Playwright test isolation is incorrect, leading to random failures
>>> [!note] Migrated issue <!-- Drupal.org comment --> <!-- Migrated from issue #3554549. --> Reported by: [jessebaker](https://www.drupal.org/user/3546546) Related to !814 !856 !841 !815 !837 !804 !771 !730 !264 >>> <h3 id="overview">Overview</h3> <p>I believe our current playwright tests install Drupal/create a database <em>per worker</em> but should instead do this <em>per test case</em>.</p> <p>The current configuration means that a file full of tests could be run on a non-clean install based on Playwright deciding to re-use workers that have reached the end of their test file.</p> <h3 id="proposed-resolution">Proposed resolution</h3> <p>Investigate test speed/stability when changing the <code>scope</code> of the Drupal install step from <code>worker</code> to <code>test</code>. </p> <h3 id="ui-changes">Remaining work</h3> <ul> <li>Re-enable disabled tests</li> <li>Add some tests back to the shared worker</li> <li>Add an eslint rule to block direct import of playwright's test object</li> <li>Re-enable testing for firefox and webkit</li> </ul>
issue