Skip to content

Test Report — Year Planner for React (RD-01 through RD-12) ​

Date: 2026-11-13 Scope: M1 foundation/temporal core/classic grid; M2 layers, visibility, paint/erase, ribbons, hit index and print planning; RD-06 rendered view, theming, slots, inspector, adapters and A3 print output; RD-07 pointer, keyboard and touch interaction, selection matrix, previews and focus; RD-08 controlled commands, validation, versions, rollback and accepted-not-applied handling; RD-09 accessibility semantics, focus, live regions, contrast and axe evidence; RD-10 traceability, browser/visual/print matrices, timezone and SSR evidence, performance budgets, issue #3 optimization, security/privacy gates, dependency audit and CI evidence; RD-11 documentation, Storybook and API reference; RD-12 release versioning, OIDC automation, package metadata, governance and the release checklist; final internal-use licence, functional layout guide and 56-theme refresh Workflow: CodeOps (exec-plan), plans m1-foundation-core, m2-layers-ribbons, m3-react-view, m4-interaction-engine, m5-controlled-commands, m6-accessibility, m7-quality-gates, m8-documentation and m9-release-maintenanceVerdict: PASS for RD-01 through RD-12, with the release-time owner actions recorded in docs/releasing.md. The manual NVDA/VoiceOver audit remains parked; no WCAG conformance claim is made.


Commands ​

bash
pnpm install --frozen-lockfile
pnpm verify
pnpm verify:full
pnpm test:quality
pnpm test:timezones
pnpm test:audit
pnpm test:licenses
pnpm test:environment
pnpm test:perf            # report mode; PERF_STRICT=1 gates the issue #3 oracle
pnpm bench                # report-first; --enforce gates the 2x ceiling
pnpm test:flaky           # triage only (repeat-each=3, Chromium)
pnpm test:consumer
pnpm run docs

verify runs:

text
toolchain check → package and tools typecheck → lint → unit/component coverage
→ quality contracts → license report check → build → API Extractor report check
→ bundle size → Storybook build → TypeDoc → publint/attw

verify:full adds Playwright, the three-timezone suite, the dependency audit, the performance budget run, the benchmark orchestrator and the clean-consumer install.

Browser locale fallback regression (2026-11-14) ​

The published 0.3.1 Armenian story translated planner messages but showed English calendar dates in Chromium. Its native Intl.DateTimeFormat omitted Armenian, Georgian and Icelandic, while the Node test environment supported them. The old browser smoke test checked the message switch without asserting month/day labels.

  • Before implementation: 16 targeted React locale assertions and all 4 Chromium language tests failed on the English fallback.
  • After review corrections: all 54 targeted React tests and all 5 Chromium language tests pass. The browser tests include printed month and weekday labels and Arabic digits in the built Armenian Storybook component.
  • Regression coverage includes calendar names, accessible dates, inspector periods, runtime switching, exclusive all-day ends, year/timezone/DST boundaries, numeric/clock extensions and invalid locale input.
  • Full local release verification passed on an isolated snapshot of the exact release inputs. This preserved the unrelated marketing PDF, which the docs binary-asset contract rejects in the working copy.
Release evidenceResult
Package and quality tests226 core, 568 React, 10 language and 255 quality tests passed
Browser matrix187 passed, 74 intentional platform/feature skips
Timezones / packed consumerThree 226-test timezone runs and packed consumer passed
PerformanceProduction suite passed; report-mode initial-render misses remain recorded
Independent reviewAll three user-approved findings fixed; 18/18 browser/native comparisons passed
PR #15All active CI checks passed; merged as 55a6154
Release 43591487674Full verify, build, publish and registry smoke passed
npmAll three packages at 0.3.2 / latest; SLSA provenance references 55a6154 and release.yml
Pages 43592773688Deployment passed
Live ChromiumArmenian շբթ, հնվ, 4 հնվ, 2016 թ.; Georgian and Icelandic calendar/print labels; Arabic digits ٤ հնվ, ٢٠١٦ թ.; no page errors

Results ​

CheckResult
Core tests226 passed
React tests542 passed (51 files)
Core coverage97.58% lines, 95.78% branches, 99.50% functions
React coverage94.58% lines, 86.09% branches, 95.29% functions
Playwright172 passed, 74 skipped (PDF and axe are Chromium-only; forced-colors is Chromium-only; touch is Chromium-only; one pending-preview timing test is quarantined on all desktop engines). The full matrix now runs as five independent shards
BrowsersChromium, Firefox, WebKit, Pixel 5 Chromium, iPhone 13 WebKit, Chromium touch
A3 print (M1 view)1 page, 1191.12 × 842.88 pt
A3 print (RD-06 24 months)2 pages, 1191.12 × 842.88 pt
A3 print (RD-06 36 months)3 pages, 1191.12 × 842.88 pt
A3 print boundaries15/28/29 months print 2/2/3 pages; 12/14/18/24/36 verified
Quality tests255 passed (traceability, public API, coverage, size, CSS minifier, CI, package, license, audit, environment and release contracts)
Dependency auditPASS — no high or critical advisories; two moderate dev-only (glob, minimatch) accepted
LicensesPASS — THIRD_PARTY_NOTICES.md current, including palette attribution
Clean consumerPASS — packed types (tsc --noEmit), CSS import and production build; registry mode added for post-release
StorybookPASS — static build succeeds; 186 stories across 16 files
API ExtractorPASS — reports checked in
TypeDocPASS — docs/api/core, docs/api/react and docs/api/languages generated
Docs sitePASS — classic layout, licence, localization, browser-test and theme pages build (501 files)
Bundle sizeReact entry + CSS 46.35 kB gzip / 60 kB; total 69.59 kB gzip / 72 kB (core, React, adapters and CSS, including the English fallback). Languages 86.95 kB gzip / 95 kB for all 45 catalogues; largest single language 2.22 kB / 8 kB
Module formatESM-only — "type": "module", import-only exports, ES build
CIverify, browser-plan, browser-build, sharded browsers, security, performance and dco jobs configured; browser evidence and benchmark artifacts uploaded
Browser versionsChromium 158.0.8127.18, Firefox 159.0, WebKit 27.2 (Playwright 1.65.0)
Timezones226 core tests pass under each of UTC, Pacific/Kiritimati and America/Los_Angeles (678 runs total; final gate)

