Fixed & documented, not-just-marketed:
- HomeScreen now derives summary card from LocalDataManager.myResidences
with VM fallback — populated PNG genuinely differs from empty.
- DocumentsScreen added same LocalDataManager fallback pattern + ambient
subscription check (bypass SubscriptionHelper's singleton gate).
- ScreenshotTests.setUp seeds the global DataManager singleton from the
fixture per variant (subscription/user/residences/tasks/docs/contractors/
lookups). Unblocks screens that bypass LocalDataManager.
Honest coverage after all fixes: 10/34 surface-pairs genuinely differ
(home, profile, residences, contractors, all_tasks, task_templates_browser
in dark mode, etc.). The other 24 remain identical because their VMs
independently track state via APILayer.getXxx() calls that fail in
Robolectric — VM state stays Idle/Error, so gated "populated" branches
never render.
Root architectural fix needed (not landed here): every VM's xxxState
should mirror DataManager.xxx reactively instead of tracking API results
independently. That's a ~20-VM refactor tracked as follow-up in
docs/parity-gallery.md "Known limitations".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previous run left edit_document at 0/4 because the record task hadn't
recorded it; the other 39 surfaces' goldens were optimized in-place by
zopflipng (no visual change). Gallery HTML/markdown regenerated to
reflect 160 Android goldens (40 surfaces × 4 variants).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Late writes from the previous recordRoborazziDebug pass. Brings Android
coverage from 17 → 21 surfaces.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
scripts/build_parity_gallery.py walks both golden directories and pairs
Android↔iOS PNGs by filename convention into docs/parity-gallery.html —
a self-contained HTML file with relative <img> paths that renders
directly from gitea's raw-file view (no server needed).
Current output: 34 screens × 71 Android + 58 iOS images, grouped per
screen with sticky headers and per-screen anchor nav.
docs/parity-gallery.md: full workflow guide — verify vs record, adding
screens to both platforms, approving intentional drift, tool install,
size budget, known limitations.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the synthetic theme-showcase ScreenshotTests with real screens
rendered against FixtureDataManager.empty() / .populated() via
LocalDataManager. GallerySurfaces.kt manifest declares 40 screens.
Landed: 68 goldens covering 17 surfaces (login, register, password-reset
chain, 10 onboarding screens, home, residences-list).
Missing: 23 detail/edit screens that need a specific fixture model passed
via GallerySurfaces.kt — tracked as follow-up in docs/parity-gallery.md.
Non-blocking: these render silently as blank and don't fail the suite.
Android total: 2.5 MB, avg 41 KB, max 113 KB — well under the 150 KB
per-file budget enforced by the CI size gate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- SnapshotGalleryTests rendered at displayScale: 2.0 (was native 3.0)
→ 49MB → 15MB (~69% reduction)
- Records via SNAPSHOT_TESTING_RECORD=1 env var (no code edits needed)
- scripts/optimize_goldens.sh runs zopflipng (or pngcrush fallback)
over both iOS and Android golden dirs
- scripts/{record,verify}_snapshots.sh one-command wrappers
- Makefile targets: make {record,verify,optimize}-snapshots
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Records 58 baseline PNGs across 29 primary SwiftUI screens × {light, dark}
for the honeyDue iOS app. Covers auth, password reset, onboarding,
residences, tasks, contractors, documents, profile, and subscription
surfaces — everything that's instantiable without complex runtime context.
State coverage is empty-only for this first pass: views currently spin up
their own ViewModels which read DataManagerObservable.shared directly, and
the test host has no login → all flows render their empty states. A
follow-up PR adds an optional `dataManager:` init param to each
*ViewModel.swift so populated-state snapshots (backed by P1's
FixtureDataManager) can land.
Tolerance knobs: pixelPrecision 0.97 / perceptualPrecision 0.95 — tuned to
absorb animation-frame drift (gradient blobs, focus rings) while catching
structural regressions.
Tooling: swift-snapshot-testing SPM dep added to the HoneyDueTests target
only (not the app target) via scripts/add_snapshot_testing.rb, which is an
idempotent xcodeproj-gem script so the edit is reproducible rather than a
hand-crafted pbxproj diff. Pins resolve to 1.19.2 (up-to-next-major from
the 1.17.0 plan floor).
Blocks regressions at PR time via `xcodebuild test
-only-testing:HoneyDueTests/SnapshotGalleryTests`.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
testTagsAsResourceId is Android-only; its use in commonMain broke
compileKotlinIosSimulatorArm64. Wrap behind expect fun — Android impl
sets the semantic, other platforms return Modifier unchanged. Blocks
P3 iOS parity gallery otherwise.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces DataManagerEnvironmentKey + EnvironmentValues.dataManager so
SwiftUI views can resolve DataManagerObservable via @Environment, mirroring
Compose's LocalDataManager ambient on the Kotlin side.
