test(docx-markdoc): bound real ILPA annotation projections independently - #995
Conversation
Two consecutive Node22 CI runs time out while one test aggregates two real ILPA sources and two destination projections under a single thirty-second limit. Register the same four source/destination combinations as individual cases so a slow runner can complete and report each one. Keep every hyperlink, named-style, size and destination-relationship assertion and the thirty-second per-case bound; do not skip inputs, raise timeouts, weaken assertions or change compiler/runtime behavior. The full annotation file now reports twenty-one passing tests instead of eighteen, with each public projection independently accountable. Keep this shipping-blocker repair separate from the focused native cleanup PR. Ref: #982, #993
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
LLM gate (advisory)All evaluated rules passed - 5 pass, 0 warn, 0 error, 11 skipped, 16 total FindingsNone. All 16 rules (5 evaluated, 11 skipped)
|
|
Independent dynamic Claude Opus 5 review APPROVE at exact The reviewer executed both old and new test bodies, counted all 22 original assertion evaluations and 24 new evaluations, proved four distinct callbacks, and ran the complete 21-test annotation file twice. The split preserves the per-case 30-second limit, while explicitly expanding the aggregate group budget. It adds two imports rather than skipping product work. Historical compiler performance has not been bisected; local timing extrapolations are not CI measurements. Actual required Linux Node22 job passed, with all 113 package tests passing. The four public ILPA compiler projections measured:
Proof: https://github.com/UseJunior/safe-docx/actions/runs/35271960534/job/105373751003 . These are actual logged completion times, not the review's extrapolations. Node20 coverage and the required CI corpus are still running. Automerge remains unarmed until CI completes. No check or protection bypass. Advisory complete verdict: 5 pass / 0 warn / 0 error / 11 skipped. Separate combined runtime verification passed all six local gates (3513 passed / 57 platform or opt-in skips) and required21corpus / 23-row manifest; that supplements but does not replace this PR's checks or future exact-head receipts. Non-blocking review observation: untouched SDX-MDOC-103 has tighter default-timeout headroom (actual Node22 3626ms); monitor separately rather than silently changing its timeout in this focused PR. Word and full-package XSD validity are not newly claimed. |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
|
Review precision correction, with independent dynamic Opus re-adjudication at unchanged The initial local review mistakenly required four passing Markdoc cases on Node20. Reading the actual complete The measured four-case proof is Node22 only:10575/7687/7443/7644ms; all113Markdoc tests passed. No Node20 Markdoc coverage is claimed. Node20's actual required shared-library coverage job still must finish green before merge, with no bypass. Its inherited package-matrix omission is a separate characterization/CI follow-up rather than an implicit configuration expansion in this focused test-only PR. Complete corrected local report: |
Post-merge smoke passedMerged and built exact detached main: All six mandatory gates passed: build, workspace lint, full suite (3429 passed / 47 platform or opt-in skips), spec coverage, conformance citations and conformance document. Focused public ILPA annotation suite 21/21 passed. Built-code smoke independently imported and compiled BOTH bundled public ILPA forms × comment/footnote (4/4), checking preserved link text, Hyperlink style, 18-half-point size and valid External relationships in the actual destination parts. Inputs matched the Git blobs from the exact merge; all runtime imports came from this smoke worktree's newly built dist. Production docs: READY Visual evidence: skipped—test registration only, no document/compiler behavior change. Word and full-package XSD validity are not newly claimed. First 50 smoke-log linesLast 50 smoke-log lines |
1 participant
Summary
Split the real ILPA hyperlink annotation regression into four named cases: whole-of-fund/deal-by-deal × comment/footnote. Each case retains the existing 30-second timeout, public input, import assertions, compiler entry point, hyperlink formatting checks and destination-part external relationship checks.
The prior single callback aggregated all four full-document projections and timed out in required Node22 CI twice on PR993 and once on PR994. This change explicitly increases the aggregate time available across independently bounded cases; it does not claim unchanged total budget or a product performance improvement. No runtime, fixture, CI configuration, skip, coverage exclusion or assertion is changed.
Ref: #982. Unblocks final integration checks for #979 and #985 as well. Does not close those runtime issues.
Verification
.peer-review/ilpa-ci-blocker-opus5-20260917.review.txt.Actual Node22 CI now passed all113Markdoc tests; the four cases measured10575/7687/7443/7644ms. A second dynamic Opus clarification retained APPROVE and corrected an initial review premise: Node20's coverage path tests core/compare/MCP plus the two runtime packages, not Markdoc. No Node20 execution of these four cases is claimed. The actual required Node20 shared-library coverage job must still pass; its inherited package omission is a separate follow-up, not a reason to bypass a check or silently change CI scope here.
Visual evidence is skipped: this is test registration only, with unchanged document/compiler behavior. Word remains unverified. Hosted API build/deploy is out of scope.
Non-blocking review observations: total work grows by two imports (~6% locally); Node22 timing predictions are extrapolations, not measurements; the untouched SDX-MDOC-103 real-input test has tighter CI timing headroom and should be monitored separately. No historical compiler performance bisect was performed. Actual required CI remains decisive.