RD-13 Storybook demos ​

  • Navigation starts with Demos: select a range, move an item, resize an item, paint a layer, navigate by keyboard and raise a context menu. Every demo has a play function.
  • Demos/Event Log shows timestamped selection, command and context-menu callbacks beside the planner.
  • Every one of the 186 stories across 16 files has a story-level description; the Reference catalogue remains intact.
  • The Fluent UI and MUI context-menu examples are host-owned and dev-only.

The coverage thresholds are enforced by Vitest: core 95% and React 80% for lines, branches, functions and statements.

RD-07 evidence ​

AreaEvidence
Modesauto, inspect, select, paint, erase, create, move, resize; modes are a TypeScript union; no runtime mode validator is claimed
SelectionPlain/Shift/Ctrl/Cmd date and item matrix; anchor/focus transitions; visual-order item ranges; merged/split ranges; Ctrl/Cmd+A; reserved Ctrl/Cmd+Shift
PointerGrid-root delegation, 4 px threshold, pointer capture, rAF move coalescing, edge auto-scroll, lost-capture cancel, one-shot click suppression
CommandsMove, resize, paint, erase, create and delete through the existing pipeline; duration/grab offset; one-day minimum; disabled ranges rejected
PreviewPer-row clipped date/move/resize/paint/create previews; fixed marquee; invalid red preview; pointer-events: none
KeyboardArrows skip disabled dates; Shift+Arrow range; Space toggle; Enter inspector; Escape cancel/clear; Delete focused/sole item
FocusCommit focus restoration; hidden/removed item pruning; nearest enabled-date fallback; polite live region; controlled selection sync
Touchoff/auto stay scroll-first with touch-action: auto; on enables editing with touch-action: none; real touch E2E
UndoUncontrolled move/resize/paint/erase/create/update/delete; redo cleared by a new edit; controlled mode reports UNDO_UNAVAILABLE
Host controlsHost buttons/inputs in cells keep pointer and key events; no nested control inside an item proxy
SecurityHidden/locked/timed/annotation items cannot start edits; no phantom dates; diagnostics carry no hidden payload

RD-06 evidence ​

AreaEvidence
Public view14/36-month grid, 39 CSS tracks, 37 ARIA columns, sticky header/footer, zoom 0.75–1.5, scroll container and scrollToDate
ItemsCross-month fragments, exact slot widths, deterministic lanes, one focus proxy per item (one tab stop)
Annotation/overflowAnnotation badges, compact 3-lane plus +N popover, hidden exclusion
TooltipDefault title/date span and renderTooltip replacement
CellsgetDateMeta background/text/label/ARIA/holiday/disabled/content and annotation overlay
SlotsrenderCell, renderItem, renderAnnotation, renderOverflow, renderEmpty, renderInspector, renderTooltip; structural components
Proxy contractrole="button" proxy, no nested interactive controls, sibling resize handles, aria-pressed selection
Theming24 tokens, 56 presets, provider, cssVars, implemented unstyled, data-yp-theme
StateControlled/uncontrolled items/visibility/selection/filter; dev warnings; deterministic seeding
CommandsLocal read-only/version/period checks, host validation, timeout, accept/reject, retry/dismiss, uncontrolled undo/redo, annotation host-ownership
InspectorDetails, visibility/editability, overflow, create/move/resize/paint/erase/delete, keyboard reachable
AdaptersFluent UI v9 and MUI token mappings; real Buttons proven in a dev-only demo fixture across three browsers; no design-system dependency in the library
PrintHidden per-page tree, 37 headers/footers, 1/2/3 A3 pages, per-page scale, labels/weekends/ISO weeks/ribbons
SSRNode renderToString plus jsdom hydration with no mismatch errors
CSPDocumented style-src-attr 'unsafe-inline'; Playwright CSP-header test
SecurityMetadata never rendered; item style and color allowlist; hidden data and stale tooltips excluded

Theme refresh evidence ​

AreaEvidence
Catalogue19 Material 2014 families, 35 Fluent UI v9 families and high-contrast/print; exactly 56 public names
Default and removalsmaterial-teal; classic-blue, classic-yellow and modern have no aliases
Token mappingMaterial 50/100/500/900 and Fluent neutral/Background2/BorderActive/Foreground2 mapped to the existing 24 tokens
ContrastAll 56 presets measured on surface, muted, weekend and context for 4.5:1 ribbon/text and 3:1 focus/border
UtilitiesMonochrome high-contrast and print; print keeps printScale: 1
Demo and StorybookSelectors enumerate the catalogue; Material, Fluent and utility groups; Themes control covers all presets
Fixtures2016 material-teal, 2017 fluent-marigold, PRT-01 in fluent-marigold
Baselines42 visual tests regenerated across Chromium, Firefox and WebKit; print baselines regenerated
Print boundaries12/14/15/18/24/28/29/36-month A3 page counts verified from PDFs
AccessibilityAxe matrix includes material-teal, fluent-marigold, high-contrast, print, overflow, error and 36-month states
NoticesReact package ships palette attribution; root notice repeats it
SizeCompact local palette data and a minified published CSS keep the React surface at 69.59 kB / 72 kB. Optional language catalogues are measured separately: all 45 at 86.95 kB / 95 kB, largest single language at 2.22 kB / 8 kB

