Matter Switch: Attempt re-profiling on device software updates #2466
Conversation
|
Invitation URL: |
|
Minimum allowed coverage is Generated by 🐒 cobertura-action against 199f355 |
| SwitchFields.DIMMABLE_LIGHT_ID = 0x0101 | ||
| SwitchFields.COLOR_TEMPERATURE_LIGHT_ID = 0x010C | ||
| SwitchFields.EXTENDED_COLOR_LIGHT_ID = 0x010D | ||
| SwitchFields.ON_OFF_PLUG_IN_UNIT_ID = 0x010A |
There was a problem hiding this comment.
can be reverted, but I changed the device type fields to more closely match the spec-defined device types (ON_OFF_PLUG -> ON_OFF_PLUG_IN_UNIT) and I cut the DEVICE_TYPE out of the ones that had it, since it's kinda self-explanatory. If we think it's not, I might move them into their own file with a DeviceType header or something.
| local main_endpoint = switch_utils.find_default_endpoint(device) | ||
| local profile_name = nil | ||
|
|
||
| local server_onoff_eps = device:get_endpoints(clusters.OnOff.ID, { cluster_type = "SERVER" }) | ||
| if #server_onoff_eps > 0 then | ||
| SwitchDeviceConfiguration.create_child_switch_devices(driver, device, server_onoff_eps, main_endpoint) | ||
| -- workaround: finds a profile for devices of the Light Switch device type set that break spec and implement OnOff as 'server' instead of 'client'. | ||
| -- note: since the Light Switch device set isn't supported, these devices join as a matter-thing. | ||
| if switch_utils.detect_matter_thing(device) then | ||
| profile_name = SwitchDeviceConfiguration.match_light_switch_device_profile(device, main_endpoint) | ||
| end | ||
| end |
There was a problem hiding this comment.
This may look like a change in functionality but it is not.
When removing the initialize_buttons_and_switches helper function, it became clear there was some formatting that no longer made a lot of sense. Aka, the semi-arbitrary profile_found = true statements and the
if profile_found then
return
end
statement based around those declarations that had been in place since more logic was hidden by the helper function.
So what I did was remove that profile_found logic, and now have match_light_switch_device_profile use the same try_metadata_update as everything else (at the bottom of match_profile), only returning the profile name instead of doing an actual update.
I would have done the same for the button logic, but it gets a little messy due to the electrical energy stuff. This will be fixed though in the powerTopology PR, so we can streamline that too. I think having one point of contact with try_metadata_update is clearer, catches more edge cases by default, and is easier to maintain, which is why I'm moving the logic in that direction.
1 participant
Description of Change
We store data related to the device software version (saved from the BasicInformation cluster's softwareVersion attribute) on the driver st_store. This leverages that version to attempt a re-profiling if this is ever updated. The same data is what is leveraged by the hub to re-interview a device on a device update as well.
I also took the opportunity to remove the
initialize_switch_and_buttonshelper function inmatch_profile. This seemed like an extra layer of complexity that was no longer required at this point. Can revert if wanted but it changes no functionality so I see it as a nice change.Summary of Completed Tests