E2E: onboarding redirect (/wizard) breaks post-login assumptions

[tomqwu]

Problem

• Multiple E2E tests assume login redirects to /app/schedule, but app can redirect to /wizard when onboarding is incomplete.
• This causes failures like test_solver_workflow + mobile_responsive (mobile_login_flow).

Evidence

• Full rebaseline: 1 fail after 105 passed (maxfail=1): tests/e2e/test_mobile_responsive.py::test_mobile_login_flow expected /app/schedule, got /wizard.
• Solver workflow previously failed similarly.

Plan

1. Centralize a helper/fixture to ensure onboarded state for E2E (detect /wizard then POST /api/onboarding/skip and proceed).
2. Update affected tests to use the helper.
3. Re-run tests/e2e (maxfail=1 loop) until onboarding redirect class is gone.

Acceptance

• No E2E tests fail solely due to /wizard redirect after login.
• All updated tests still validate schedule loads (page-title etc.).

Notes

• Short-term patches applied in: tests/e2e/test_solver_workflow.py and tests/e2e/test_mobile_responsive.py; should be refactored into shared helper.

[tomqwu]

Update: pushed commit a1c6059 on PR #1 to reduce onboarding /wizard redirect flakes in E2E. Added shared skip_onboarding helper and refactored tests to use it; verified mobile_login_flow passes. Next step: rerun full e2e with --maxfail=1 and keep burning down failures.

[tomqwu]

Follow-up: fixed remaining mobile_responsive onboarding flake () by allowing /wizard redirect + calling skip_onboarding_from_storage; verified test passes.

[tomqwu]

Follow-up (avoid shell backticks): Fixed tests/e2e/test_mobile_responsive.py::test_mobile_touch_gestures onboarding redirect flake by allowing /wizard and calling skip_onboarding_from_storage(page), then navigating to /app/schedule. Verified via: poetry run pytest -vv tests/e2e/test_mobile_responsive.py::test_mobile_touch_gestures (PASS).

[tomqwu]

Pinned down the intermittent 5xx that was still breaking the maxfail rebaseline:

• Endpoint: GET /api/onboarding/progress
• Symptom: 500 with UNIQUE constraint failed: onboarding_progress.person_id
• Cause: get-or-create race when multiple concurrent requests hit onboarding progress.

Fix landed: make OnboardingService.get_progress() idempotent by catching IntegrityError, rolling back, and reloading existing progress.

Commit: 3eea414

Next: rerun poetry run pytest -q tests/e2e --maxfail=1 to see what the next deterministic failure is.

[tomqwu]

Update: E2E stability check — repeated runs are now clean.

• Full E2E stability reruns: poetry run pytest -q tests/e2e --maxfail=1 -> 177 passed, 51 skipped (no failures) on multiple consecutive runs.
• Focused flake-hotspot recheck: settings_language_change + volunteer_empty_schedule + mobile_touch_gestures -> 3/3 passed.

This suggests the /wizard onboarding redirect + onboarding_progress idempotency fixes have stabilized the suite. Will keep an eye out for any recurrence, but things look good now.

[tomqwu]

Closed this issue.
Evidence: The Onboarding Dashboard is now functional and handles the SPA redirects correctly.
Local screenshot for review: /Users/tomwu/clawd/signupflow/test-artifacts/onboarding_dashboard_final.png