>>> [!note] Migrated issue <!-- Drupal.org comment --> <!-- Migrated from issue #3619898. --> Reported by: [mably](https://www.drupal.org/user/3375160) Related to !84 >>> <h3>Problem/Motivation</h3> <p>Eight listener registrations across seven test classes subscribe to AuditableEvent by hand, each rebuilding the same three lines against the container's dispatcher. They fall into two families, and neither has an owner.</p> <p>Four collect the events a call dispatches. WriteControllerTest and ReadControllerTest declare a private $auditEvents property, register a listener in setUp() that appends to it, and filter it through a private helper; the two helpers differ only in whether they match the channel as well as the action. ConsumerFloodGuardTest and DormancyTestFormTest want the same thing and get it a different way, with a local $captured array closed over by reference inside the test method.</p> <p>Four make the dispatch fail. VaultTest, ConsentFlowTest (twice) and GarbageCollectorTest register a priority-1000 listener that throws, to prove that a delete rolls its transaction back, that an approval does not commit past a failed audit, and that a per-item guard contains the failure and lets the loop continue. Three of the four also carry their own "which event should throw" condition, spelled out inline.</p> <p>This is test hygiene with no runtime effect. It is worth doing because the two families are exactly the kind of near-identical fixture <span class="drupalorg-gitlab-issue-link project-issue-status-info project-issue-status-2"><a href="https://www.drupal.org/project/yoyaku/issues/3619693" title="Status: Fixed">#3619693: Retire every duplicated method body, in tests as well as production</a></span> and <span class="drupalorg-gitlab-issue-link drupalorg-gitlab-link-wrapper"><a href="https://git.drupalcode.org/project/kessai/-/work_items/3619740" class="drupalorg-gitlab-link">https://git.drupalcode.org/project/kessai/-/work_items/3619740</a></span> found survive a clean clone-detector report: the bodies differ enough to escape a detector and are the same decision repeated.</p> <h3>Proposed resolution</h3> <p>One trait in tests/src/Traits, with a method per family: one that starts collecting and gives the collected events back filtered, and one that makes the dispatch throw for a named event. The collecting method has to be a method call rather than a property read, because PHPStan cannot see a subscriber write a property and narrows it to the empty array it was last assigned; that is what <span class="drupalorg-gitlab-issue-link drupalorg-gitlab-link-wrapper"><a href="https://git.drupalcode.org/project/pdv/-/work_items/3619891" class="drupalorg-gitlab-link">https://git.drupalcode.org/project/pdv/-/work_items/3619891</a></span> hit in WriteControllerTest.</p> <p>Keep the fault-injection predicate at the call site: which event should blow up is what each of those four tests is about, and one of them matches on the context payload rather than on the action alone. The priority does not stay there. This summary first said to leave 1000 explicit at every call site, and writing it showed that to be wrong: the number means the same thing in all four places and a bare 1000 at a call site says nothing, so it belongs in the trait as a named constant carrying its reason.</p> <p>No new coverage, and no test may change what it asserts. Each class keeps its own assertions; only the registration and the filtering move.</p> <h3>Remaining tasks</h3> <ul> <li>Add the trait and convert the four collecting registrations.</li> <li>Convert the four fault-injection registrations.</li> <li>Confirm the suite passes with the same assertion counts it has now.</li> </ul> <h3>User interface changes</h3> <p>None. Test code only.</p> <h3>API changes</h3> <p>None. Test code only.</p> <h3>Data model changes</h3> <p>None. Test code only.</p> <h3>Release notes snippet</h3> <p>None needed: no shipped code changes.</p> <p>AI-Generated: Yes (Claude Code was used to help draft this issue summary and to survey the duplicated listeners it describes. I reviewed it before posting; there is no code on this issue yet.)</p>