ci / test (push) Failing after 1m29s
The prefs were a free dict compared as text with the `get` reply, so lint and sanity could not see that several keys never converge: - `get radio.silence` -> '30 min (now x1), quiet ...', `get batt.icon` -> 'on (level 3/5, ...)', `get dutycycle` -> '50.0%': a change on every run, then a failure after the `set`; - `advert.interval` is stored halved (25 reads back as 24, and the node takes only 0 or 60-240), `agc.reset.interval` in steps of 4; - `set repeat <anything but off>` means ON, so `repeat: 0` turned it on; - lat/lon compared to 1e-3 (~100 m), so a smaller move never went out; - errors like 'ERR - ...', 'unknown config: ...', 'Invalid ...' did not match the expected reply and ended as a timeout after five resends. module_utils/meshcore_prefspec.py now holds kind, range, step, tolerance and a reply parser for every pref of the meshcore-tth CLI (CommonCLI.cpp, CommonRadioPrefs.cpp, NodeMaintenance.cpp, CommonGpsPrefs.cpp); the module refuses a value the node would refuse or alter before it sends anything, in check mode too. Unknown keys still go, compared as before, with a warning. radio.txstall (RAM only) and radio.rxps (compound) are refused with a reason. New keys that are plain commands in the firmware: gps (off/on/duty), gps.period, gps.firstfix and powersaving; gps.period goes first and conflicts with gps off/duty. Tests: replies as the firmware prints them and as a bench board and three repeaters answered; with MESHCORE_TTH_SRC set, a drift check that every key is still in the CLI sources.