With the current implementation, we run into a deadlock with the lock
from the state machine interfering with the accessLock on the module.
We can not wait for the state machine to finish while having the
accessLock locked by write_target. As a consequence, when restarting
the state machine we should not wait, but remember the state function
to call and postpone the restart after the cleanup has finished.
For this, we want to know the status before calling the state function.
- create HasState mixin, using doPoll for driving the machine
- StatusCode decorator for assigning a status to a state function
- remove the state machines 'threaded' option
- 'Retry' is now a unique value instead of a class. The retry period
is determined by the (fast) poll interval.
- return 'Finish' instead of None for finishing the machine. returning
None for state function is now an error, as this might happen
easily inadvertently.
Change-Id: Icb31367442f10e98be69af3e05a84f12ce5cc966
