Skip to content

CHANGELOG.md

Latest commit

 

History

History
596 lines (578 loc) · 37.5 KB

CHANGELOG.md

File metadata and controls

596 lines (578 loc) · 37.5 KB

Changelog

All notable changes to Reactor will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning once a 1.0.0 release is cut. While the project is pre-1.0 and labeled experimental, the public API surface may change between releases without notice.

[Unreleased]

Changed (breaking)

  • .Margin(double, double) and .Padding(double, double) parameter order swapped to match CSS shorthand convention. Was (horizontal, vertical); now (vertical, horizontal). This aligns with CSS — padding: 16px 14px; means top/bottom = 16, left/right = 14, vertical first. Any existing positional 2-arg call site in the repo has been migrated to the named-arg form (.Margin(horizontal: 16, vertical: 8)) which preserves layout regardless of parameter order; recommend the same for external callers. Pre-1.0 breaking change is intentional — the original ordering was a layout-rotation footgun for agents and humans with CSS muscle memory. (spec 038 §3 — feedback from 525-run corpus / WPF-vs-CSS mental model)

Added

  • mur check --trace <path> — append one JSONL row per parsed diagnostic to <path> (in addition to stdout) for offline mining. Schema: {ts, code, severity, file, line, col, msg, receiver_type?, member?, mode}. Source code text is never written; absolute paths outside the project root are redacted to <external>. (spec 038 §0.3)
  • Tier-2 Roslyn semantic suggester for mur check. Covers CS1061, CS0103, CS0117, CS1503, CS7036 against Microsoft.UI.Reactor.* symbols; emits → try: <text> // [<evidence>] on the diagnostic line above the per-code confidence threshold (default 0.75). Tier-1 analyzer-ID hints still win ties. (spec 038 §5, §1.1–§1.6)
  • Per-code emit thresholds for the Tier-2 SymbolSuggester (src/Reactor.Cli/Check/Suggesters/Thresholds.cs) calibrated against the spec-037 50-run corpus. CS1061 raised to 0.80 (the structural-rewrite fixes in the corpus would otherwise risk false positives); CS0103 / CS0117 / CS1503 / CS7036 held at 0.75 default. Tuning harness lives in tests/Reactor.Tests/CheckCommandTests/Tuning/; first run snapshot at docs/specs/tasks/038-tuning-reports/2026-05-10-50run.md. (spec 038 §1.8, Data Checkpoint B)
  • EC1 5×N eval (2026-05-10): reactor-kanban-mur-check beats baseline on cost mean (−24%), cost median (−33%), and wall-time variance (CV 24% vs 81%); paired analysis wins 4 of 5 rounds. reactor-calc-mur-check regresses (+21% cost) because the suggester's per-invocation overhead (~5–8s) does not amortize on ~150 LoC projects with no API exploration surface to skip. Finding captured as a new spec 038 §11 risk + §14 open question on a project-size / diagnostic-count gate; merge to main pending product decision on path. No code change in this entry — eval result + spec doc updates only.
  • MUR_TELEMETRY=1 opt-in: appends (code, suggester, confidence, evidence_short) per emitted suggestion to ~/.mur/telemetry/<yyyy-mm-dd>.jsonl. Local-first, scoped to the active project; no source code, file paths, or machine identifiers logged. (spec 038 §10, §1.7)
  • mur check --suggest-threshold <N> — gate Tier-2 suggestions by per-invocation unique CS-prefixed diagnostic count. Default 3, set 0 to always emit. Resolution of the EC1 calc-vs-kanban split: small builds (1–2 errors) skip the ~5–8 s Tier-2 setup the agent doesn't need; larger structural failures still get suggestions. Counts the same dedup key EmitDiagnostics uses. (spec 038 §11 risk row, §14 #8)
  • Data Checkpoint C (spec 038 / spec 037): 525-pair mining corpus mirrored into docs/specs/tasks/038-tuning-reports/2026-05-11-525run-source/ (1,027 fixes / 1,233 ranker rows / 104 clusters from gpt-5.5). Analysis in 2026-05-11-525run.md. Cross-agent reproducibility bar still open — a second-agent drop is required before Phase-3 rule PRs. Top Phase-3 targets surfaced: CS0117/Theme *Background → SolidBackground, CS1061/*Element WinUI-name → Reactor-shortcut family, CS1955/GridSize missing-parens-on-factory. Tier-2 per-code thresholds held at current values; gate threshold (3) empirically defensible at 28.7% emit rate. No code change in this entry — calibration + docs only. (spec 038 §1.8, Data Checkpoint C)
