PR implements ghost section break cleanup in trackChangesAcceptorAst.ts:856-871 (acceptAllChanges) but does not apply this logic to the core acceptChanges primitive in accept_changes.ts:451-454, causing these sibling implementations to drift apart despite the requirement that they agree on all cases as noted in trackChangesAcceptorAst.test.ts:948-953.

All 16 rules (5 evaluated, 11 skipped)

Rule	Verdict	Detail
read_file response metadata parity	SKIPPED	paths not touched by this PR
Live DOM namespace-safe OOXML writes	SKIPPED	paths not touched by this PR
Complex-field revisions preserve complete accept/reject state machines	PASS	The PR only modifies paragraph-level section break removal logic and does not touch field atomization, validateFieldStructure, w:fldChar, w:instrText, w:delInstrText, or collapsed-field comparison logic.
Field validation per story, not global	SKIPPED	paths not touched by this PR
Revision IDs seeded from all revision-bearing side parts	SKIPPED	paths not touched by this PR
Accept/reject sweep side parts and caches	PASS	The PR does not meet the precondition as it only touches AST candidate projection helpers in packages/docx-compare/src/tagged/trackChangesAcceptorAst.ts and does not modify DocxDocument, REVISION_STORY_PART_PATHS, accept_changes, reject_changes, or any side-part revision markup.
DocumentViewNode.heading stays canonical	SKIPPED	paths not touched by this PR
AI-author parity across entry points	SKIPPED	paths not touched by this PR
Property-change wrapper discipline	SKIPPED	paths not touched by this PR
SUPPORT.md Table A drift vs. implementation	SKIPPED	paths not touched by this PR
Table A / Table B boundary on side-part revisions	SKIPPED	paths not touched by this PR
Canonical-emission surface completeness	SKIPPED	paths not touched by this PR
Unit-test quality (avoid tautological / change-detector tests)	PASS	The newly added tests in packages/docx-compare/src/tagged/formattingFidelity.test.ts:267 and packages/docx-compare/src/tagged/trackChangesAcceptorAst.test.ts:229 are independent of the system under test, asserting concrete semantic XML structures and exact fidelity scores from first principles without relying on SUT mocks or behavioral snapshots.
Re-derived facts vs canonical sources	WARN	The PR implements ghost section break cleanup in trackChangesAcceptorAst.ts:856-871 (acceptAllChanges) but does not apply this logic to the core acceptChanges primitive in accept_changes.ts:451-454, causing these sibling implementations to drift apart despite the requirement that they agree on all cases as noted in trackChangesAcceptorAst.test.ts:948-953.
.openspec tag ↔ test-assertion drift	PASS	The added .openspec tag in review/packages/docx-compare/src/tagged/formattingFidelity.test.ts:269 matches a test body that fully exercises and asserts the scenario's GIVEN, WHEN, and THEN criteria without any tag-stuffing.
Library stays general (no downstream-domain leakage)	SKIPPED	paths not touched by this PR

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...docx-compare/src/tagged/trackChangesAcceptorAst.ts	85.71%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!

[stevenobiajulu]

Final CI update: all required checks are complete and successful at head 3b9f80095c753bcb4d99771b5adabf906fd0eebb; the PR is READY and GitHub reports a clean merge state. Neutral/skipped advisory or deployment-only jobs are non-failing. No merge, automerge, deployment, or publication was performed.