[Question] browser.newContext failed why?

[mtt87]

This is randomly happening for me only on GitLab CI, when I run tests locally everything is fine 100% of the times.

I'm using this config

e2e:
  stage: e2e
  image: mcr.microsoft.com/playwright:focal
  script:
    - npm run test:e2e

"test:e2e": "DEBUG=pw:api playwright test e2e",

const config: PlaywrightTestConfig = {
  workers: 1,
  retries: 0,
  timeout: 150000,
  globalSetup: './globalSetup',
}

In my global setup I signup 3 different type of users and save their 3 browser context into json storage.
Something like this

async function globalSetup() {
  const browser = await chromium.launch({
    slowMo: 2000,
  })
  const baseContext = await browser.newContext({
    baseURL: frontendBaseUrl,
  })
  // do stuff
  await baseContext.context().storageState({ path: 'baseStorageState.json' })
    const ownerContext = await browser.newContext({
    baseURL: frontendBaseUrl,
  })
  // do stuff
  await ownerPage.context().storageState({ path: 'ownerStorageState.json' })
  // repeat

  await browser.close()
}

Then I have a test.describe.serial("3 roles", ...) that is testing operations on the 3 roles. I know serial is not advised but I need to because I have a use case where the first user has to invite others in their "workspace" and the other tests can only start when the invite has been created and then consumed.

test.describe.serial('Console', () => {
  let inviteLinkAdmin: string
  let inviteLinkViewer: string
  let browser: ChromiumBrowser

  test.beforeAll(async () => {
    browser = await chromium.launch({
      slowMo: 500,
    })
  })

  test.afterAll(async () => {
    await browser.close()
  })

  test('OWNER ROLE', async () => {
    const context = await browser.newContext({
      storageState: 'ownerStorageState.json',
      baseURL: frontendBaseUrl,
    })
    // do stuff
  })

  test('VIEWER ROLE', async () => {
    const context = await browser.newContext({
      storageState: 'viewerStorageState.json',
      baseURL: frontendBaseUrl,
    })
    // do stuff
  })

As you can see I have DEBUG=pw:api enabled but it still doesn't tell me exactly why browser.newContext failed, is there a way to know?

[shirshak55]

is this because of baseURL: frontendBaseUrl? Maybe the frontendBaseUrl is not accessible that's why it timed out ?

[mtt87]

is this because of baseURL: frontendBaseUrl? Maybe the frontendBaseUrl is not accessible that's why it timed out ?

Everything is possible 😄 but I would say unlikely, it's a public URL served by Vercel, I didn't notice any downtime this morning, and while I was running the test on CI I was also running it on my laptop, same config same urls.

[dgozman]

As you can see I have DEBUG=pw:api enabled but it still doesn't tell me exactly why browser.newContext failed, is there a way to know?

The error message is indeed not helpful. Looking at the logs, there is browser.close started right before browser.newContext failed, so my guess would be: the test times out after 150 seconds, afterAll closes the browser and that makes browser.newContext fail.

Most likely browser.newContext times out, perhaps similar to #9788?

[mtt87]

Okay that sounds possible, I'll close the issue as it's most likely related.
Perhaps Playwright should not try to continue with the test if it reached the timeout and not show that line as error but maybe it's just a little improvement not a bug. Thank you!