Write widget tests for mentor info popup and list fallback
epic-geographic-peer-mentor-map-ui-accessibility-task-015 — Write Flutter widget tests for MentorInfoPopup and MentorListFallback. Popup tests: loads from BLoC state without network call, focus trap is active, close button dismisses popup, deep-link button navigates correctly. List fallback tests: renders all filtered mentors, live region announces count update, each item has correct semantic label, view toggle preserves list position.
Acceptance Criteria
Technical Requirements
Execution Context
Tier 5 - 253 tasks
Can start after Tier 4 completes
Implementation Notes
For focus trap tests, use tester.sendKeyEvent(LogicalKeyboardKey.tab) and assert that the focused widget's key is still a descendant of the popup's key. Inject a MockGoRouter via ProviderScope or a constructor parameter so navigation assertions do not require a real routing context. For live region tests, ensure the Semantics widget uses the liveRegion: true property rather than politeAnnouncement — the test must validate the correct flag. The list scroll preservation test should store the ScrollController in the BLoC or a state holder accessible to both the list and the map view so that the controller is not disposed on view toggle.
Testing Requirements
Widget tests using flutter_test, MockBloc, and a MockGoRouter. Test focus trap behaviour by asserting FocusManager.instance.primaryFocus remains within the popup subtree after simulated keyboard navigation. Test live region by asserting SemanticsNode.hasFlag(SemanticsFlag.isLiveRegion) on the count text node. Scroll position preservation: use ScrollController and record controller.offset before and after ViewToggled event.
Achieve full branch coverage for empty list, single item, and multi-item states.
Flutter's map canvas (flutter_map) does not natively support semantic focus traversal for screen readers, meaning map markers may be entirely invisible to VoiceOver/TalkBack users. If the accessible list fallback is not treated as a first-class view, screen reader users will have no access to the feature.
Mitigation & Contingency
Mitigation: From sprint 1, treat mentor-list-fallback as a fully featured primary view, not an afterthought. Implement and test it in parallel with the map canvas. Make the view-toggle-button keyboard focusable and announced on every screen state. Conduct VoiceOver testing on device before submitting each PR touching UI components.
Contingency: If map canvas accessibility cannot be achieved for marker focus traversal, make the view-toggle-button the default focus target on screen load for VoiceOver users (detected via screen-reader-detection-service) so they are immediately directed to the list fallback without needing to discover the toggle.
The mentor-info-popup must occupy no more than 40% of visible map area on small screens. On devices with screen heights under 667px (iPhone SE), overlapping with the filter panel or obscuring most of the map could severely degrade usability.
Mitigation & Contingency
Mitigation: Implement the popup as a bottom sheet capped at 40% of screen height with a ScrollView for overflow content. Test on iPhone SE (375x667pt) and the smallest commonly used Android form factor in the device lab. Define max-height as a percentage constant in location-privacy-config or design tokens.
Contingency: If the popup cannot fit all required fields within 40% height on smallest targets, truncate assigned contact count and certification badge to icons-only in the compact view, with a 'View Profile' button always visible at the bottom of the popup regardless of scroll position.
Filter state must remain perfectly synchronised between the map view and the list fallback. If the filter panel emits state that is not consumed identically by both views, coordinators switching between views will see inconsistent mentor sets, eroding trust in the feature.
Mitigation & Contingency
Mitigation: Store active filter criteria in a single shared Riverpod provider owned by the map-view-screen and consumed by both map-marker-widget (via mentor-location-service) and mentor-list-fallback. Write integration tests that apply a filter, switch views, and assert identical mentor counts in both views.
Contingency: If filter sync proves brittle, simplify to a single filter state object passed explicitly as a constructor argument to both views on each rebuild, eliminating indirect state sharing.