Specification evidence ​

AreaTest files
Dates, ISO weeks, leap yearspackages/core/src/date.spec.test.ts, date.impl.test.ts
Half-open intervals and subtractionpackages/core/src/interval.spec.test.ts
Periods and timed projectionpackages/core/src/period.spec.test.ts
Classic layout and dynamic monthspackages/core/src/layout/classic-year.spec.test.ts, classic-year.impl.test.ts
Entity model and validationpackages/core/src/model/validation.spec.test.ts, validation.impl.test.ts
Visibility, filters and editabilitypackages/core/src/model/visibility.spec.test.ts, visibility.impl.test.ts
Paint and erasepackages/core/src/model/paint.spec.test.ts, paint.impl.test.ts
Ribbon fragments, lanes and overflowpackages/core/src/ribbon/fragments.spec.test.ts, fragments.impl.test.ts
Hit indexpackages/core/src/ribbon/hit-index.spec.test.ts, hit-index.impl.test.ts
Print planningpackages/core/src/ribbon/print.spec.test.ts, print.impl.test.ts
Shared contractspackages/core/src/contracts.spec.test.ts, model/selection.impl.test.ts
Theme and providerpackages/react/src/theme.spec.test.tsx, theme.impl.test.tsx
View shell, state and geometryYearPlanner.spec.test.tsx, YearPlanner.ssr.spec.test.tsx, YearPlanner.hydration.spec.test.tsx, YearPlanner.impl.test.tsx, geometry.impl.test.tsx, state.impl.test.tsx
Items, annotations and slotsitems.spec.test.tsx, items.impl.test.tsx
Commandscommands.spec.test.tsx, commands.impl.test.tsx
Inspectorinspector.spec.test.tsx, inspector.impl.test.tsx
Print and adaptersprint.spec.test.tsx, print.impl.test.tsx, adapters.spec.test.tsx, adapters.impl.test.tsx, styles.spec.test.tsx
Interactionpackages/react/src/interaction/selection.spec.test.tsx, gesture.spec.test.tsx, pointer.spec.test.tsx, keyboard.spec.test.tsx, interaction.integration.spec.test.tsx
Interaction implpackages/react/src/interaction/selection.impl.test.tsx, gesture.impl.test.tsx, pointer.impl.test.tsx, keyboard.impl.test.tsx
Interaction E2Etests/e2e/interaction.spec.ts, tests/e2e/interaction-touch.spec.ts
Accessibilitya11y.spec.test.tsx, a11y.impl.test.tsx, announcements.spec.test.tsx, announcements.impl.test.tsx, accessibility.spec.test.tsx, accessibility.impl.test.tsx, interaction/row.spec.test.tsx, interaction/row.impl.test.tsx, styles.spec.test.tsx
Accessibility E2Etests/e2e/accessibility.spec.ts (eight axe states plus keyboard, skip link, hidden item)
Integrationpackages/react/src/integration.spec.test.tsx
E2Etests/e2e/classic-year.spec.ts
Red phasesdocs/m1-red-phase.md, docs/m2-red-phase.md, docs/m3-red-phase.md, docs/m4-red-phase.md

The Playwright suite generates real Chromium PDFs:

text
rd06-24-months-a3.pdf: Pages 2, 1191.12 x 842.88 pts (A3)
rd06-36-months-a3.pdf: Pages 3, 1191.12 x 842.88 pts (A3)
classic-2016-a3.pdf:   Pages 1, 1191.12 x 842.88 pts (A3)

Each RD-06 page uses its core-planned uniform scale and repeats the weekday header, month labels, weekend tints, ISO weeks, ribbons and annotations.

API documentation ​

API Extractor reports are committed at:

  • packages/core/etc/year-planner-core.api.md
  • packages/react/etc/react-year-planner.api.md

TypeDoc generates docs/api/core/index.html, docs/api/react/index.html and a landing page at docs/api/index.html. Generated HTML is not committed.

RD-08 evidence ​

AreaEvidence
Request identityOne ID per request; active/settled/obsolete state; collision suffix; late and obsolete responses ignored
ContextOptional baseRevision; optional AbortSignal third argument on validation and request callbacks
ValidationLocal checks first; error diagnostics block even when permitted; warnings/info display as inspector notices
VersionsPositive-integer item versions; expectedVersion on update/move/resize/delete; expectedVersions on paint; STALE_REVISION
OutcomesRuntime normalization; malformed outcomes map to HOST_REJECTED; accepted paint coverage is proved
PendingPending item outline and aria-busy; gesture preview retained until settlement; one command per item
RollbackPreview clears; focus returns to the source item; paint/create fall back to the gesture date; selection is preserved
Accepted-not-appliedControlled props are checked immediately and on change; timeout produces ACCEPTED_NOT_APPLIED and reverts
RetryExplicit host rejection, thrown call and timeout matrix; no delete, timed-out create/paint or stale replay
Uncontrolled paritySame reducer, version increments, paint residuals and undo/redo; controlled mode reports UNDO_UNAVAILABLE
Browser evidencetests/e2e/commands.spec.ts in Chromium, Firefox and WebKit: accept, reject, stale, timeout, never-apply, warning, pending preview and capacity mock
SecurityDiagnostics render as text; local hidden-item diagnostics omit titles/dates; host prose is host-owned; no authorization inferred

