mirror of
https://github.com/MBombeck/HealthLog.git
synced 2026-11-10 05:12:14 +00:00
`/` runs its reads server-side and hands them to TanStack through a HydrationBoundary, so the first HTML already carries real tiles. React threw that away on every cold load. The app shell hydrates first and its mount answers `/api/auth/me`; the dashboard sits below `loading.tsx`, hydrates later, and its first client render therefore read an account the server never had. That disagreed with the streamed HTML, React reported #418 and re-rendered the whole page. The prefetch paid for its queries and delivered none of its paint. The account payload is now withheld from the hydration render wherever it is only rendered. `useAccountOnceMounted()` answers null until the component has mounted, which is what the server rendered, and hands the payload over on the next render. The dashboard page, the caregiver card, the getting-started checklist, the achievements tile and the unit labels read it that way. `useAuth` itself still reads the cache, because the document vault redirects off `user.modules` and a first render claiming "no account" bounces it to `/`. Two tiles disagreed without any account involved. A query reports `isLoading` true on the client's first render and false on the server, so both now gate on whether the cell holds an answer. Nothing in CI ran the shipped configuration: the suite's server turns the prefetch off so route mocks keep governing the dashboard. A second server runs it on, and one spec asserts that no hydration error is reported and that the tile node the server streamed is still the node in the DOM once the page settles. Measured over nine cold loads with the CPU throttled 4x: seven bailouts before and none after, median LCP 869 ms before and 145 ms after. FCP and the time to the first tile in the DOM are unchanged, because the streamed paint was never the problem. Keeping it was.
156 lines
7.0 KiB
TypeScript
156 lines
7.0 KiB
TypeScript
import { defineConfig, devices } from "@playwright/test";
|
|
|
|
import {
|
|
SSR_PREFETCH_BASE_URL,
|
|
SSR_PREFETCH_PORT,
|
|
} from "./e2e/setup/ssr-prefetch-server";
|
|
|
|
/**
|
|
* Playwright configuration for the HealthLog E2E suite.
|
|
*
|
|
* The suite covers the smoke-level user paths (auth redirect, login
|
|
* form, public version endpoint, locale switch, axe-core) without
|
|
* needing seeded data — every spec either runs against an unauthed
|
|
* surface or uses route interception to stub out the API. Specs that
|
|
* need a logged-in user are kept narrow and flagged in their describe
|
|
* block; CI runs them against a worker that seeds a deterministic test
|
|
* user on startup.
|
|
*
|
|
* To run locally: `pnpm dlx playwright install --with-deps chromium`
|
|
* once, then `pnpm e2e`.
|
|
*/
|
|
export default defineConfig({
|
|
testDir: "./e2e",
|
|
testIgnore: ["setup/**"],
|
|
globalSetup: "./e2e/setup/global-setup.ts",
|
|
timeout: 30_000,
|
|
expect: { timeout: 10_000 },
|
|
fullyParallel: true,
|
|
forbidOnly: !!process.env.CI,
|
|
failOnFlakyTests: !!process.env.CI,
|
|
// Two retries on CI: under a loaded shared runner, DOM-settle-gated
|
|
// assertions (wizard step transitions, list refetch → card provenance,
|
|
// disclosure toggles) intermittently sample a mid-transition frame. A
|
|
// retry distinguishes transient contention from a persistent break, while
|
|
// failOnFlakyTests keeps either outcome visible to CI. The default expect
|
|
// timeout is lifted from 5s → 10s for the same settle headroom.
|
|
retries: process.env.CI ? 2 : 0,
|
|
workers: process.env.CI ? 2 : undefined,
|
|
reporter: process.env.CI
|
|
? [["github"], ["html", { open: "never" }]]
|
|
: [["list"], ["html", { open: "never" }]],
|
|
|
|
use: {
|
|
baseURL: process.env.E2E_BASE_URL ?? "http://localhost:3000",
|
|
trace: "retain-on-failure",
|
|
screenshot: "only-on-failure",
|
|
video: "retain-on-failure",
|
|
// v1.18.6 — block the service worker in the test context. The
|
|
// production build registers `/sw.js`, whose v1.18.6 data branch
|
|
// serves allowlisted `/api/*` GET reads network-first. A worker-
|
|
// originated `fetch` is NOT subject to Playwright's `page.route`
|
|
// interception, so the SW would re-fetch the real (empty) backend
|
|
// and serve that — bypassing every per-spec `route.fulfill` mock and
|
|
// breaking read-after-write assertions (a just-created row never
|
|
// surfaced). Blocking the worker keeps the route mocks authoritative
|
|
// without weakening the shipped SW behaviour. Real users keep the
|
|
// offline data cache; only the test harness opts out.
