0.1.0a6
Setup no longer fails on an inverter that closes the Modbus session after a
read times out. 0.1.0a5 reported "Lost the connection to the inverter:
read_input_registers(13264, 15)" and retried forever on such an inverter; a
lost connection now gets one retry per component before the poll is abandoned.
The control test puts every register back by writing the raw word it held
before the test, checked by reading the register itself, so a restore no
longer depends on the write scale under test.
Known issue: the export power limit (register 13074) is written at a tenth of
the requested value on at least one firmware -- a master/slave SH10RT-V112 pair
-- that does not multiply writes by ten. Under investigation.