ci: boot the runner image and run playwright against it

.github/workflows/security.yml

@@ -56,10 +56,40 @@ jobs:
       - run: npm run test:run
       - run: npm run migrate:smoke
       - run: npm run build
-      - run: docker build --target runner .
+      - run: docker build --target runner -t auth-runner:ci .
       - run: docker build --target worker .
+      # The image that ships is the one under test: nonce CSP, HSTS, the
+      # standalone server and migrate-on-boot only exist in production mode.
+      # --network host lets the container reach the service containers on
+      # localhost and lets Playwright reach it on 127.0.0.1:3000. The
+      # Turnstile pair is Cloudflare's documented always-pass test keypair.
+      - name: Boot the runner image
+        run: |
+          docker run -d --name auth-runner --network host \
+            -e DATABASE_URL -e REDIS_URL -e OIDC_PRIVATE_KEY_PEM -e OIDC_KEY_ID \
+            -e AUTH_BASE_URL=http://127.0.0.1:3000 \
+            -e OAUTH_CSRF_SECRET=ci-oauth-csrf-secret \
+            -e TELEGRAM_BOT_WEBHOOK_SECRET=ci-telegram-webhook-secret \
+            -e TURNSTILE_SECRET_KEY=1x0000000000000000000000000000000AA \
+            -e TURNSTILE_SITE_KEY=1x00000000000000000000AA \
+            auth-runner:ci
       - run: npx playwright install --with-deps chromium
+      - name: Wait for the runner to be ready
+        run: |
+          for i in $(seq 1 45); do
+            if curl -fsS http://127.0.0.1:3000/api/health/ready; then exit 0; fi
+            sleep 2
+          done
+          echo "runner image did not become ready" >&2
+          docker logs auth-runner
+          exit 1
       - run: npm run test:e2e
+        env:
+          PLAYWRIGHT_BASE_URL: http://127.0.0.1:3000
+      - if: failure()
+        run: docker logs auth-runner
+      - if: always()
+        run: docker rm -f auth-runner
 
   sdk:
     runs-on: ubuntu-latest

README.md

@@ -89,8 +89,14 @@ With `REDIS_URL` set, rate-limit counters live in Redis and are shared across
 test files within one run, so a test that exercises a rate-limited path must
 use keys unique to itself.
 
-CI (`.github/workflows/security.yml`) runs the unit and integration suites,
-the Playwright end-to-end scenarios against Postgres and Redis services, `npm audit`,
+`npm run test:e2e` starts the dev server on port 3100 by itself. Set
+`PLAYWRIGHT_BASE_URL` to point the suite at an already-running server
+instead (CI sets it to the booted production image).
+
+CI (`.github/workflows/security.yml`) runs the unit and integration suites
+against Postgres and Redis services, boots the production runner image
+against the same services and runs the Playwright end-to-end scenarios
+against it, and runs `npm audit`,
 the migration smoke test, the production build, and both Docker image builds
 on every push. The `sdk` job separately builds and type-checks the Node SDK.
 

playwright.config.ts

@@ -1,22 +1,23 @@
 import { defineConfig, devices } from "@playwright/test";
 
+// CI points PLAYWRIGHT_BASE_URL at the booted production image and manages
+// that process itself; locally the dev server on 3100 is started (or reused).
+const baseURL = process.env.PLAYWRIGHT_BASE_URL || "http://127.0.0.1:3100";
+
 export default defineConfig({
   testDir: "./tests/e2e",
   timeout: 30_000,
   use: {
-    baseURL: "http://127.0.0.1:3100",
+    baseURL,
     trace: "on-first-retry",
   },
-  projects: [
-    {
-      name: "chromium",
-      use: { ...devices["Desktop Chrome"] },
-    },
-  ],
-  webServer: {
-    command: "npm run dev -- -H 127.0.0.1 -p 3100",
-    url: "http://127.0.0.1:3100/login",
-    reuseExistingServer: !process.env.CI,
-    timeout: 120_000,
-  },
+  projects: [{ name: "chromium", use: { ...devices["Desktop Chrome"] } }],
+  webServer: process.env.PLAYWRIGHT_BASE_URL
+    ? undefined
+    : {
+        command: "npm run dev -- -H 127.0.0.1 -p 3100",
+        url: "http://127.0.0.1:3100/login",
+        reuseExistingServer: !process.env.CI,
+        timeout: 120_000,
+      },
 });