No view migrations yet — views continue to read DataManagerObservable.shared
directly. The actual screen-level substitution (fake DataManager for
parity-gallery / tests / previews) lands in P1 when ViewModels gain an
optional init param that accepts the environment-resolved observable. For
this commit we only need the key so P1 can wire against it.
Note: the iosSimulator Kotlin compile is broken at baseline (bb4cbd5)
with pre-existing "Unresolved reference 'testTagsAsResourceId'" errors
across 20+ screen files — Android-only semantics API imported in
commonMain. Swift-parse of the new file succeeds. Verified by checking
out bb4cbd5 and rerunning ./gradlew :composeApp:compileKotlinIosSimulatorArm64.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Swaps direct `DataManager.xxx` access for `LocalDataManager.current.xxx`
across every Compose screen under ui/screens/** that references the
singleton. Each composable resolves the ambient once at the top of its
body and reuses the local val for subsequent reads — keeping rewrites
minimal and predictable.
Screens touched:
- HomeScreen (totalSummary)
- ResidencesScreen (totalSummary)
- ResidenceDetailScreen (currentUser)
- ResidenceFormScreen (currentUser)
- ProfileScreen (currentUser + subscription)
- ContractorDetailScreen (residences)
- subscription/FeatureComparisonScreen (featureBenefits)
- onboarding/OnboardingFirstTaskContent (residences × 3 sites)
No behavior change — in production the ambient default resolves to the
same DataManager singleton. The change is purely so tests, previews, and
the parity-gallery can `CompositionLocalProvider(LocalDataManager provides fake)`
to substitute a fake without tearing screens apart.
Files under ui/subscription/** and ui/components/AddTaskDialog.kt also
reference DataManager but live outside ui/screens/** (plan's scope) —
flagged for a follow-up pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a narrow IDataManager contract covering the 5 DataManager members
referenced from ui/screens/** (currentUser, residences, totalSummary,
featureBenefits, subscription) and a staticCompositionLocalOf ambient
(LocalDataManager) that defaults to the DataManager singleton.
No screen call-sites change in this commit — screens migrate in P0.2.
ViewModels, APILayer, and PersistenceManager continue to depend on the
concrete DataManager singleton directly; the interface is deliberately
scoped to the screen surface the parity-gallery needs to substitute.
Includes IDataManagerTest (DataManager is IDataManager) and
LocalDataManagerTest (ambient val is exposed + default type-checks to the
real singleton). runComposeUiTest intentionally avoided — consistent with
ThemeSelectionScreenTest's convention, since commonTest composition
runtime is flaky on iosSimulator.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
(a) liveRegion + error semantics on form error surfaces so TalkBack
announces them when they appear:
- Shared ErrorCard (used by LoginScreen, RegisterScreen,
VerifyEmail/ResetCode, ForgotPassword, ResetPassword)
- OnboardingCreateAccountContent inline error row
- JoinResidenceScreen inline error row
(b) focusRequester + ImeAction.Next on multi-field forms:
- LoginScreen: auto-focus username, Next→password, Done→submit
- RegisterScreen: auto-focus username, Next chain through
email/password/confirm, Done on last
(c) navigateUp() replaces navController.popBackStack() for simple back
actions in App.kt (6 screens) and MainScreen.kt (3 screens), where
the back behavior is purely navigation-controlled.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces ~163 raw .dp values with design-system tokens per CLAUDE.md rule.
Covers most visible screens (Tasks, Residences, Profile, Documents,
dialogs, kanban, forms). Adds AppSpacing/AppRadius imports where missing.
Remaining sites are geometric/canvas values (stroke widths, icon sizes,
non-standard values like 6.dp/14.dp/20.dp) or don't map to existing
tokens.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Records initial golden set + wires verifyRoborazziDebug into CI. Diffs
uploaded as artifact on failure. ScreenshotTests @Ignore removed.
Root cause of the prior RoboMonitoringInstrumentation:102 failure:
createComposeRule() launches ActivityScenarioRule<ComponentActivity>
which fires a MAIN/LAUNCHER intent, but the merged unit-test manifest
declares androidx.activity.ComponentActivity without a LAUNCHER filter,
so Robolectric's PM returns "Unable to resolve activity for Intent".
Fix: switch to the standalone captureRoboImage(path) { composable }
helper from roborazzi-compose, which registers
RoborazziTransparentActivity with Robolectric's shadow PackageManager
at runtime and bypasses ActivityScenario entirely.
Also pin roborazzi outputDir to src/androidUnitTest/roborazzi so
goldens live in git (not build/) and survive gradle clean.