  • mur check Phase 2 — MSBuild passthrough + deterministic pre-emit ranker. mur check [<path>] [mur-flags] [-- <msbuild-args>] — anything after a bare -- is forwarded verbatim to dotnet build. mur injects --nologo, -v:m, and -p:Platform={host arch} only if the same flag is not named in the passthrough section (detection by flag name, not value). When --trace is on, the trace records the effective dotnet build argv as a kind: "command" header row so replays are bit-faithful. New mode flags: --strict (promote warnings to errors), --final (emit every diagnostic — pre-merge sweep), --quiet (errors only). --emit-threshold <float> overrides the per-mode ranker default (0.6 iteration / 0.0 final). Pre-emit ranker (src/Reactor.Cli/Check/Ranker/PolicyTable.cs) suppresses noise mid-iteration (CS1591, CS0168, IDE0xxx, NU1701/NU1605, MSB3245/MSB3270/MSB3277, CS8600–CS8625 nullable warnings) while always emitting errors. (spec 038 §8, Phase 2.1–2.3)
  • tools/Reactor.MurCheckGuardrail — offline guardrail that audits a pair of --trace files (one iteration, one --final) against PolicyTable's universal-error floor invariant. Fails CI if a future policy-table edit would let a real build error get suppressed mid-iteration. The "universal floor" rule (Error severity always scores 1.0 regardless of code family) makes the invariant hold by construction today; the guardrail is the regression test that catches accidental violations. (spec 038 §8 Phase 2.4)
  • plugins/reactor/skills/reactor-build-and-check/SKILL.md updated for the iteration / --final workflow. EC2 measured 0/10 production value on the strong "explicit done gate" framing across 6 variant runs, so the framing was softened post-batch: --final is now documented as an optional pre-merge sweep (for human review / CI ship-readiness gates), explicitly NOT a task-completion requirement. SKILL anchor wording: "When mur check exits 0, you are done." Same wording in the legacy root SKILL.md. (spec 038 §8 Phase 2.5)
  • Phase-2.x — gate-input regression fix in CheckCommand.ShouldEmitSuggestions. The initial Phase-2 implementation counted the post-ranker emittable list when deciding whether to run the Tier-2 suggester. EC2 (3-round preview) measured Tier-2 firing collapse from EC1's 80% to 0% on kanban-mur because nullable warnings (CS8602/etc) were filtered out of the emittable list before the gate-count, closing the gate on builds EC1 had left open. Fixed by counting the full parsed diagnostics list — the gate measures build complexity, not stdout visibility. Regression test RankerTests.Suggest_gate_counts_full_parsed_list_not_post_ranker_emittable locks the behavior; fails the build if the bug is reintroduced. (spec 038 §14 #8)
  • Phase-2.x — EC2 5×N PASS by median (2026-05-11). reactor-calc-mur-check beats base on every metric (cost −5.1%, tokens −5.8%, turns −5.1%, wall −7.9%; variance 1.9× tighter). reactor-kanban-mur-check at cost median parity ($3.30 = $3.30); mean dragged to +5.7% by R2 outlier (n=5, R2-excluded mean is −3.3%). First-build OK 5/5 on both variant arms. --final invocation 0/10 across both projects (SKILL framing doing its job). Tier-2 firing 0/10 — gate correctly inhibits on small-batch iteration patterns; closing the kanban token gap is Phase-3's scope (rules > fuzzy match). Criterion-2 guardrail audit deferred to a harness retrofit (post-run mur check --final against the final workspace state to generate the iter+final trace pair the guardrail tool audits). Phase 2 cleared to merge to main. (spec 038 §1.8 EC2 acceptance, §8, §11)
  • Phase 3.1 / 3.1a — Tier-3 rule infrastructure scaffolded. New surface under src/Reactor.Cli/Check/Rules/: IRulePattern contract (Name, Provenance, DiagnosticCodes, DeclaredTargets, TryMatch), RuleContext + RuleSuggestion records, RuleRegistry (reflection discovery of IRulePattern implementations in Reactor.Cli, Default singleton, dedup on Name collisions, BestMatch with disable list and self-disable-on-unresolved-target reporting, Statuses for --list-rules), and RuleSymbolResolver (per-CSharpCompilation cached symbol lookup via ConditionalWeakTable — spec §3.1a's contract that rules never string-match MemberAccess.Name.ValueText). New CLI flags --disable-rule <Name> (repeatable, warns on unknown names) and --list-rules (short-circuits dotnet build, prints the name/provenance/status table, exits 0). SuggesterOrchestrator runs rules alongside Tier-2; spec §6 "rule wins over Tier-2 fuzzy match" preserved; rules can match diagnostic codes outside Tier-2's SupportedCodes so CS1955 / Theme-lookup rules are unblocked. tests/Reactor.Tests/CheckCommandTests/Rules/RuleTargetResolutionTests.cs is the §3.1a CI gate — instantiates every registered rule against a live Reactor Compilation (full assembly references, the inverse of TestCompilation.Create) and asserts every declared target resolves. Passes vacuously today; becomes load-bearing the moment the first rule lands. 35 new unit tests covering contract shape, registry discovery and edge cases (duplicates, throwing rules, self-disable), resolver cache identity, orchestrator rule-vs-Tier-2 precedence, and ArgsParser round-trip. Phase-3 rule PRs themselves remain blocked on the second-agent corpus drop (cross-agent reproducibility bar #2 of the Validation Gate). (spec 038 §3.1 + §3.1a)
