ci: run E2E tests on every PR, surface on release (non-blocking) (#842)

.github/workflows/e2e.yml

@@ -0,0 +1,59 @@
+name: E2E
+
+on:
+  workflow_call:
+    inputs:
+      ref:
+        description: Ref to check out (defaults to the calling workflow's ref)
+        required: false
+        type: string
+
+jobs:
+  e2e:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.ref || github.ref }}
+
+      # Why: electron-vite's globalSetup invokes a full app build, which
+      # compiles native modules via node-gyp. Mirrors the install step in
+      # pr.yml's verify job so E2E doesn't hit missing-toolchain errors.
+      - name: Install native build tools
+        run: sudo apt-get update && sudo apt-get install -y build-essential python3
+
+      # Why: Electron on Linux needs an X display even when the app
+      # suppresses mainWindow.show() via ORCA_E2E_HEADLESS. xvfb provides a
+      # virtual framebuffer so Chromium can initialize without a real display.
+      - name: Install xvfb
+        run: sudo apt-get install -y xvfb
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 24
+
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          run_install: false
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Run E2E tests
+        run: xvfb-run --auto-servernum pnpm run test:e2e
+
+      # Why: Playwright retains traces/screenshots only on failure. Uploading
+      # them as an artifact makes post-mortem debugging on CI possible without
+      # re-running locally.
+      - name: Upload Playwright traces
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: playwright-traces
+          path: test-results/
+          retention-days: 7
+          if-no-files-found: ignore

.github/workflows/pr.yml

@@ -43,3 +43,6 @@ jobs:
 
       - name: Build
         run: pnpm build
+
+  e2e:
+    uses: ./.github/workflows/e2e.yml

.github/workflows/release.yml

@@ -63,6 +63,18 @@ jobs:
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
+  # Why: E2E runs alongside the release for visibility (failures surface as a
+  # red check on the tag), but it is NOT in `release`'s needs list. Releases
+  # already take a while and the suite is already a required check on PRs, so
+  # gating here would mostly just delay shipping without adding much signal.
+  # Matches the pattern used by noqa's app deploy, which runs E2E with
+  # continue-on-error so failures are visible but don't block the deploy.
+  e2e:
+    needs: resolve-release
+    uses: ./.github/workflows/e2e.yml
+    with:
+      ref: ${{ needs.resolve-release.outputs.ref }}
+
   release:
     needs:
       - resolve-release

tests/playwright.config.ts

@@ -26,6 +26,11 @@ export default defineConfig({
   // substantially. The few visible-window tests that still rely on real
   // pointer interaction are marked serial in their spec file instead.
   fullyParallel: true,
+  // Why: Playwright defaults to workers=1 on CI, which would serialize all
+  // specs on the ubuntu-latest runner (4 vCPUs) and waste headroom. Each test
+  // launches an isolated Electron instance with its own userData dir, so they
+  // don't share state — we can safely fan out to match the runner's vCPU count.
+  workers: process.env.CI ? 4 : undefined,
   forbidOnly: !!process.env.CI,
   retries: process.env.CI ? 1 : 0,
   reporter: 'list',