Lucas BergerandClaude Opus 4.8
c7955a46b9
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
5.5 KiB
Phase 7: Mobile Test Harness - Discussion Log
Audit trail only. Do not use as input to planning, research, or execution agents. Decisions are captured in CONTEXT.md — this log preserves the alternatives considered.
Date: 2026-06-10 Phase: 7-Mobile Test Harness Areas discussed: Device profile(s), Test-data strategy, Assertion approach, Stack lifecycle / baseURL
Device profile(s)
| Option | Description | Selected |
|---|---|---|
| iPhone only | Single iPhone profile; matches Apple-member UX constraint; fastest, misses Android-Chrome layout | |
| iPhone + Pixel | Two-profile matrix covering both ecosystems; ~2x runtime | ✓ |
| iPhone + small-Android | iPhone + narrow Android profile to stress tightest viewport |
User's choice: iPhone + Pixel
Follow-up: engine fidelity
| Option | Description | Selected |
|---|---|---|
| Faithful engines | iPhone → WebKit, Pixel → Chromium; most faithful; adds WebKit to CI image | ✓ |
| Chromium-only | Both Chromium, iPhone viewport/UA/touch only; lighter, matches playwright-cli |
User's choice: Faithful engines Notes: Accepts heavier browser image for true WebKit/Chromium rendering fidelity.
Test-data strategy
| Option | Description | Selected |
|---|---|---|
| Seed DB fixtures | Insert deterministic rows before run; realistic end-to-end render path | |
| Mock API routes | Playwright route-fulfill canned JSON; hermetic, bypasses real API | |
| Chrome/empty-states only | No population; assert nav/drawers/tap-targets/empty copy; smallest scope | |
| Hybrid: seed + empty | Seed DB for populated views + keep empty-state assertions; broadest coverage | ✓ |
User's choice: Hybrid: seed + empty Notes: Captured constraint — seed must be deterministic and reset per run (SC #3, repeatable day-over-day); runs in global-setup against MariaDB :3306, targets shared calendar id 10 + lists.
Assertion approach
| Option | Description | Selected |
|---|---|---|
| Structural + tap-targets | Role/locator + measured box checks (≥44px, no overflow); portable, stable; misses visual drift | |
| Both (structural + screenshots) | Add toHaveScreenshot; catches visual regressions but flaky cross-env | |
| Screenshots-primary | Lean on visual snapshots; highest flakiness/maintenance |
User's choice: Other (free text) — "Defer this decision to research and for you to decide as it needs to be robust and low maintenance. I have to imagine this has been done elsewhere before and should be well documented." Notes: Marked as research question, not locked. Claude's steer: lead with structural + tap-target measurement; add screenshots only if research finds a non-flaky CI-baseline pattern. Schedule-X widget makes naive pixel snapshots drift-prone.
Stack lifecycle / baseURL
| Option | Description | Selected |
|---|---|---|
| baseURL + readiness wait | Configurable baseURL, readiness gate; caller owns stack bring-up; matches existing dev stack | ✓ (Claude, per "you decide") |
| Playwright webServer | Auto-start vite; can't own multi-container API/DB/Redis stack | (partial — vite only) |
| You decide | Pick best fit for SC #4 + local ergonomics | ✓ |
User's choice: You decide
Notes: Claude's recommendation — baseURL (env-driven) + global-setup readiness gate on /health; caller (operator locally / compose in CI) brings up the stack; optional webServer for vite only with reuseExistingServer: !CI; API+MariaDB+Redis stay compose-managed. Satisfies SC #4.
Claude's Discretion
- Assertion approach — deferred to research (robust/low-maintenance, host↔CI portable).
- Stack lifecycle — "you decide"; recommendation captured above, planner may refine env-var name / webServer wiring.
- Spec file location/structure, failure trace/artifact capture, npm-script + Makefile wiring — not discussed; planner's discretion following existing conventions.
Deferred Ideas
- "Gitea CI — full regression + Docker publish" todo (score 0.6) — belongs to Phase 8, which consumes this harness. Reviewed, not folded.