  • Spec 038 Phase-3 vocab table at docs/specs/tasks/038-vocab-table.csv (§3.0 prerequisite for any Class-B rule PR). 20 rows covering WPF / Silverlight / WinUI 2 / WinUI 3 → Reactor vocabulary translations, seeded from the 525-run report's Phase-3 priority targets plus desk research against skills/reactor.api.txt. (spec 038 §3.0)
  • GridSizePxRenameRule (Class-A induced): CS0117 on Microsoft.UI.Reactor.GridSize where the missing member is Pixel, Pixels, or Fixed — the WPF / WinUI / legacy-XAML names — suggests GridSize.Px(...) with the same numeric argument. Cross-agent reproducibility STRONG: 5 events in gpt-5.5 + 4 events in sonnet-4.6 = 9 events combined, 100% rewrite target is Px(...) on every row. 5 unit tests (3 positive covering all three legacy names, 2 negative). Bar #5 (independent reviewer signoff) pending. (spec 038 §3.2, §6 Class A, Validation Gate bar #2)
  • TextBlockStyleHintRule (Class-A induced): CS1061 or CS0117 on Microsoft.UI.Reactor.Core.TextBlockElement where the missing-member name is Style — suggests Reactor's fluent text helpers (.FontSize, .Bold, .SemiBold, .Italic, .Foreground) directly on the element. Reactor doesn't expose a Style member; the WPF/WinUI mental model reaches for TextBlock.Style = SomeTitleStyle style-resource attachment. Cross-agent reproducibility STRONG-after-collapse: 2 events gpt-5.5 (fluent .Style(...) shape) + 3 events sonnet-4.6 (record with { Style = ... } shape) = 5 events combined; the rule covers BOTH syntactic shapes in one rewrite. 5 unit tests. (spec 038 §3.2, §6 Class A)
  • ThemeBackgroundSuffixRule (previously Class-B, promoted to Class-A by the cross-agent audit). Rule shape unchanged; the file-header comment now records the audit's bar #2 evidence (16+11=27 events across both corpora on the (CS0117, Theme, other) key) alongside the original vocab-table-citation justification. The Class-A / Class-B distinction is about evidence type, not rule shape; this rule was authored from the vocab table first and the corpus later confirmed it. (spec 038 §6 Class A re-classification)
  • Critical fix — CompilationLoader now resolves ProjectReference outputs. Without this, every Tier-3 rule self-disabled on real mur check invocations against Reactor apps: the loader only parsed project.assets.json's targets section (NuGet packages), so Reactor itself (a project reference for every sample app) was invisible. New code walks libraries.<id> entries with type=project, reads the referenced csproj's <AssemblyName> (or falls back to the basename), and locates the most-recently-built matching .dll under that project's bin/ subtree. Regression locked by CompilationLoaderTests.Resolves_ProjectReference_built_dll_from_project_assets_json. Unit tests passed before this fix because they use synthetic in-memory compilations; end-to-end smoke against samples/apps/wordpuzzle was what surfaced the silent failure mode. (spec 038 §5 + §6 + §3.1a)
  • Suggest-gate carve-out for Tier-3 rules. The gate (--suggest-threshold, default 3 unique CS-prefixed diagnostics) was wrapping the entire suggester block — when closed, neither Tier-2 nor rules ran. The gate exists to suppress Tier-2 fuzzy match's noise on small builds; rules are precision-anchored (Roslyn ISymbol binding) and shouldn't be subject to that calibration. SuggesterOrchestrator now takes a tier2Enabled bool; CheckCommand.Run always builds the orchestrator (when the compilation loads) and passes the gate result in. Tier-3 rules always run when their diagnostic code surfaces; Tier-2 stays gated. Two new tests lock this down. This is the EC2 watch-item ("Phase-3 rules are the right lever — not Phase-2.x gate tuning") finally addressed in code. (spec 038 §11 + §14 #8, EC2 watch-item)
