Fixed
- Node-RED could crash after a deploy with "This callback function has
already been called by someone else; it can only be called one time"
(reported in the homematic-forum). The homematic devices node did not stop
when Node-RED stopped it: a deploy while it was still waiting for the
CCU's device list let the old node publish its devices anyway, after the
new one had started. Every HomeKit device then had its handlers twice,
each read and write ran through both, and the first write the CCU did not
accept terminated Node-RED. The node now stops waiting and publishing when
it is stopped and lets go of the CCU connection, and a redeploy no longer
adds a device's identify handler a second time. If Node-RED did crash this
way, a restart of Node-RED (not a deploy) cleared the doubled handlers. - HmIP blinds and shutters: the second and third virtual receiver (when
enabled in the device list) moved the first one. The command went to
the channel address with the channel number appended twice (e.g.
…:4:5instead of…:5), which the CCU reads as the first receiver;
the log showed such addresses for the first receiver as well
(…:4:4). Position and tilt are sent to the receiver's own channel now.