test(docx-core,docx-mcp): final regression suite for canonical emission (#143) by stevenobiajulu · Pull Request #175 · UseJunior/safe-docx

stevenobiajulu · 2026-05-08T02:19:39Z

Summary

FINAL sub-PR for #120. Builds the comprehensive regression suite that locks in the canonical-emission contract introduced by #135–#142. After this lands, #120 fully closes out and the umbrella checklist on #118 ticks #120.

Three test groups across two packages:

Group	Location	Count	Purpose
A — Per-primitive emission	`docx-core/src/integration/canonical-emission-regression.test.ts`	10	For every Table A primitive, invoke with ctx and assert expected revision element
B — Round-trip with comparison	same file	4	Apply primitive, run compareDocuments, assert AI revisions survive semantically
C — Tool integration through SessionManager	`docx-mcp/src/integration/canonical-emission-mcp.test.ts`	10	For every Table A MCP tool, exercise via SessionManager + save + unzip

Total: 24 new tests locking in the surface.

The litmus test (Group B)

The replace_text round-trip test deliberately locks in two claims:

Claim	Result
AI author identity IS preserved through compareDocuments	✅ Asserted
AI date timestamps are NOT preserved (regenerated by comparison)	✅ Asserted as a gap

This is a known partial-correctness boundary that #126 (remove comparison from default save path) will close. The test asserts the current gap explicitly so future maintainers get an alert when comparison gains date-preservation:

expect(
  allDatesAreFixed,
  'Comparison currently regenerates dates. If this test now passes (all dates preserved), comparison gained date-preservation — update this test to assert preservation instead. See #126.',
).toBe(false);

SUPPORT.md updates

Verified by [120.8] (#143) regression test. note appended to each of the 20 write-time emitter rows in Table A.
compare_documents and save (tracked branch) rows are intentionally NOT marked as verified — they're comparison-time, not write-time.
replaceParagraphTextRange row reflects the rPrChange follow-up replaceParagraphTextRange should emit w:rPrChange when run formatting changes #173.
addCommentReply row reflects addCommentReply should emit body revision markup OR SUPPORT.md should be softened #174 Table B classification.

Peer review

Reviewer	Severity	Finding	Resolution
Codex	Medium	SUPPORT.md verification scope drift (22 Table A rows but only 20 verified)	Confirmed correct — the 2 unverified are comparison-time legacy paths; explicitly noted in their existing notes
Codex	High	Group B assertions too weak; wouldn't catch wrapper regeneration with same author	Fixed — strengthened to capture `{kind, id, author, date, textContent}` tuples; the strengthened test caught a real gap (date regeneration) which is now locked in as a known limitation
Codex	Low	Group C is tool-integration, not MCP dispatch	Fixed — describe block renamed to "Tool integration through SessionManager: canonical revision emission"; comment clarifies scope
Both	Normal	rPrChange and addCommentReply gaps need follow-ups	Filed as #173 (rPrChange) and #174 (addCommentReply); SUPPORT.md cross-references both
Gemini	—	"Approval Recommended. Coverage maps perfectly to the 20 primary emitters"	Confirmed

Verification

npm run build -w @usejunior/docx-core   # clean
npm run lint:workspaces                  # clean
npm run test:run -w @usejunior/docx-core
# Test Files  83 passed (83)
# Tests  1208 passed | 1 skipped (1209)
npm run test:run -w @usejunior/docx-mcp
# Test Files  71 passed (71)
# Tests  720 passed (720)

Total impact across the #120 chain

Sub-PR	Tests added (cumulative)
#135 [120.0] foundation	docx-core 805 → 810
#136 [120.1] text	810 → 1102 (+9)
#137 [120.2] document	1102 → 1110
#138 [120.3] comments	1118 → 1147
#139 [120.4] footnotes	1147 → 1155
#140 [120.5] layout	1155 → 1166
#141 [120.6] clear_formatting	docx-mcp 696 → 696
#142 [120.7] author config	docx-mcp 696 → 710
#143 [120.8] regression (this PR)	+24 across both packages

Total chain: ~138 new tests covering the canonical-emission shift end-to-end.

Follow-ups filed (out of scope)

Extend accept_changes / reject_changes to handle revisions in footnotes.xml and comments.xml #165 — accept_changes side-part extension
Extend inferStartingRevisionIdState to scan side parts (comments.xml / footnotes.xml) #171 — inferStartingRevisionIdState side-part scanning
replaceParagraphTextRange should emit w:rPrChange when run formatting changes #173 — replaceParagraphTextRange w:rPrChange emission
addCommentReply should emit body revision markup OR SUPPORT.md should be softened #174 — addCommentReply Table B reclassification

Closes

[120.8] Regression suite for canonical revision emission across the surface #143 ([120.8] of Make tracked changes the canonical representation for the supported surface #120)
After merge: closes Make tracked changes the canonical representation for the supported surface #120 and ticks the umbrella checklist on [Tracking] SafeDocX: scoped tracked-changes-as-canonical, comparison decoupled #118.