  • First Class-A induced rule: GridSizeFactoryParensRule (CS1955 on Microsoft.UI.Reactor.GridSize.Auto() → suggest GridSize.Auto — i.e. drop the parens, since Auto is a static property and only Star(double) / Px(double) are methods). Cross-agent reproducibility STRONG: 146 events combined across the gpt-5.5 525-run (110 events) and claude-sonnet-4.6 525-run (36 events) corpora, top-frequency cluster in both at 10.7% / 9.8% of fixes respectively. Both corpora are unanimously about Auto — every captured row's diag.member field is exactly "Auto". First cross-tier rule: CS1955 is outside Tier-2's SupportedCodes, so the orchestrator's RulesCoverCode path is now load-bearing for at least one diagnostic code. 5 unit tests (3 positive fixtures from distinct cross-corpus run_ids, 2 negative — lookalike Acme.GridSize in a user namespace plus a synthetic non-CS1955 diag gate test). Validation Gate cleared on bars #1–#4 + #6; bar #5 (independent reviewer signoff) pending. Cross-agent audit recorded at docs/specs/tasks/038-tuning-reports/2026-05-11-cross-agent-audit.md. (spec 038 §3.2, §6 Class A, Validation Gate bar #2)
  • §3.1a per-rule performance bound test: RulePerformanceTests.BestMatch_median_under_per_rule_budget ([Trait("Category","Perf")]) asserts RuleRegistry.Default.BestMatch median across 1000 iters on the canonical CS1061-on-ButtonElement.OnClick fixture stays under 0.5 × rule_count × 4 ms (4× CI slack matches CompilationLoaderTests convention). Was deferred until the first rule landed; now load-bearing for the rule set. (spec 038 §3.1a)
  • Cross-agent reproducibility audit at docs/specs/tasks/038-tuning-reports/2026-05-11-cross-agent-audit.md comparing gpt-5.5 and claude-sonnet-4.6 525-run corpora on receiver-typed clusters. Verdicts: three Class-A targets STRONG (CS1955/GridSize/other 146 events; CS0117/Theme/other 27; CS0117/GridSize/renamed_member 7); two more strong after rule-design collapse (TextBlockElement member rename
    • TemplatedListViewElement generalized over <T>); one striking gpt-5.5-only signal (CS1955/GridElement, 29 events, zero in sonnet — deferred to a third corpus drop). Closes Data Checkpoint C's cross-agent-reproducibility gap. (spec 038 §3.0, Validation Gate bar #2)
  • First three Class-B vocabulary-translation rules: ThemeBackgroundSuffixRule (CS0117 on Theme with member ending in Background → Theme.SolidBackground, cluster C0019, 16 events); AlignmentShortcutRule (CS1061 on Reactor *Element receivers for HorizontalAlignment / VerticalAlignment → .HAlign(...) / .VAlign(...), cluster C0017 + adjacent ≈ 22 events); and ButtonOnClickFactoryMoveRule (CS1061 on ButtonElement.OnClick → Button(..., onClick: ...) factory named-arg, explicitly naming .OnTapped as the wrong sibling to keep agents from reaching for the gesture event). Both bind target types via RuleSymbolResolver (no string matching); the rule-target resolution CI gate is now load-bearing. 10 new unit tests (positive fixtures cite their source run_ids from the 525-run corpus; hand-authored extensions are tagged [Trait("Origin", "VocabHandAuthored")]). PRs remain blocked on Validation Gate bar #5 (independent reviewer signoff) — the artifacts are "ready for review", not "ready to merge". (spec 038 §3.2, §6 Class B)
  • .Margin(...) and .Padding(...) per-side overloads now default unspecified sides to 0.0. Enables agent-intuitive call shapes like .Margin(top: 12), .Padding(left: 8, right: 8) that previously failed to compile (CS7036: no matching overload). 525-run corpus shows 198 build failures from agents writing this exact shape against the prior all-required signature — far and away the highest-frequency failure-driver in the drop. Eliminating it is a single-line code edit per overload but a large agent-productivity unlock. Reactor.Tests adds CSS-ordering + per-side + positional-overload regression tests. (spec 038 §3 follow-up — surfaced during Phase-3 rule authoring)
  • Cheatsheet in plugins/reactor/skills/reactor-getting-started/SKILL.md now shows the named-arg Button("Save", onClick: handler) form alongside the positional one, with an explicit anti-pattern comment naming .OnClick(...) and .OnTapped(...) as the wrong fixes for click intent. The cheatsheet's .OnTapped((s, e) => ...) example is now anchored to non-Button surfaces (Border / Image / ScrollView) with a back-reference to the Controls section — the prior parenthetical Button carve-out was easy to miss mid-build. (spec 038 §3 — agent-facing skill updates)