|
|
serviceWorkers: "block",
|
|
// HealthLog ships dark mode as the default (Dracula theme) — `globals.css`
|
|
// sets `color-scheme: dark` on the root and the `<ThemeProvider>` defaults
|
|
// to "system". Playwright's stock context is `colorScheme: "light"`, which
|
|
// means axe-core was scanning a layout users never actually see (Dracula
|
|
// greens on a light card → 1.18 contrast). Forcing the test theme to dark
|
|
// matches what real users render on first paint and is the only honest
|
|
// a11y baseline for this app.
|
|
colorScheme: "dark",
|
|
// Issue #490 — pin the browser timezone. Without this the context
|
|
// inherits the HOST zone (UTC on CI, Europe/Berlin on a local Mac),
|
|
// so any surface that renders a clock or a day boundary could pass
|
|
// locally and skew two hours / a calendar day on CI. Berlin matches
|
|
// the app's own display fallback, so CI and local render identically.
|
|
timezoneId: "Europe/Berlin",
|
|
},
|
|
|
|
projects: [
|
|
{
|
|
name: "chromium-desktop",
|
|
use: {
|
|
...devices["Desktop Chrome"],
|
|
viewport: { width: 1280, height: 720 },
|
|
colorScheme: "dark",
|
|
},
|
|
},
|
|
{
|
|
name: "chromium-mobile",
|
|
use: {
|
|
// Pixel 5 — Chromium-based mobile profile so CI only needs
|
|
// `playwright install chromium` instead of also pulling
|
|
// webkit. iPhone-13 is intentionally avoided here; the
|
|
// mobile-Safari smoke is exercised by the iOS app suite.
|
|
...devices["Pixel 5"],
|
|
colorScheme: "dark",
|
|
},
|
|
},
|
|
],
|
|
|
|
// Spin up the production build for E2E. `pnpm build` is run
|
|
// separately by CI before the suite — locally, set E2E_SKIP_WEB_SERVER=1
|
|
// to point at an already-running dev server.
|
|
webServer: process.env.E2E_SKIP_WEB_SERVER
|
|
? undefined
|
|
: [
|
|
{
|
|
command: `mkdir -p .next/standalone/.next/static .next/standalone/public && cp -R .next/static/. .next/standalone/.next/static && cp -R public/. .next/standalone/public && PORT=3000 HOSTNAME=127.0.0.1 ${JSON.stringify(process.execPath)} .next/standalone/server.js`,
|
|
url: "http://localhost:3000/api/version",
|
|
timeout: 60_000,
|
|
reuseExistingServer: !process.env.CI,
|
|
stdout: "ignore",
|
|
stderr: "pipe",
|
|
env: {
|
|
...process.env,
|
|
// Native document rendering is fail-soft and outside the browser
|
|
// suite. Disable it here so queued local thumbnail jobs cannot
|
|
// crash the shared server inside Skia during unrelated scenarios.
|
|
NATIVE_CANVAS: "off",
|
|
// The dashboard RSC wrapper server-prefetches the snapshot into
|
|
// the first HTML (HydrationBoundary). Playwright's route mocks
|
|
// only see CLIENT fetches, so an SSR-embedded snapshot would
|
|
// bypass `mockDashboardSnapshot` and every dashboard spec would
|
|
// assert against the seeded account instead of its fixture.
|
|
// Disable the prefetch for the e2e server — the suite keeps the
|
|
// deterministic client-fetch path.
|
|
DASHBOARD_SSR_PREFETCH: "false",
|
|
},
|
|
},
|
|
// The SHIPPED configuration, on its own port.
|
|
//
|
|
// Every spec above runs against a server with the dashboard's RSC
|
|
// prefetch disabled, which is the one thing self-hosters never run:
|
|
// the flag defaults ON. That gap let a React #418 hydration bailout
|
|
// reach production on `/` for a signed-in account, on both viewports,
|
|
// on every cold load — the client's first render read query cells the
|
|
// server never had, so React discarded the streamed dashboard and
|
|
// rebuilt it. The suite could not see it, because with the prefetch
|
|
// off both sides paint the same skeleton.
|
|
//
|
|
// One extra server on 3100 closes that. Only
|
|
// `dashboard-ssr-prefetch-hydration.spec.ts` talks to it, and it
|
|
// asserts the one thing the mocked specs cannot: that the page the
|
|
// server streamed is the page the browser keeps.
|
|
{
|
|
command: `PORT=${SSR_PREFETCH_PORT} HOSTNAME=127.0.0.1 ${JSON.stringify(process.execPath)} .next/standalone/server.js`,
|
|
url: `${SSR_PREFETCH_BASE_URL}/api/version`,
|
|
timeout: 60_000,
|
|
reuseExistingServer: !process.env.CI,
|
|
stdout: "ignore",
|
|
stderr: "pipe",
|
|
env: {
|
|
...process.env,
|
|
NATIVE_CANVAS: "off",
|
|
DASHBOARD_SSR_PREFETCH: "true",
|
|
},
|
|
},
|
|
],
|
|
});
|