This reverts #116200
We changed the polling logic to avoid polling if all chars are marked as watchable
to avoid crashing the firmware on a very limited set of devices as it was
more in line with what iOS does. In the end, the user ended up replacing
the device in #116143 because it turned out to be unreliable in other
ways. The vendor has since issued a firmware update that may resolve
the problem with all of these devices.
In practice it turns out many more devices
report that chars are evented and never send events. After a few months
of data and reports the trade-off does not seem worth it since
users are having to set up manual polling on a wide range of
devices. The amount of devices with evented chars that do not
actually send state vastly exceeds the number of devices that
might crash if they are polled too often so restore the previous
behavior
fixes#138561fixes#100331fixes#124529fixes#123456fixes#130763fixes#124099fixes#124916fixes#135434fixes#125273fixes#124099fixes#119617
* fix(alert): check can_ack prior to acking
* fix(alert): add test for when can_acknowledge=False
* fix(alert): warn on can_ack blocking an ack
* Raise error when trying to acknowledge alert with can_acknowledge set to False
* Rewrite can_ack check as guard
Co-authored-by: Franck Nijhof <frenck@frenck.nl>
* Make can_ack service error msg human readable because it will show up in the UI
* format with ruff
* Make pytest aware of service error when acking an unackable alert
---------
Co-authored-by: Franck Nijhof <frenck@frenck.nl>
fix sync date for multiple devices
do not use handler for explicit update devices as internal communication lib do not provide which device is updated
use ha update loop
copy data object to prevent rewrite data from internal lib
allow more time to process response before log warning about long wait for response and make log message more clear
Previously, when the source sensor's state remains constant, the derivative
sensor repeats its latest value indefinitely.
This patch fixes this bug by consuming the state_reported event and updating
the sensor's output even when the source sensor doesn't change its state.
* tweaks
* mysql
* mysql
* Update homeassistant/components/recorder/history/modern.py
* Update homeassistant/components/recorder/history/modern.py
* Update homeassistant/components/recorder/const.py
* Update homeassistant/components/recorder/statistics.py
* Apply suggestions from code review
* mysql
* mysql
* cover
* make sure db is fully init on old schema
* fixes
* fixes
* coverage
* coverage
* coverage
* s/slow_dependant_subquery/slow_dependent_subquery/g
* reword
* comment that callers are responsible for staying under the limit
* comment that callers are responsible for staying under the limit
* switch to kwargs
* reduce branching complexity
* split stats query
* preen
* split tests
* split tests
* Update NS integration to show first next train instead of just the first.
* Handle no first or next trip.
* Remove debug statement.
* Remove seconds and revert back to minutes.
* Make use of dt_util.now().
* Fix issue with next train if no first train.
* Bump Music Assistant client to 1.1.0
* Add some casts to help mypy
* Add handling of the new media types in Music Assistant
* mypy cleanup
* lint
* update snapshot
* Adjust tests
---------
Co-authored-by: Franck Nijhof <git@frenck.dev>
Brett Adams
Joost Lekkerkerker
J. Nick Koston
StaleLoafOfBread
Filip Agh
Marcel van der Veldt
Juan Grande
Guido Schmitz
G Johansson
Martreides
M-A
LaithBudairi
Bram Kragten
Robert Resch
Robert Svensson
Jan Bouwhuis
Josef Zweck
Erik Montnemery
Ivan Lopez Hernandez
Jan-Philipp Benecke
J. Diego Rodríguez Royo
Michael Hansen
Jeef
starkillerOG
Paulus Schoutsen