  • Spec 038 EC3-final watch-item: rule_fired trace event. When a Tier-3 rule attaches a suggestion to a diagnostic, mur check --trace now writes one structured row per fire: {kind: "rule_fired", rule, code, confidence, evidence, file, line, mode}. Per-rule firing-rate audits collapse from multi-step content scans against events.jsonl agent tool outputs to a 1-line jq over the trace file. Tier-2 suggestions deliberately do not emit this row — Tier-2 firing rates are visible via the opt-in MUR_TELEMETRY=1 channel. (spec 038 §0.3, EC3-final watch-item)
  • Spec 038 §3.1a residual: trace-channel structured warning hook for self-disabled rules. TraceWriter.WriteRuleSelfDisabled(rule, target) emits {kind: "rule_self_disabled", rule, unresolved_target, mode}. SuggesterOrchestrator threads an optional onRuleSelfDisabled callback through to RuleRegistry.BestMatch; CheckCommand.Run wires it to the active trace writer when --trace <path> is set, dedup'd per-invocation per-rule. Stdout stays clean — agents don't read trace files, but maintainers see "rule X disabled because target Y didn't resolve" the moment a Reactor minor release breaks something. (spec 038 §3.1a)
  • EC1 re-run with the diagnostic-count gate (2026-05-11): both arms PASS. reactor-calc-mur-check cost −4% mean (was +21% in the prior batch); reactor-kanban-mur-check cost −33% mean / −39% median (was −24% mean — preserved and grew). First-build OK 5/5 both variant arms. Phase 1 acceptance bar met; Phase 1 cleared to merge to main. Watch-item carried into Phase 2: kanban CV widened (24% prior → 54%) because one of five runs hit 0 firings and took the long-tail base path — gate behavior is path-dependent on the agent's exploration order. Below the resolution threshold for a Phase-1 blocker; Phase 2 telemetry should track per-run firing counts. (spec 038 §1.8 EC1 acceptance, §11 risk row, §14 #8)
  • WindowSpec, ReactorWindow, WindowKey, WindowStartPosition, PresenterKind, WindowState, WindowIcon, WindowDipSizeChangedEventArgs, WindowClosingEventArgs, ReactorAppContext — first-class Window primitive promoted out of internal hosting wiring. ReactorApp.Run(Action<ReactorAppContext>) is the new multi-window startup surface; the existing Run<TRoot> overload is preserved as a thin wrapper. (spec 036 §3, §4)
  • ReactorApp.OpenWindow, Windows, PrimaryWindow, FindWindow, WindowOpened / WindowClosed, Exit, ShutdownPolicy, UIDispatcher — process-wide window topology. (spec 036 §4.3, §6)
  • Per-window DPI awareness — ReactorWindow.Dpi, DipScale, DpiChanged; WindowMessageMonitor (SetWindowSubclass) for WM_DPICHANGED and WM_GETMINMAXINFO; DIP→physical conversion in initial size, SetSize, SetPosition. Min/max constraints flow through WM_GETMINMAXINFO so dragging across a DPI boundary respects spec'd minimums. (spec 036 §5)
  • RenderContext.UseDpi(), parameterless UseWindowSize(), UseBreakpoint(double). (spec 036 §5.2)
  • ReactorWindow.Activated, Deactivated, SizeChanged, StateChanged, Closing, Closed events with UI-thread synchronous dispatch. Closing runs UseClosingGuard predicates first then subscribers; any false cancels. (spec 036 §6.3, §7)
  • RenderContext.UseWindow(), UseWindowState(), UseIsActive(), UseClosingGuard(Func<bool>). Tray-flyout fallback semantics match spec §7.1 (null/Normal/true/no-op). (spec 036 §7)
  • RenderContext.UseOpenWindow(WindowKey, WindowSpec, Func<Component>)
    • Component.UseOpenWindow mirror — open or reuse a secondary window keyed by WindowKey. Identity-stable across re-renders; spec changes flow through ReactorWindow.Update; parent unmount does not auto-close the child. (spec 036 §4.3 / §15.6)
  • ReactorWindow.PersistedScope — per-window Core.WindowPersistedScope, disposed when the window closes. RenderContext.UsePersisted(_, _, PersistedScope.Window) now resolves to this per-window store, so two windows of the same component class hold independent persisted state. (spec 036 §3.4 / §4.4 — closes spec 033 §7.5.)
  • ShutdownPolicy.OnPrimaryWindowClosed exits when the primary window closes (not just when the snapshot empties); OnLastSurfaceClosed considers tray icons (Phase 8 fills the registry). The default zero-window startup-callback path now exits under OnLastSurfaceClosed too when no tray icons were opened. (spec 036 §6.2)
  • IWindowPersistenceStore, PackagedSettingsStore, JsonFileStore, and ReactorApp.WindowPersistenceStore — pluggable per-window placement persistence. Default auto-detect picks the WinRT settings store for packaged apps and a hand-rolled, AOT-safe JSON file store (1 MB cap, atomic write-then-rename, base64-per-id) for unpackaged apps. WindowSpec.PersistenceId opts in; placement saves on close and restores on first show via WindowPlacementCodec with a monitor- layout fingerprint borrowed from WinUIEx.WindowManager. (spec 036 §8)