RD-09 evidence ​

AreaEvidence
SemanticsRoot region, labelled grid, aria-selected cells, 37 columns, generated shortcut description, item names with period/layer/pattern/status/selection/clipping
Grid structureItem proxies moved to a grid-sibling overlay; axe aria-required-children passes in every scanned state
FocusOne roving tab stop for cells, proxies and the active row’s overflow; Home/End row bounds; disabled-date fallback
Skip linkFirst focusable planner control opens the inspector and focuses the item list
Item listVisible, keyboard-reachable list with rendered-fragment filtering, overflow markers and hidden-item exclusion
Live regionsVisible status bar; hidden polite and assertive regions; pending/accepted/rejected/timeout/stale/aborted/not-applied mapped once
ContrastComputed per-fragment text colour; all built-in presets pass 4.5:1 ribbon text and 3:1 focus/border
Target size24 px lanes, 24×24 resize handles, 24 px overflow, 28 px overflow entries, 26 px inspector buttons
Non-colour cuesSelected outline/weight/underline, dotted holiday underline, +N overflow, layer patterns/labels
MediaForced-colours borders/outlines, reduced-motion rules, print hides status/skip/live regions
Automated axeEight states: blue, yellow, high-contrast, compact, expanded, overflow dialog, command error, AC3 Fluent/MUI and 36 months — all zero serious/critical
Keyboard E2EArrow/Shift+Arrow/Space/Enter, one tab stop per composite, skip link, move-by-week, hidden-item fallback, no nested proxy controls
Manual ATParked by user decision 2026-11-11; Orca attempt recorded; no NVDA/VoiceOver run; no WCAG conformance claim

Performance baseline (RD-10, issue #3) ​

Command: pnpm bench --out docs/benchmarks/rd10-baseline (the 10k cases were run in short chunks under one run id on the 14 GiB reference host; the aggregator writes the same report). Environment: Intel Core i5-3470S (4 vCPU), 15 GB, Ubuntu 26.04, Node 24.21.0, Playwright 1.65.0, production preview at 1600×1200. Run: 2026-11-11T20:30:00.000ZStatus: pre-optimization baseline. Values above the budget are recorded, not hidden.

Scenarionp50p95BudgetStatus
Layout 10,000 items, compact1088.16 ms108.07 ms250 mspass
Layout 10,000 items, expanded10108.61 ms295.20 ms250 msmiss
Initial render 500 items5414.40 ms565.40 ms150 msfail
Initial render 2,000 items52,167.90 ms2,247.60 ms150 msfail
Initial render 10,000 items541,920.00 ms50,457.90 ms150 msfail
Initial render 12/14/18/24/36 months5401–543 ms420–615 ms150 msfail
Drag 500 items, compact26316.70 ms33.40 ms16.7 msfail
Drag 2,000 items, compact26016.70 ms300.00 ms16.7 msfail
Drag 10,000 items, compact (10 steps)4916.70 ms8,916.40 ms16.7 msfail
Drag 500 items, expanded26016.70 ms33.40 ms16.7 msfail
Drag 500 items, 36 months26916.70 ms100.00 ms16.7 msfail
Overflow open/close, 10,000 items593.50 ms115.30 ms100 msmiss
Issue #3 production drag, 500 items3—33.40 ms16.7 msmiss
Issue #3 development drag, 500 items26816.70 ms416.60 ms16.7 msfail

The production 500-item drag is 33.4 ms p95; the development drag is 416.6 ms p95, a development-overhead factor of about 12×. CPU profiles are captured under test-results/bench/profiles, named with the run id (issue-3-production, issue-3-dev, 36-month-500).

pnpm bench is report-first so this baseline can be recorded honestly. pnpm bench --enforce and PERF_STRICT=1 pnpm test:perf gate the 2× ceiling and the issue #3 oracle after the optimization. The 10,000-item drag records 10 steps because the 14 GiB reference host cannot complete 60 steps without exhausting memory; the sample count is printed beside every p95.

Post-optimization results (issue #3) ​

Command: pnpm bench --out docs/benchmarks/rd10-after (same chunked reference-host method) Run: 2026-11-11T21:30:00.000ZChange: row/cell/item/print/inspector memoization, stable interaction and command inputs, per-row preview isolation and a preview-content comparator. No virtualization or external store.

ScenarioBefore p95After p95Long tasks before → after
Drag 500 items, compact33.40 ms16.70 ms27 → 2
Drag 2,000 items, compact300.00 ms16.80 ms62 → 2
Drag 10,000 items, compact (10 steps)8,916.40 ms66.60 ms20 → 9
Drag 500 items, expanded33.40 ms16.80 ms45 → 2
Drag 500 items, 36 months100.00 ms16.70 ms62 → 4
Issue #3 production drag, 500 items33.40 ms16.70 ms64 → 4
Issue #3 development drag, 500 items416.60 ms16.70 ms62 → 2
Overflow open/close, 10,000 items115.30 ms118.30 ms—
Initial render 500 items565.40 ms666.20 ms—
Initial render 10,000 items50,457.90 ms43,436.20 ms—

The production 500-item drag now meets the 16.7 ms frame budget. The issue #3 oracle (after p95 ≤ 0.9× baseline or ≤ 16.7 ms, three repeats, no long-task or input-delay regression) passes under PERF_STRICT=1 pnpm test:perf.

Two results remain documented misses rather than claims:

  • Overflow with 10,000 items stays around 118 ms against the 100 ms budget; the overflow list itself is unchanged by this optimization.
  • Initial render (500/2k/10k) is still far above the 150 ms budget. It is measured and recorded but not claimed; the issue #3 scope was interaction responsiveness. The 10,000-item drag remains a stress case with 10 steps and a recorded ceiling exemption.