…on (#143) Closes #143. **FINAL sub-PR for #120** — closes out the umbrella shift. Builds the comprehensive regression suite that locks in the canonical- emission contract introduced by #135-#142. Three test groups: **Group A — Per-primitive emission** (10 tests in docx-core): For every Table A primitive, invoke with ctx and assert the expected revision element is emitted with valid w:id/w:author/w:date attributes. Walks SUPPORT.md as the source of truth. Locks current behavior including the known rPrChange gap (#173) and addCommentReply Table B classification gap (#174). **Group B — Round-trip with comparison** (4 tests in docx-core): Apply primitive with ctx, run compareDocuments, assert AI revisions survive semantically through the comparison pipeline. The litmus test for the umbrella's correctness story. The replace_text round-trip test deliberately locks in TWO claims: - AI author identity IS preserved through compareDocuments - AI date timestamps are NOT preserved (regenerated by comparison) This is a known partial-correctness boundary that #126 (remove comparison from default save path) will close. The test asserts the current gap explicitly so future maintainers get an alert when comparison gains date-preservation — the failure message tells them to update the assertion to require preservation, marking #126 progress. **Group C — Tool integration through SessionManager** (10 tests in docx-mcp): For every Table A MCP tool, exercise via SessionManager + save + unzip. NOT full MCP-dispatch end-to-end (those would also exercise server.ts CallToolRequestSchema handler) — naming clarifies the scope. Asserts revision elements with w:author='SafeDocX' in the saved buffer. SUPPORT.md updates: - 'Verified by [120.8] (#143) regression test.' note appended to each of the 20 write-time emitter rows in Table A. - compare_documents and save (tracked branch) rows are intentionally NOT marked as verified (they're comparison-time, not write-time). - replaceParagraphTextRange row reflects the rPrChange follow-up #173. - addCommentReply row reflects the #174 Table B classification. Peer review: Gemini (LGTM) + Codex (4 findings, all addressed): 1. SUPPORT.md verification scope drift — confirmed only 20 of 22 Table A rows are verified; the 2 unverified are comparison-time legacy. 2. Group B too weak — strengthened to capture pre/post-compare revision tuples and lock in the date-regeneration gap explicitly. 3. Group C is tool-integration not MCP dispatch — describe block renamed for clarity. 4. rPrChange and addCommentReply gaps — filed as #173 and #174 with SUPPORT.md cross-references. Verification: npm run build -w @usejunior/docx-core # clean npm run lint:workspaces # clean npm run test:run -w @usejunior/docx-core Test Files 83 passed (83) Tests 1208 passed | 1 skipped (1209) npm run test:run -w @usejunior/docx-mcp Test Files 71 passed (71) Tests 720 passed (720) After this lands, #120 fully closes out and the umbrella checklist on #118 ticks #120. Total tests added across the entire #120 chain: 1102 → 1208 (docx-core, +106) and 688 → 720 (docx-mcp, +32) = +138 new tests covering the canonical-emission shift end-to-end. Follow-ups filed (out of scope for this PR): - #165 — accept_changes side-part extension - #171 — inferStartingRevisionIdState side-part scanning - #173 — replaceParagraphTextRange rPrChange emission - #174 — addCommentReply Table B reclassification No changes to baselines/.

vercel · 2026-05-08T02:19:44Z

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

Project	Deployment	Actions	Updated (UTC)
site	Ready	Preview, Comment	May 8, 2026 2:26am

- Remove unused W import and ISO_Z_RE constant - Correct epic label from 'Document Editing' to 'Document Comparison' - Type-strengthen toPartMap so Record<K, string> destructuring is well-typed CI's tsc was stricter than local lint cache; these errors only surfaced on the clean CI build.

codecov · 2026-05-08T02:32:09Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

github-actions Bot added the test label May 8, 2026

stevenobiajulu enabled auto-merge (squash) May 8, 2026 02:19

vercel Bot deployed to Preview May 8, 2026 02:26 View deployment

stevenobiajulu merged commit 7dc0513 into main May 8, 2026
22 checks passed

This was referenced May 8, 2026

Make tracked changes the canonical representation for the supported surface #120

Closed

[Tracking] SafeDocX: scoped tracked-changes-as-canonical, comparison decoupled #118

Open

This was referenced May 26, 2026

epic: ship LLM-Based Quality Gate (3-phase rollout) #252

Open

feat(ci): add LLM-Based Quality Gate (Phase 1, advisory) #253

Merged

This was referenced May 27, 2026

docs(scripts): align conformance script headers with canonical-README contract #271

Merged

fix(ci): harden LLM-gate against npm config-load failures (closes #273) #274

Merged

fix(ci): pass --repo to gh pr view in workflow_dispatch path (closes #275) #276

Merged

stevenobiajulu deleted the 143-regression-canonical-emission-20260508 branch May 27, 2026 17:07

github-actions Bot mentioned this pull request Jun 3, 2026

test(docx-comparison): fragmented-field fast-check arbitrary for the Lean spec bridge #296

Merged

7 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test(docx-core,docx-mcp): final regression suite for canonical emission (#143)#175

test(docx-core,docx-mcp): final regression suite for canonical emission (#143)#175
stevenobiajulu merged 2 commits into
mainfrom
143-regression-canonical-emission-20260508

stevenobiajulu commented May 8, 2026

Uh oh!

vercel Bot commented May 8, 2026 •

edited

Loading

Uh oh!

Uh oh!

codecov Bot commented May 8, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

stevenobiajulu commented May 8, 2026

Summary

The litmus test (Group B)

SUPPORT.md updates

Peer review

Verification

Total impact across the #120 chain

Follow-ups filed (out of scope)

Closes

Uh oh!

vercel Bot commented May 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

codecov Bot commented May 8, 2026

Codecov Report

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

vercel Bot commented May 8, 2026 •

edited

Loading