After _control_heating() is executed, current_operation() is correctly called but _is_device_active() still reports the old state if the heater switch, at least in my case. The resulting climate state is incorrect until the next refresh, which apparently occurs only when _sensor_changed() gets called (it can be minutes after).
I think the state of the heater switch should be forced to update at the end of _control_heating(), but I don't know how to do that...
A simple sleep() fixes it, but obviously is just a temporary workaround, I'm not really expecting this PR to be actually committed, unless there's no other solution.
* Migrate callbacks to use schedule_update_ha_state
* Migrate MQTT sensor callback to async
* Migrate wemo to not update inside schedule_update_ha_state
* Make MQTT switch async
* Fix nx584 test
* Migrate tellstick callback
* Migrate vera callback
* Alarm control panel - manual: use async callbacks
* Run the switch rest tests that work
* Separate platform and presentation units in climate
* Fix unit tests
Maybe
* Fix unit tests some more
Maybe
* Rename _platform_unit_of_measurement to temperature_unit
* Fix tests for renamed attribute
* Update ecobee to use only range setpoints
* Update nest to only use range setpoints
* Update demo to use only range for ecobee test device
* Update test
* Fetch unit from ecobee
* generic_thermostat did not have state
* generic_thermostat test update
Fabian Affolter
Lev Aronsky
Pascal Vizeli
Gianluca Barbaro
r-jordan
Lewis Juggins
Paulus Schoutsen
Sean Dague
Russell Cloran
John Arild Berentsen