  • WindowSpec.Backdrop is now seeded as a window-level default through BackdropApplier.SetWindowDefault, so the first frame paints the declared material even when the root component tree carries no BackdropChoice modifier. Tree-level modifiers still win on subsequent renders. (spec 036 §3.3)
  • Owned-window relationship via WindowSpec.Owner — applies the Win32 GWLP_HWNDPARENT slot at construction time and force-hides the owned window from the taskbar / Alt-Tab. Owner-close cascades to owned children with WindowCloseReason.OwnerClosed; if any owned guard cancels, the owner-close cancels too. (spec 036 §9)
  • ReactorWindow.Progress (TaskbarProgress, with TaskbarProgressState enum: None / Indeterminate / Normal / Paused / Error) and ReactorWindow.Overlay (TaskbarOverlay with Icon / AccessibleDescription). Both lazy-initialize the ITaskbarList3 COM wrapper through TaskbarComSingleton so apps that never touch the shell surface pay no startup cost. (spec 036 §11.1 / §11.2)
  • ReactorWindow.SetThumbnailToolbar(IReadOnlyList<ThumbnailToolbarButton>) / ClearThumbnailToolbar() — up to seven buttons; first call uses ThumbBarAddButtons, later calls use ThumbBarUpdateButtons. Validation rejects > 7, duplicate Ids, empty Ids, null OnClick. Click dispatch hooks WM_COMMAND in WindowMessageMonitor. HICONs are released on ReactorWindow.Dispose. (spec 036 §11.5)
  • JumpList, JumpListItem, JumpListItemKind — process-scoped jump list. Packaged path uses Windows.UI.StartScreen.JumpList; unpackaged falls back to a hand-rolled ICustomDestinationList wrapper (JumpListComInterop) gated by runtime Package.Current detection through the new PackageRuntime helper. AppUserModelId, ShowRecent, ShowFrequent are settable. JumpListItem.ForUri(...) factory is the recommended way to build entries — pairs with LaunchActivation.TryResolve<TRoute>(map) for the navigation handoff. (spec 036 §11.3 / §11.6)
  • LaunchActivation parsing — OnLaunched now reads Microsoft.Windows.AppLifecycle.AppInstance.GetActivatedEventArgs for File / Protocol / Toast activations and falls back to the WinUI LaunchActivatedEventArgs.Arguments + Environment.GetCommandLineArgs for jump-list / tray re-launches. LaunchActivation.TryResolve<TRoute> bridges the launch argument string into the existing DeepLinkMap<TRoute> so jump-list / tray entries become a one-liner navigation handoff. (spec 036 §11.6, implementation-time addition)
  • ReactorTrayIcon + TrayIconSpec — system-tray icon as a peer of ReactorWindow. ReactorApp.OpenTrayIcon, TrayIcons snapshot, FindTrayIcon, TrayIconOpened / TrayIconClosed events; mirrored on ReactorAppContext. Hidden message-only window (TrayHiddenWindow) routes Shell_NotifyIcon callbacks back to the UI thread under NOTIFYICON_VERSION_4 semantics. Click, DoubleClick, RightClick events fire on the UI thread. Update(spec) diffs icon / tooltip / visibility; Close / Dispose removes the icon and unregisters from ReactorApp.TrayIcons. OnLastSurfaceClosed now reads the real TrayIconCount and re-evaluates on tray close so a tray-only app exits cleanly when the final icon goes away. (spec 036 §11.4)
  • RenderContext.UseTrayIcon(TrayIconSpec) + Component.UseTrayIcon mirror — opens (or reuses by key) a tray icon scoped to the calling component. The trailing UseEffect cleanup closes the icon on unmount; spec changes flow through Update via a record-keyed UseEffect. (spec 036 §11.4)
  • Seven live-shell selftest fixtures under tests/Reactor.AppTests.Host/SelfTest/Fixtures/WindowModelFixtures.cs: WindowModel_LifecycleEvents, _ClosingEventCancels, _TaskbarProgressLiveCom, _ThumbnailToolbarLiveCom, _PersistedScopeIsolated, _TrayIconRoundTrip, _UseOpenWindowReusesByKey. They exercise the public surface against real HWND / ITaskbarList3 / Shell_NotifyIcon COM, opening secondary ReactorWindows through ReactorApp.OpenWindow and cleaning up under ShutdownPolicy.Explicit so they don't kill the host harness. 33/33 assertions pass alongside the full 2314-assert selftest matrix. (spec 036 §0.5 / §0.6 / §11)
  • Devtools windows.list / windows.activate / windows.close / windows.open MCP tools (spec 036 §10). windows.list returns id, key, title, DIP size, DPI, state, isMain — driven by a new WindowRegistry.Attach(ReactorWindow, ...) overload that retains the back-reference. windows.open is gated by the same component allowlist as switchComponent so loopback callers can't spawn arbitrary types; windows.close honors UseClosingGuard and surfaces cancelled: true instead of hanging. The devtools WindowRegistry is now driven from ReactorApp.WindowOpened / WindowClosed events so secondary windows opened via OpenWindow are tracked too. CLI and skills/devtools.md plumbed.