Known gaps and honest limitations ​

GapStatus
Full command lifecycle: rollback, stale responses, accepted-not-appliedImplemented — RD-08
Batch multi-move, alt+drag subtraction, drag between layers/resourcesDeferred from R1
Native touch-scroll E2ETouch gating and touch-action are tested; full scroll physics is RD-10
Controlled-host undo historyNot implemented; rejected with UNDO_UNAVAILABLE
Fluent UI v9/MUI example storiesReal Buttons are proven in the demo fixture; accessibility stories cover the states
Annotation editingRendered but host-owned in this slice (AR-31)
Geometry via arbitrary cssVarsCSS variables change visually; use theme tokens or units for layout geometry
Full axe auditAutomated axe matrix green in Chromium; no serious/critical violations
Manual NVDA/VoiceOver auditParked by user decision 2026-11-11 — low priority, not an R1 blocker, no Linux path; WCAG conformance is not claimed
Performance benchmarks (500/2k/10k items)Before/after evidence in docs/benchmarks/rd10-baseline.md and rd10-after.md; drag now 16.7 ms p95. Initial render and 10k overflow remain documented misses
Pending-preview E2EOne geometry-sensitive timing test is fixme on all desktop engines. The state is covered by the React command tests and standalone probes (5/5 runs: preview present, command state pending). Revisit with a serial drag run before release
36-month drag variance16.7–33.3 ms p95 across repeated runs; the checked-in after run recorded 16.7 ms and the budget is met or missed within one frame
React entry tree-shakingThe published entry is a single-file ESM bundle, so importing one export retains the component. The consumer proof asserts packed types, the CSS import and a production build; per-export module splitting is deferred and no tree-shaking claim is made
Mobile touch and RTLTouch editing and marquee are implemented; RTL-specific pointer behavior is not
SVG shape plane and edge-path connectorsDeferred; edge-path emits INVALID_RENDERER_OPTION
Published packagesNot published; pre-alpha
Cross-browser PDF outputChromium baseline only

Toolchain note ​

Vite 8 / Vitest 5 crashed with Illegal instruction on the reference CPU (Intel i5-3470S, no AVX2) through the Rolldown native binding. The verified fallback is Vite 7.3.8 + Vitest 4.1.12, recorded in plan AR #15 and requirements AR #79. Vite 8 / Vitest 5 remain the target on compatible hardware. React 19-only is the support policy (requirements AR #80). Vitest 4.1.12 removes the critical tinypool advisories; the two remaining moderate advisories are development-only.

RD-12 — Release, versioning and maintenance ​

Date: 2026-11-12 Scope: lockstep 0.2.0, package metadata and changelogs, the human release PR, the manual OIDC release workflow, governance/support documents and the release checklist. Verdict: release automation and evidence are complete; the first live publication, repository visibility change, environment creation and branch protection remain maintainer actions.

Commands and results ​

CommandResult
pnpm verifyPASS — 220 core, 508 React, 185 quality tests
pnpm release:checkPASS — versions, changelogs, tarballs, source maps and npm pack --dry-run
pnpm release:publish --dry-run --allow-dirtyPASS — core then React packed, validated and published in npm dry-run mode; no upload
node scripts/apply-branch-protection.mjs --recordPASS (exit 0) — records the private-repo HTTP 403 and the missing second reviewer
pnpm changeset status --since=origin/mainPASS — no pending changesets; the release tree is versioned
bash -n scripts/clean-consumer.shPASS — registry and packed modes parse

What changed ​

  • Both packages are at 0.2.0 in the RD-12 record; RD-13 moves them to 0.3.0, the first published version. They have generated changelogs, package READMEs, checked-in LICENSE copies and complete publish metadata.
  • scripts/release-check.mjs validates versions, changelogs, tarballs, source maps and the npm dry-run allowlist; release:publish refuses pending changesets.
  • scripts/release-publish.mjs packs with pnpm, verifies the exact core dependency, compares published integrity, publishes core before React with npm provenance, and creates one v<version> GitHub release.
  • .github/workflows/release.yml is manual, main-only and environment-guarded; its build job has no id-token, and its publish job installs nothing.
  • Every workflow action is pinned to a full commit SHA; PRs get an informational changeset status check.
  • SUPPORT.md, an expanded SECURITY.md, path-specific CODEOWNERS, grouped Dependabot updates and docs/releasing.md cover governance and support.
  • scripts/apply-branch-protection.mjs checks/applies the required checks, one approval, no direct pushes and the second-reviewer prerequisite.

Release-time owner actions ​

  • Run npm login, pnpm release:publish --bootstrap --skip-github-release for 0.3.0, configure the npm trusted publisher, then publish 0.3.1 through the workflow within the two-day validation window.
  • Enable private vulnerability reporting, Dependabot alerts and Pages; create the npm-production environment; make the repository public; add a second reviewer; apply branch protection.
  • The first OIDC release produces the provenance attestation; the 0.3.0 bootstrap does not, and 0.2.0 is never published.

Known gaps ​

  • The live npm publication and the GitHub release were not executed in this environment (no npm authentication, private repository).
  • Branch protection is unavailable on the current private plan.
  • The manual NVDA/VoiceOver audit remains parked; no WCAG conformance claim.

Repository state ​

  • Source-available internal-use licence; signed origin certification and an explicit contribution grant.
  • ESM-only package; no CJS/UMD build.
  • pnpm verify and pnpm verify:full pass locally.
  • M1 plan: 48/48 tasks; M2 plan: 46/46; RD-06 plan: 46/46; RD-07 plan: 33/33; RD-09 plan: 55/55.
  • No deferred feature is presented as implemented.

