ci(e2e): serve a production build instead of the dev server

The suite has been failing on CI while staying green locally: 19 failures
spread across admin navigation, the editor publish flow, bookmarking, the
feed sidebar and moderation. Identical failures on develop and on every
branch off it, so nothing in the feature work caused them.

They all share a cause. Playwright's webServer ran `next dev`, so the
first request to each route blocked on an on-demand Turbopack compile.
Locally that is under a second; on a cold runner with three workers
compiling at once it outlasts the 10s expect timeout — hence assertions
like `toHaveURL(/admin/users)` polling 13 times and giving up while the
navigation was still compiling.

CI now builds once and serves it. Measured on the same machine, a cold
route costs ~0.7-1.2s under `next dev` and ~0.02-0.03s prebuilt.

Local runs keep `next dev` for the fast feedback loop.

EMAIL_AUTH_ENABLED is set for the job because a production build runs with
NODE_ENV=production, which would otherwise disable the passwordless
provider that dev turns on implicitly.

Author: NiallJoeMaher

.github/workflows/e2e-tests.yml

@@ -41,6 +41,9 @@ jobs:
     env:
       DATABASE_URL: "postgresql://postgres:secret@localhost:5432/postgres"
       NEXTAUTH_URL: http://localhost:3000/api/auth
+      # A production build sets NODE_ENV=production, which would otherwise turn
+      # off the passwordless provider that dev enables implicitly.
+      EMAIL_AUTH_ENABLED: "true"
       GITHUB_ID: ${{ secrets.E2E_GITHUB_ID }}
       GITHUB_SECRET: ${{ secrets.E2E_GITHUB_SECRET }}
       NEXTAUTH_SECRET: ${{ secrets.NEXTAUTH_SECRET }}
@@ -91,6 +94,12 @@ jobs:
           npm run db:migrate
           npm run db:seed
 
+      # Playwright serves this build (see webServer in playwright.config.ts).
+      # Tests used to run against `next dev`, where the first hit on a route
+      # blocked on an on-demand compile for longer than the assertion timeouts.
+      - name: Build app
+        run: npm run build:e2e
+
       - name: Run Playwright tests
         id: playwright-tests
         run: npx playwright test

package.json

@@ -5,6 +5,8 @@
   "scripts": {
     "dev": "next dev",
     "dev:e2e": "ENV=E2E next dev",
+    "build:e2e": "ENV=E2E next build",
+    "start:e2e": "ENV=E2E next start",
     "build": "next build",
     "ci-build": "next build",
     "vercel-build": "if [ \"$VERCEL_ENV\" = \"production\" ]; then npm run db:migrate; fi && next build",

playwright.config.ts

@@ -68,10 +68,19 @@ export default defineConfig({
 
   outputDir: "playwright-report",
 
-  /* Run your local dev server before starting the tests */
+  /* Run your local dev server before starting the tests.
+
+     CI serves a production build instead. Against `next dev`, the FIRST visit
+     to each route waits on an on-demand Turbopack compile, which on a cold
+     runner routinely outlasts the 10s expect timeout — that is what made the
+     admin-nav, editor-publish, bookmark, feed and moderation specs fail there
+     while the same suite stayed green locally. A prebuilt server has no
+     per-route compile step, so those assertions see the page immediately.
+     The workflow runs `build:e2e` before invoking Playwright. */
   webServer: {
-    command: "npm run dev:e2e",
+    command: process.env.CI ? "npm run start:e2e" : "npm run dev:e2e",
     url: "http://127.0.0.1:3000",
     reuseExistingServer: !process.env.CI,
+    timeout: 120_000,
   },
 });