  • Microsoft.UI.Reactor.Hooks.UseMemoCells / UseMemoCellsByKey / UseMemoCellsByIndex — cell-level memoization hooks (extension methods on RenderContext, plus matching Component shims) for high-frequency list/grid bodies. Cells whose item value (and declared deps) haven't changed since the previous render are reused by reference; the reconciler short-circuits on ReferenceEquals and skips diffing entirely. (spec 034 §C)
  • REACTOR_HOOKS_007 analyzer + codefix — warns when a UseMemoCells builder lambda closes over a value that isn't declared in the params deps list, which would silently render stale. The codefix appends the missing capture to the deps slot. Indirect captures through helper methods are a documented blind spot. (spec 034 §C)
  • "Memoizing list cells" section in docs/guide/advanced.md covering the three overloads, when each is the right hammer, the gen2 trade-off, and the analyzer-as-safety-net story. (spec 034 §C)
  • tests/stress_perf/StressPerf.ReactorOptimized — sibling bench variant that demonstrates the spec-034 §B direct-record-initializer idiom for inner-loop cell construction. The naive StressPerf.Reactor variant stays unchanged and remains the framework-level baseline; the new optimized sibling is the reference implementation of the perf-tips skill. Wired into run_stocks_grid_baseline.ps1, run_bench_aot_publish.sh, run_benchmark.sh, and run_sweep_arm64.ps1. (spec 034 §B)
  • "Hot loops" section in docs/guide/advanced.md documenting when to reach for direct record initializers, the trade-offs vs the fluent chain, and a side-by-side worked example. Source template at docs/_pipeline/templates/advanced.md.dt. (spec 034 §B)
  • Expr(Func<Element?>) factory in Microsoft.UI.Reactor.Factories for inline block-expression bodies inside a DSL tree, removing the ((Func<Element?>)(() => …))() cast ceremony. Pure composition — no hooks, no memoization, no reconciler boundary. (spec 033 §5)
  • IPersistedStateScope interface, PersistedScope enum (Window / Application), ApplicationPersistedScope (process-wide singleton at ApplicationPersistedScope.Default, capacity 4096), and WindowPersistedScope (per-host instance, capacity 1024). All backed by an internal LruCache<TKey,TValue>. New RenderContext.UsePersisted<T>(key, initial, PersistedScope) overload makes the scope explicit. (spec 033 §2)
  • Microsoft.UI.Reactor.Factories.RenderEachTime(Func<RenderContext, Element>) — explicit factory for "inline component with own hooks that re-renders every parent render". Replaces the soft-deprecated Func(...) for the rare cases that genuinely want always-re-render semantics. (spec 033 §4)
  • Microsoft.UI.Reactor.GridSize value type with Auto / Star(weight) / Px(pixels) smart constructors, implicit conversion to Microsoft.UI.Xaml.GridLength, and a strict invariant-culture string parser (Parse). New typed Grid(GridSize[], GridSize[], …) factory overload. (spec 033 §1)
  • samples/InteropFirst — XAML-window-hosts-Reactor demonstration with shared ObservableCollection<Order>, shared ICommands bridged through CommandInterop.FromCommand, and shared App.xaml brush resources flowing through props into a Reactor Component<TProps>. (spec 033 §7)
  • BackdropKind enum and .Backdrop(BackdropKind) / .Backdrop(Func<SystemBackdrop?>) modifier on the root tree for declarative Mica / Acrylic on Reactor-hosted windows. ReactorHost applies the modifier at the end of each reconcile pass and resets the window's backdrop on dispose; ReactorHostControl that does not own its window no-ops with a one-shot debug log. (spec 033 §6)
  • ElementRef<T> typed-ref wrapper (Microsoft.UI.Reactor.Input), UseElementRef<T>() hook (Microsoft.UI.Reactor.Hooks), and a strongly-typed .Ref<T,TElement>(...) modifier overload. The typed surface removes the (Button)ref.Current cast at consumers and adds a DEBUG-only assertion when a typed ref is bound to an element of the wrong concrete type. AOT-safe and reflection-free at the public surface. (spec 033 §3)
  • Component.UsePersisted<T>(key, initial, PersistedScope) three-arg overload so component subclasses can declare the persisted-state scope (Window vs Application) explicitly at the call site, matching the RenderContext.UsePersisted overload added earlier. (spec 033 §2)

