[Bug]: UI Mode buffers all test steps when root-level testDir conflicts with project-specific testDirs

[AlonMiz]

Version

1.47.2 (also happens on latest)

Steps to reproduce

1. Create a playwright.config.ts with BOTH a root-level testDir AND project-specific testDir:

import {defineConfig, devices} from '@playwright/test';

export default defineConfig({
  testDir: './integration',  // ❌ Root-level testDir pointing to non-existent directory
  
  projects: [
    {
      name: 'chromium',
      use: {...devices['Desktop Chrome']},
      testDir: './tests/e2e'  // ✅ Project-specific testDir (actual location)
    }
  ]
});

2. Create a simple test in ./tests/e2e/example.spec.ts:

import {test, expect} from '@playwright/test';

test('example test', async ({page}) => {
  await page.goto('https://playwright.dev');
  await page.locator('h1').waitFor();
  await expect(page).toHaveTitle(/Playwright/);
});

3. Run Playwright UI Mode:

npx playwright test --ui --project=chromium

4. Observe the behavior in the UI Mode sidebar

Expected behavior

Test steps should appear in real-time in the UI Mode sidebar as they execute:

• "page.goto" appears immediately when navigation starts
• "locator.waitFor" appears when waiting begins
• "expect" appears when assertion runs
• Timeline shows progress in real-time

Actual behavior

All test steps are buffered and appear only after the test completes or fails:

• UI Mode sidebar shows "Running" with a loading indicator
• Timeline stays at 0s
• No steps appear in the sidebar during test execution
• After ~30 seconds (or when test completes), ALL steps dump at once
• Timeline jumps from 0s to final duration instantly

Screenshot of buffering behavior:

• Sidebar shows "Running" with blank steps list
• Timeline at 0s despite test executing
• All actions appear simultaneously after completion

Additional context

Root Cause Analysis

The issue occurs because:

1. Playwright's test discovery mechanism tries to scan BOTH:

   • Root-level testDir: './integration' (which doesn't exist)
   • Project-level testDir: './tests/e2e' (which exists)

2. This creates a race condition in the test discovery process

3. The UI Mode reporter waits for the root-level discovery to complete before streaming steps

4. Since ./integration doesn't exist, the discovery times out (~30 seconds)

5. During this timeout, all test steps are buffered instead of streamed to the UI

Workaround

Remove or comment out the root-level testDir:

export default defineConfig({
  // testDir: './integration',  // ❌ Remove this line
  
  projects: [
    {
      name: 'chromium',
      use: {...devices['Desktop Chrome']},
      testDir: './tests/e2e'  // ✅ Only use project-specific testDir
    }
  ]
});

Suggested Fix

Playwright should either:

1. Validate configuration: Warn when both root-level and project-specific testDir are set
2. Handle gracefully: Prioritize project-specific testDir and ignore root-level when projects are defined
3. Document clearly: Add a warning in the documentation about this limitation
4. Fail fast: If root-level testDir doesn't exist, fail immediately instead of timing out

Impact

This issue affects:

• Developer experience in UI Mode (appears broken/frozen)
• Debugging efficiency (can't see steps in real-time)
• Large monorepos with multiple projects (common pattern)

Environment

System:
  OS: macOS Sequoia
  CPU: (10) arm64 Apple M1 Pro
Binaries:
  Node: 20.19.0 - ~/.nvm/versions/node/v20.19.0/bin/node
  npm: 10.8.2 - ~/.nvm/versions/node/v20.19.0/bin/npm
npmPackages:
  @playwright/test: 1.47.2 => 1.47.2

[yossiyedid]

same here, strange behavior.