fix: Remove process.exit(0) from global teardown to fix CI failure reporting

.github/workflows/ci.yml

@@ -249,14 +249,29 @@ jobs:
 
     - name: Run e2e tests
       run: |
-        xvfb-run --auto-servernum --server-args="-screen 0 1280x960x24" -- pnpm --filter @vibetree/desktop exec playwright test --reporter=list || TEST_EXIT_CODE=$?
+        # Run tests with timeout wrapper to handle worker teardown issues
+        # The timeout command will kill the process if it hangs, and we capture the actual test result separately
+        timeout --signal=KILL 900 xvfb-run --auto-servernum --server-args="-screen 0 1280x960x24" -- pnpm --filter @vibetree/desktop exec playwright test --reporter=list 2>&1 | tee /tmp/playwright-output.txt || true
+
+        # Check if tests actually passed by looking at the output
+        # Playwright prints "X passed" at the end for successful runs
+        # We check for "X passed" and make sure there's no "X failed" pattern
+        # Note: "1 error was not a part of any test" is the worker teardown timeout which we ignore
+        if grep -qE "[0-9]+ passed" /tmp/playwright-output.txt && ! grep -qE "[0-9]+ failed" /tmp/playwright-output.txt; then
+          TEST_EXIT_CODE=0
+          echo "Tests passed!"
+        else
+          TEST_EXIT_CODE=1
+          echo "Tests failed or had errors!"
+        fi
+
         # Kill any lingering Electron processes to prevent worker teardown timeout
         pkill -9 -f electron || true
         pkill -9 -f xvfb || true
         # Wait a moment for processes to die
         sleep 2
-        # Exit with original test exit code
-        exit ${TEST_EXIT_CODE:-0}
+        # Exit with actual test result
+        exit $TEST_EXIT_CODE
       env:
         CI: true
         DISPLAY: ':99.0'

.gitignore

@@ -62,3 +62,6 @@ packages/server/.env
 **/playwright-report/**
 **/test-results/**
 
+# E2E test exit status file
+**/.exit-status
+

apps/desktop/e2e/helpers/test-launcher.ts

@@ -34,9 +34,15 @@ export async function launchElectronApp(options: LaunchOptions = {}): Promise<El
 
 /**
  * Close Electron app quickly using process.exit to prevent worker teardown timeout
- * We use process.exit(0) because electronApp.close() can be too slow (>5 seconds)
- * and cause worker teardown timeouts. The trade-off is acceptable since these are
- * ephemeral test instances.
+ *
+ * We use process.exit(0) because electronApp.close() hangs for this app - the app's
+ * cleanup process takes too long (>2 minutes), causing test timeouts. This is NOT
+ * a Playwright issue (the 30s timeout was removed in v1.19), but rather specific
+ * to this app's shutdown behavior.
+ *
+ * The trade-off: Playwright's exit code becomes unreliable because process.exit(0)
+ * causes connection loss that Playwright interprets as an error. The CI workflow
+ * handles this by parsing test output instead of relying on exit codes.
  *
  * IMPORTANT: We must cleanup fork processes before calling process.exit(), because
  * process.exit() bypasses Electron's before-quit event where cleanup normally happens.

apps/desktop/playwright.config.ts

@@ -7,7 +7,9 @@ export default defineConfig({
   retries: process.env.CI ? 2 : 0,
   workers: 1, // Use only 1 worker for Electron tests
   reporter: 'html',
-  globalTeardown: './e2e/global-teardown.ts',
+  // Note: globalTeardown removed - the CI workflow handles process cleanup
+  // externally via pkill commands to avoid worker teardown timeout issues.
+  // Using process.exit() in globalTeardown would mask test failures.
   // Increased timeout to 120s (2 minutes) to allow more time for worker teardown
   // The worker teardown timeout uses the same value as the test timeout
   // This helps prevent "Worker teardown timeout of 60000ms exceeded" errors