RD-11 — Documentation, Storybook and API reference ​

Date: 2026-11-12 Scope: the guide set, Storybook states, TSDoc/API Extractor/TypeDoc, the docs-site artifact and the demo evidence.

Commands and results ​

CommandResult
pnpm verifyPASS — 220 core, 508 React, 138 quality tests
pnpm api-report:ciPASS — both reports checked in, no warnings
pnpm docs:buildPASS — TypeDoc core and React, 0 warnings with fatal validation
pnpm storybook:buildPASS — 183 stories across 15 files
pnpm docs:sitePASS — 241 files; Storybook at /, TypeDoc at /api
pnpm docs:site -- --check-onlyPASS — the built artifact resolves every relative asset and landing link
pnpm verify:fullPASS — 143 E2E passed / 40 skipped; timezone suite 220 × 3; perf 7 passed / 1 skipped; bench 8 passed / 8 skipped; clean consumer PASS
Final tree re-checkThe release gate above ran on the final product code. Two later pnpm verify re-runs under host load average > 9 hit timer-sensitive failures (inspector resize, axe 36-month); each failed test passed alone. The affected suites are unchanged by the documentation work.

What changed ​

  • Fourteen topic guides plus the commands, security and architecture pages; docs/README.md maps every topic.
  • The README quickstart renders a 14-month view from static data with controlled selection and visibility and a print action. The install note is honest: the packages are not on a public registry, so the packed-tarball path is documented.
  • Storybook adds the exact 2017 fluent-marigold PRT-01 print story and Fluent/MUI provider wrapper stories, and configures the built-in Storybook 10 viewport.
  • TSDoc covers every public report entry, @public tags are on the public source surface, and @example is bound to the five entry points.
  • The docs-site script builds the packages first, writes Storybook to the repo root, copies TypeDoc to /api, and verifies relative links. The PR job builds and checks the artifact; the Pages deploy is manual and main-only.
  • The demo renders a shared navigation on every route, including the hidden-data proof at ?security=1.

Deviations ​

  • RD-11 names @storybook/addon-viewport; Storybook 10 provides the viewport in core, so no addon is installed.
  • RD-11 AC7 says "install the package"; the packages are unpublished, so the quickstart uses the packed-tarball/consumer path until the release milestone.
  • The docs site is built and uploaded but not published; the deploy job waits for the public-release Pages enablement.

Known gaps ​

  • The manual NVDA/VoiceOver audit remains parked; no WCAG conformance claim is made. Automated axe evidence is green.
  • No versioned docs and no Storybook MDX guide pages in R1; the guides are repository markdown and the published site is Storybook plus TypeDoc.
  • No live Pages deployment.

RD-13 — Product quality ​

Date: 2026-11-13 Scope: supported-size performance, two-dimensional host fit, header typography, the host-owned context-menu contract, the Storybook Demos group and the 0.3.0 release preparation. Verdict: implementation and evidence complete; the live 0.3.0 publication and repository setup remain maintainer actions.

Results ​

CommandResult
pnpm testPASS — 226 core, 535 React tests
pnpm test:qualityPASS — 211 contracts, including 193 traceability rows
PERF_STRICT=1 playwright ... RD-13-AC-03PASS — initial render 242.0 / 346.4 ms p95; click 32.5 / 59.6 ms p95; drag work 4.9 ms/frame, input delay 14.2 ms
playwright test tests/e2e/fit.spec.tsPASS — 1280 × 800, 1440 × 900, 1024 × 768; 24 px cells and ribbon proxies
playwright test tests/e2e/context-menu.spec.tsPASS — pointer item request, selection, keyboard date request, one request per key
playwright test tests/e2e/storybook.spec.tsPASS — the built Demos story opens its host menu, reports the target and returns focus; Fluent and MUI stories open theirs; the month band and header keep the native menu
pnpm storybook:buildPASS — 183 stories across 15 files, six Demos play stories and the event log
pnpm docs:site -- --check-onlyPASS — 249 files; Storybook at /, TypeDoc at /api
pnpm release:checkPASS — lockstep 0.3.0 versions, changelogs, tarballs and source maps
pnpm release:publish --dry-run --allow-dirtyPASS — core then React packed and validated in npm dry-run mode
pnpm verify:fullPASS — 163 E2E passed / 56 skipped; timezone 226 × 3; audit PASS (two moderate dev-only advisories); performance 10 passed / 1 skipped; benchmark 11 passed / 11 skipped plus the report; clean packed consumer PASS

What changed ​

  • Performance. Visibility is resolved once per model, items are projected once, lanes use a max-end index, the date index is set-backed and a date click re-renders one month row. The large inspector item list starts collapsed. The supported 2,000-item data size meets the revised 400 / 550 ms initial-render, 60 / 80 ms click, 8 ms drag-work and 16.7 ms input-delay gates.
  • Fit. display.fit defaults to "contain" and measures a bounded host region. Columns scale to a 24 px floor; the vertical budget chooses one to three 24 px lanes and scrolls internally when the floor is reached. A 14-month view fits 1280 × 800 with one lane; the benchmark route pins fit: "none" so performance numbers stay comparable.
  • Typography. The month label is centred with a short month and two-digit year. Header and cell sizes come from root CSS variables (heading 11 px, month 13 px); print typography is unchanged.
  • Context menu. onContextMenuRequest reports date, item and annotation targets with the pointer day, coordinates and a restoreFocus closure. Right-click selects an unselected item without a gesture; keyboard requests are deduped; touch keeps the native long-press; Fluent and MUI examples are host-owned.
  • Demos. Storybook opens with Demos: select, move, resize, paint, keyboard and context menu, plus a timestamped event log. All 186 stories have a story-level description. The context-menu stories run in a bounded shell so the full 14-month view and the host menu are visible in one laptop viewport; the Demos story renders its own menu instead of hiding the feedback below the grid. A browser check runs the built Storybook, so a story that renders but does nothing visible fails the gate.
  • Release. Both packages are 0.3.0. docs/releasing.md records the 0.3.0 bootstrap and the 0.3.1 provenance sequence; 0.2.0 is never published. RD-13 AC8 stays partial until the owner records the registry attestation.