Changed

  • Spec 034 — Element allocation reduction. Three independent allocation cuts in one PR: bucketed ElementModifiers (transparent storage shim, ~−11% bytes/tick on the 4,900-cell stress grid), direct-record-initializer idiom for inner cell loops (~−60% bytes per cell), and UseMemoCells cell-level memoization. Verified at PR-close on ARM64 Release with full ETW Present-tracking across 10/20/50/100% mutation, all eight stress_perf variants: ReactorOptimized at 10% mutation reaches 17.1 Effective Refresh/s — within noise of DirectX (17.2) and Wpf (17.9), and +66% over naive Reactor (10.3). Reconcile-time win on the same A/B: −76% at 10% (32.5 ms → 7.9 ms), −61% at 20%, −31% at 50%, −12% at 100% — memo's win tracks the partial-reuse opportunity exactly as predicted. DirectX runs away at saturation (50%+) — no allocating framework can keep up there. Component A in isolation (naive Reactor pre-shim vs post-shim, same source, no app-code changes) shows renders/sec within run-to-run noise at 20/50/100% — its win is allocation-side, not renders-side, on this hardware. See docs/specs/034-element-allocation-reduction.md § "Verified close-out — 2026-05-03" for the full eight-variant matrix and reads. (spec 034)
  • ElementModifiers now stores layout and visual fields in LayoutModifiers / VisualModifiers sub-records. Existing call sites are unaffected — public properties (Padding, Margin, Foreground, Background, …) shim through to the appropriate bucket on read and write. Perf-critical inner loops may construct buckets directly via the new Layout = … / Visual = … initializer slots to avoid a fat ElementModifiers clone per fluent step. (spec 034 §A)
  • PersistedStateCache rewritten over an LRU cache with eviction-on-full semantics. The previous "refuse new keys when 4096 entries are present" policy is replaced — later, hotter keys are no longer starved by the first 4096 keys ever recorded. Application-scope registers an Windows.System.MemoryManager.AppMemoryUsageIncreased handler and trims to 25% of capacity when the OS reports OverLimit / High. Best-effort: hosting models that do not expose the event log a notice and carry on. Key validation now requires non-empty keys ≤ 256 chars. (spec 033 §2)
  • GridDefinition gains a strongly-typed constructor accepting GridSize[] for columns and rows. The legacy string-array constructor is preserved for backward compatibility. (spec 033 §1)
  • ApplicationPersistedScope and WindowPersistedScope now emit one-line Debug.WriteLine diagnostics on construction, disposal, and (for the application scope) memory-pressure trim. Logs only counts and capacity — never keys or values, since keys may be derived from user-controlled identifiers in apps. (spec 033 §7.10)
  • samples/Reactor.TestApp/Demos/PersistedDemo, NavigationDemo, and samples/apps/regedit migrated to the explicit UsePersisted(key, initial, PersistedScope.Window) overload to document per-window intent at the call site. (spec 033 §2)

Deprecated

  • Microsoft.UI.Reactor.Factories.Grid(string[], string[], params Element?[]) is marked [Obsolete]. Use the strongly-typed Grid(GridSize[], GridSize[], params Element?[]) overload with GridSize.Auto / GridSize.Star(weight) / GridSize.Px(pixels) instead. Slated for removal in the next minor release. (spec 033 §1)
  • Microsoft.UI.Reactor.Factories.Func(Func<RenderContext, Element>) is marked [Obsolete]. Replace with Memo(ctx => …) (render once + state changes) or RenderEachTime(ctx => …) (always re-render). Slated for removal in the next minor release. (spec 033 §4)

Removed

  • ReactorHost.MainDispatcherQueue (internal static, first-host-wins capture). Cross-thread setState marshalling and AutoSuggest's RaiseStateChanged now route through ReactorApp.UIDispatcher. ReactorHost ctor seeds UIDispatcher for embedded ReactorHostControl scenarios that bypass ReactorApp.Run. (spec 036 §4.3)

Fixed

Security