testing-playwright-e2e

Use web-first assertions like toHaveText that auto-wait and retry, never call waitForTimeout, and wait for specific conditions such as URLs or network responses. Prove stability by running new specs with --repeat-each=5 and parallel workers before marking them done.

Follow the priority order: getByRole first, then getByLabel, getByPlaceholder, getByText, and getByTestId, reserving CSS or XPath selectors as a last resort. Locators tied to user-visible semantics survive DOM and style refactors.

Log in once per role in a setup project and save the session with storageState, then reference that state file in dependent projects via playwright.config. Always load credentials from environment variables, never hardcode them in the repository.

Prefer fixtures over classic page object inheritance because fixtures compose, handle setup and teardown, and provide typing. Keep page objects lightweight by encapsulating locators and actions only, leaving assertions visible inside the tests.

CI failures usually come from shared test data colliding under parallel workers, order-dependent tests, or timing assumptions hidden by faster local machines. Ensure each test seeds its own unique data via API and runs independently under fullyParallel mode.

Use real services for critical flows like checkout and authentication, but mock third-party integrations such as payment providers, email, or external OAuth with page.route. Also mock error responses to test failure paths, not just the happy path.