Release-time owner actions ​

  • Run the 0.3.0 bootstrap, configure the npm trusted publisher for both packages and publish 0.3.1 within the two-day validation window.
  • Make the repository public, enable Pages and private vulnerability reporting, create the npm-production environment, add a second reviewer and apply branch protection.
  • Record npm view @blendsdk/year-planner-core versions (and the React package) proving 0.2.0 was never published, then mark AC8 proven.
  • The manual NVDA/VoiceOver audit remains parked; no WCAG conformance claim is made.

Known gaps ​

  • Annotation chips keep their 14 px design height; the 24 px floor covers date cells and ribbon item proxies.
  • The original 150 ms initial-render and 50 ms click values remain the post-0.3.0 row-windowing goal.
  • The 10,000-item overflow timing remains a documented stress miss.

Documentation website (VitePress) ​

Date: 2026-11-13 Host: GitHub Pages project site, no custom domain Base path: /year-planner/

The documentation site is one artifact built by pnpm docs:site:

PathContent
/year-planner/VitePress home, guide, tutorials, how-tos and examples
/year-planner/storybook/Static Storybook catalogue
/year-planner/api/TypeDoc reference

The build sets the VitePress base and the Storybook production base, copies both into docs/.vitepress/dist, and verifies every local link and asset against the project base. pnpm docs:site reported 423 files; the artifact includes 186 stories and the generated API tree.

Commands and results ​

CommandResult
pnpm docs:sitePASS — VitePress + Storybook + TypeDoc under /year-planner/; 423 files; links checked
pnpm docs:site -- --check-onlyPASS — the last artifact verifies without a rebuild
playwright test tests/e2e/site.spec.ts --project=chromiumPASS — home/guide/tutorial, interactive story iframe, Storybook manager, story index and API under the real base
pnpm verify:fullPASS — 163 E2E passed / 56 skipped; timezone 226 × 3; performance 10 passed / 1 skipped; benchmark 11 passed / 11 skipped; clean consumer PASS
playwright test tests/e2e/storybook.spec.ts --project=chromiumPASS — the Storybook stories also work at /year-planner/storybook/
pnpm test:qualityPASS — 211 contracts

Content ​

  • Home with a real fitted-year screenshot, feature grid, honest npm status and links to the guide, tutorials, examples, Storybook and API.
  • Tutorials: first planner, controlled commands, theming/adapters, A3 print.
  • How-tos: fit to a host, host-owned context menu, custom inspector.
  • Examples: live interactive stories, themes, print and design-system adapters.
  • Story index generated from storybook-static/index.json, so it cannot drift from the built catalogue.
  • Existing guides are the site's guide pages; links that escaped docs/ point to GitHub.

Deployment status ​

The artifact is built and verified in CI. The repository is public, Pages is enabled with GitHub Actions, and the site is published at https://blendsdk.github.io/year-planner/ with no custom domain. The deploy workflow run 43428872169 completed successfully.

Live checks on 2026-11-13: the home, /storybook/iframe.html?id=full-demo--full-demo, /examples/full-demo.html, /stories.html and /api/index.html return 200; the manager opens the full demo by default; the full demo page and the Storybook preview render the grid and the Northwind brand; the story index shows 184 stories.

Full construction demo (post-release) ​

Date: 2026-11-13 Scope: Northwind Construction's 2027 plan, the Storybook landing story and the VitePress example page.

The demo is a working application, not a mock:

  • Eight seed projects across three crews and three project types, plus four milestone annotations. Each project carries client, crew and budget metadata.
  • Double-click a date to open the allocation dialog; the start is the clicked day and the end date is inclusive. Right-click a project to edit, duplicate, mark complete or delete it.
  • The host renders a details panel, KPI strip, search/status filter, layout switch (all lanes or fit to screen), undo/redo, reset, A3 print and a theme switch.
  • The plan is stored in localStorage under year-planner.full-demo.v1, versioned and validated on load. Reset clears it. The story intentionally has no play function so Storybook does not write before the user acts.
  • Storybook opens on the full demo; the manager redirects a bare catalogue load to full-demo--full-demo. The site links to /examples/full-demo.html.

Commands and results (targeted; no full verify) ​

CommandResult
pnpm lintPASS
pnpm test:qualityPASS — 211 contracts, including 186 stories across 16 files
pnpm storybook:buildPASS — 186 stories across 16 files
pnpm docs:sitePASS — VitePress + Storybook + TypeDoc; 423 files; links checked under /year-planner/
playwright test tests/e2e/full-demo.spec.ts tests/e2e/site.spec.ts --project=chromiumPASS — 10 checks: Storybook landing, allocate/persist, context delete/undo/reset, filter/layout, site home/guide, full-demo iframe, manager, story index and API
pnpm verify:fullNot run for this post-release demo. The last full gate passed at 314f893; this change is Storybook/VitePress/docs/test only, so the owner asked to skip the full gate.

