Skip to content

045-docking-p2-review-checklist.md

Latest commit

 

History

History
136 lines (119 loc) · 6.36 KB

045-docking-p2-review-checklist.md

File metadata and controls

136 lines (119 loc) · 6.36 KB

Spec 045 Phase 2 — Human Review Gate

Use this checklist during the §2.29 review pass. It reuses the §4.7 P1 script (items 1–8) plus the ten new P2 items.

How to drive:

  1. Build the showcase: 
    dotnet build samples/apps/dock-showcase/DockShowcase.csproj -c Debug -p:Platform=x64
  2. Run it: 
    samples/apps/dock-showcase/bin/x64/Debug/net10.0-windows10.0.22621.0/DockShowcase.exe
  3. NOTE: the §2.19 unhooking removed the REACTOR_DOCK_XAML=1 A/B flip — the native renderer is the only path in this build. The Phase-1 wrapper is unavailable for side-by-side comparison.

Mark each item ✅ / ❌ / ➖ (N/A or deferred). Sign at the bottom.

Phase 1 script (re-run against P2)

  • Item 1 — Drag a center tab to each of 5 split targets (center, SplitL/T/R/B). Verify preview match, drop landing, Esc snap-back. (Native path; native renderer is the showcase default.)
  • Item 2 — Drag to each of 4 edge dock targets. Same checklist. (Native path.)
  • Item 3 — Drag a tab out of the title bar into open space; floating window appears at pointer; title bar matches adapter content. (Native renderer opens the floating window via ReactorApp.OpenWindow; the custom title bar adapter routing is §2.6 follow-up, so the fallback title is the pane's Title or DockingStrings.FloatingWindowDefaultTitle.)
  • Item 4 — Drag a floating tab back into a tab group; floating window auto-closes if last document. (Cross-host re-dock is a follow-up; tear-out + open is the proven path.)
  • Item 5 — Resize splits via splitter; min sizes respected; re-mount restores sizes. (§2.1 splitter fix from session 2 covers this; M15/M17/M19 selftests are green.)
  • Item 6 — Pin a tab to a side; click side icon; popup shows; resize popup; re-pin from popup; close from popup. (§2.5 popup open/collapse works; light-dismiss + sizer are §2.5 follow-ups.)
  • Item 7 — Save layout to JSON, quit, restart, load; layout matches. (§2.7 v2 JSON; round-trip + invariant-culture covered by unit tests.)
  • Item 8 — (negative) tear out a tab while resizing a different split; no crash. (Drag session payload is in-memory only; no cross-state interaction.)

P2-specific items

  • Item 9 — Documents vs tool windows visual distinction matches intent. (All-ToolWindow groups auto-flip to bottom-position + compact tabs via §2.8 default resolution; verify in showcase Scene A.)
  • Item 10 — Per-pane content state survives save→quit→restart →load (e.g. editor scroll position). *(Document<TState> envelope
    • JSON round-trip; full WindowPersistedScope wiring is open per §2.9. Verify shape preservation; per-pane state typed envelope round-trip via composition.)*
  • Item 11 — Ctrl+Tab pane navigator opens, navigates, closes correctly. (Deferred — needs the navigator overlay primitive. Ctrl+PageUp/PageDown navigation through tabs is wired and live (§2.10 chord set). Mark as N/A for P2 if the navigator itself isn't present.)
  • Item 12 — Layout JSON v1 file (P1 build) loads correctly in P2 build. (§2.11 migration ladder covers this; unit-tested in LayoutMigrationTests.)
  • Item 13 — Drop preview latency feels equivalent (timed where reasonable; subjective otherwise). (Hot-path ComputePreviewBounds is allocation-free per §2.20 perf budget tests; visual smoothness should match P1.)
  • Item 14 — AOT-published binary runs the showcase end-to-end. (Open — verify via dotnet publish -p:PublishAot=true of the showcase. The docking subsystem is AOT-clean by construction — JsonSerializerContext source-gen, no reflection — but the end-to-end AOT publish hasn't been driven this session.)
  • Item 15 — Run under de-DE and ar-SA (RTL); titles localize; drop targets / context-menu items localize; layout mirrors; pointer hit-tests resolve in mirrored regions. (§2.21 localization routing landed via DockingStrings.Resolver; apps wire their IntlAccessor to translate. §2.23 FlowDirection inheritance from ancestor handles the bulk; explicit drop-target glyph mirror + splitter direction inversion are §2.23 follow-ups.)
  • Item 16 — Screen reader pass (Narrator/NVDA): pane roles announced; AutomationIds stable; focus never lost; drop-target navigation keyboard-only with arrow+Enter. (§2.22: per-pane AutomationId = pane:<key> + AutomationProperties.Name from Title; host has Custom landmark + localized name. Live-region announcements are §2.22 follow-up.)
  • Item 17 — Reduced-motion: transitions disappear; static positioning correct. (No animation is wired today — reduced-motion is the default by omission. The animation pass adds the UISettings.AnimationsEnabled gate when transitions are added.)
  • Item 18 — Corrupt layout recovery: hand-edit JSON to invalid; app starts with default; error event logged; no crash dialog. (§2.7 / §2.25: corrupt JSON falls back via DockLayoutSerializer.Load returning IsFallback=true + a ReactorEventSource.DockingLayoutLoadFallback event with PII-safe category; selftest NativeDocking_Reliability_CorruptLayoutFallback_HostMounted is green.)

Sign-off

  • Reviewer: _____________________________
  • Date: _____________________________
  • PR: _____________________________
  • Net result: _____________________________
  • Followups filed (issue / PR):

Selftest baseline before review

Run before the visual pass to capture the regression-test baseline:

dotnet test tests/Reactor.Tests/Reactor.Tests.csproj -p:Platform=x64 \
  --filter "FullyQualifiedName~Docking" --no-build

"tests/Reactor.AppTests.Host/bin/x64/Debug/net10.0-windows10.0.22621.0/Reactor.AppTests.Host.exe" \
  --self-test --filter NativeDocking

Known flakes (per the spec-045-next-agent-prompt.md handoff):

  • NativeDocking_SplitterProgrammaticVisualDemo_TIMEOUT — visual demo with intentional Task.Delays.
  • M07_* / M08_* / M03_BodyBReachable / DocsByComposition_* / Reliability_Effect_BodyRendered / Sim_OverlayFound — FindText/ timing flakes that fire only in the full-filter run.

If any other selftest fails, treat as a regression and block sign-off.