36 goldens, 540KB total.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
HomeScreen + AllTasksScreen + TasksScreen now support pull-to-refresh.
forceRefresh=true per CLAUDE.md mutation pattern.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CompletionHistorySheet + contractor-picker sheet now use
material3.ModalBottomSheet. Standard Material 3 dim-behind + swipe-down
dismiss + sheet state. Inner content unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Suite6_ComprehensiveTaskTests ports iOS tests not covered by Suite5/10
(priority/frequency picker variants, custom intervals, completion history,
edge cases).
Roborazzi screenshot-regression scaffolding in place but gated with @Ignore
until pipeline is wired — first `recordRoborazziDebug` run needs manual
golden-image review. See docs/screenshot-tests.md for enablement steps.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ports representative subset of Suite8_DocumentWarrantyTests.swift
(22 of 25 iOS tests). testTags on document screens via
AccessibilityIds.Document.*. Documented deliberate skips in the
class header (5/7/8/10/11/12/16) — each either relies on iOS-only
pickers/menus or is subsumed by another ported test.
No new AccessibilityIds added — Document group already has parity.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Audits every list + detail screen (non-document) for empty/error/loading
state parity with iOS. Reuses StandardEmptyState / StandardErrorState where
possible; adds missing states where screens currently show blank on error.
- Add StandardErrorState and CompactErrorState components under
ui/components/common/ (mirrors iOS ErrorView pattern: icon + title +
message + Retry).
- ManageUsersScreen: error state previously had no retry button; now uses
StandardErrorState with a Retry CTA matching iOS ManageUsersView.
- ResidenceDetailScreen: task and contractor sub-section error cards now
use CompactErrorState with inline retry (previously plain error text).
Other audited screens (ResidencesScreen, TasksScreen, AllTasksScreen,
ContractorsScreen, ContractorDetailScreen, EditTaskScreen,
CompleteTaskScreen, TaskTemplatesBrowserScreen, TaskSuggestionsScreen,
OnboardingFirstTaskContent) already had loading + error + empty parity
via ApiResultHandler / HandleErrors / inline state machines; no changes
needed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Port SuiteZZ_CleanupTests.swift. Deletes test-prefixed residences/tasks/
contractors/documents via authenticated APILayer + TaskApi calls. Runs
alphabetically last via the SuiteZZ_ prefix. Each step is idempotent —
logs failures but never blocks the next run. Preserves one seed "Test
House" residence so AAA_SeedTests has a deterministic starting point.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ports Suite7_ContractorTests.swift. testTags on contractor screens via
AccessibilityIds.Contractor.*. CRUD + sharing + link-to-task.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Actionable IconButton/Icon instances now expose meaningful
contentDescription for TalkBack. Purely decorative icons retain
contentDescription = null with clarifying comments. Full audit of
remaining 130+ sites is follow-up work.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ports iOS HoneyDueUITests AccessibilityIdentifiers + PageObjects pattern
to Android Compose UI Test. Kotlin AccessibilityIds object mirrors Swift
verbatim so scripts/verify_test_tag_parity.sh can gate on divergence.
AAA_SeedTests bracketed first alphanumerically; SuiteZZ cleanup to follow.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Soft keyboard no longer covers input fields. Applied to every screen
with text input.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace collectAsState() with collectAsStateWithLifecycle() so StateFlows
stop collecting when the host is in background — prevents memory/CPU leaks
on lifecycle transitions.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Material 3 AutoMirrored variants flip correctly in Arabic/Hebrew.
Previous Icons.Default.ArrowBack pointed wrong direction in RTL.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Prep for parallel Android UI test suite build-out. These deps unblock
page-object tests using onNodeWithTag / createAndroidComposeRule and
cross-app permission-dialog interaction.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds acceptResidenceInvite / declineResidenceInvite to ResidenceApi
(POST /api/residences/{id}/invite/{accept|decline}) and exposes them via
APILayer. On accept success, myResidences is force-refreshed so the
newly-joined residence appears without a manual pull.
Wires NotificationActionReceiver's ACCEPT_INVITE / DECLINE_INVITE
handlers to the new APILayer calls, replacing the log-only TODOs left
behind by P4 Stream O. Notifications are now cleared only on API
success so a failed accept stays actionable.
Tests:
- ResidenceApiInviteTest covers correct HTTP method/path + error surfacing.
- NotificationActionReceiverTest invite cases updated to assert the new
APILayer calls (were previously asserting the log-only path).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Port iOS TaskAnimations.swift specs (completion checkmark, card transitions,
priority pulse) + AnimationTestingView as dev screen.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Stream M replaced the widget's task-complete path with CompleteTaskAction +
WidgetActionProcessor. Grepping confirmed the only references to
WidgetTaskActionReceiver were its own class file and the manifest entry.
Remove both.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