Repository and Pages ​

CheckResult
Repository visibilityPUBLIC via gh repo edit --visibility public
GitHub PagesEnabled with GitHub Actions as the source
Site URLhttps://blendsdk.github.io/year-planner/ (no custom domain)
Artifactdocs/.vitepress/dist (VitePress + /storybook/ + /api/)
DeploymentDocs site run 43428872169 — build and deploy succeeded
Live full demoCreate, reload persistence, context delete and reset all returned true in the browser check

The owner explicitly requested no full verify:full for this change and no custom domain. The Pages deployment is live.

Localization and lockstep versions (0.3.0) ​

The localization work added the @blendsdk/year-planner-languages package, the messages prop and 45 locale catalogues covering 41 left-to-right languages. Verification was targeted, as required by the project rule against unnecessary full runs:

CommandResult
pnpm --filter @blendsdk/year-planner-languages testPASS — 10 tests; 100% statements, branches, functions and lines
pnpm --filter @blendsdk/react-year-planner exec vitest run src/i18n.spec.test.tsxPASS — 7 formatting, integration and SSR tests
pnpm --filter @blendsdk/react-year-planner testPASS — full React suite after the message refactor
playwright test tests/e2e/i18n.spec.ts --project=chromiumPASS — German to Eastern Armenian to Japanese switch in the demo
pnpm typecheckPASS — core, React, languages, demo and Storybook
pnpm lintPASS — no errors
pnpm test:qualityPASS — 28 files, including release metadata, public API and story contracts
pnpm test:sizePASS — 46.35 / 60 kB React entry, 69.59 / 72 kB installed, 86.95 / 95 kB all languages, 2.22 / 8 kB largest language
pnpm test:versionsPASS — all seven workspaces at 0.3.0

The full verify:full gate was not run locally for this feature; CI keeps the authoritative full gate. The catalogue completeness test fails when a locale misses a key, a placeholder or a plural category.

Browser test selection and sharding (0.3.0) ​

The browser suite is grouped into ten feature groups, selected from the test map, and packed into five independent CI shards. A local ledger reuses a passing spec and browser pair only when its complete input fingerprint is unchanged. See Browser tests.

CommandResult
node scripts/browser-shards.mjs plan --gate releasePASS — 38 units, 5 shards; largest estimate 230 s
node scripts/browser-shards.mjs plan --gate task --paths packages/react/src/locale.tsPASS — 1 group, 6 units, 4 shards
node scripts/browser-shards.mjs plan --gate task --paths docs/faq.mdPASS — mode none, no browser work
node scripts/browser-shards.mjs plan --gate task --paths package.jsonPASS — mode full, every browser group
node scripts/browser-shards.mjs plan --gate task --paths new-file.txtPASS — fail-closed full selection for an unknown path
pnpm test:e2e:affected --paths packages/react/src/locale.tsPASS — 6/6 units, Chromium, Firefox and WebKit
repeat test:e2e:affectedPASS — 6 cached, 0 executed
node scripts/browser-shards.mjs run --shard chromium-2 --ciPASS — static artifacts served by playwright.ci.config.ts
pnpm test:qualityPASS — 255 contracts, including 33 browser-plan checks

Shard verification ​

All five planned shards were executed against the built demo, Storybook and docs artifacts on isolated local ports:

ShardPassedFailedSkippedDuration
chromium-13600147 s
chromium-24401124 s
firefox39016127 s
webkit39016104 s
mobile3007 s

The five jobs pass independently, so the CI matrix can run them concurrently. The full matrix is about 2.5 minutes of wall-clock time instead of about 16 minutes in one job.

First npm publication (0.3.0 bootstrap) ​

The three packages were published to npm on 2026-11-13. This was the one-time bootstrap: npm cannot generate provenance for a local publish, so 0.3.0 has no attestation. The registry now carries:

PackageVersionIntegrity
@blendsdk/year-planner-core0.3.0sha512-ZXNtZCm74kYDfca4CyIXdlM8c1/YLDEF+NMfh3eV4ejL4sW8Fn1CBo3J0Betgvahwkb/asHItvPMM4qAP3EGPQ==
@blendsdk/react-year-planner0.3.0sha512-YnAzsqJpgKgVXJg5yxMyGEKyUohxTDeSBZ3V6V1M90OCEIGRqHCTBu2HCSmFyuCiz0/bEsKeZqDA08vEOftHQA==
@blendsdk/year-planner-languages0.3.0sha512-5h+YFkPc2icb2UwPARHW6QGH3lCjOqhx8JOLr1l+syX3IFnKHvJHbUE9P5nC1k4VwLo6MrDWTWaMrZM15CDnLA==
CheckResult
pnpm release:smoke (registry)PASS — clean install, typecheck, CSS import and production build of all three exact versions
Trusted publishersConfigured for all three packages: release.yml, repository blendsdk/year-planner, environment npm-production, publish and stage-publish permissions
npm-production environmentCreated with reviewer blendsdk and a main-only deployment policy
Local npm cacheA stale negative cache briefly reported core as 404; a clean-cache install and the registry smoke passed

0.3.1 is the first provenance-signed release and validates the trusted publisher configuration.

0.3.1 provenance release ​

Release workflow run 43523190129 completed on 2026-11-13: verify, build, publish, registry smoke and documentation dispatch all passed.

CheckResult
npm versions0.3.1 at latest for all three packages
Provenancedist.attestations.provenance.predicateType is https://slsa.dev/provenance/v1 for core, React and languages
Registry smokepnpm release:smoke passed against the exact 0.3.1 versions
GitHub releasev0.3.1 published with all three package links
DocumentationPages deployment succeeded; the live site states 0.3.1

Free for internal use. Built with VitePress, Storybook and TypeDoc.