REPRODUCIBILITY (% or how often): Always
BUILD ID = OS VERSION (Settings > About product): 4.1.0.24
HARDWARE (XA2, Xperia 10…): Xperia X
UI LANGUAGE: Swedish
REGRESSION: (compared to previous public release: Yes, No, ?): Yes
DESCRIPTION:
I experience higher power consumption since updating from 4.0. There have been reports connecting this to WLAN use, but that is not the issue here, as far as I can tell.
I have tested as below to see if I can find a culprit app, to no avail. It looks more as an OS problem.
Please advice if there is anything more I should test.
PRECONDITIONS:
In order to provide consistent data, I have
Charged to 80%
Switched off all radios (mobile, WLAN, BT) and GPS
Powered off the device
Powered on the device
STEPS TO REPRODUCE:
Setup as above
Start zero or more apps
Show home screen
Leave device idle for 10+ hours
EXPECTED RESULT:
A drop in battery around 0.3%/h with four apps idle. Several days of light device use on a charge.
The four apps were Settings, Calendar, Situations and Pure Maps.
ADDITIONAL INFORMATION:
Under Koli 4.0.1 with these four apps “running” (i.e. idle), I could expect about the same discharge rate as Kvarken 4.1.0.24 shows with no apps at all.
It seems more power is consumed when I actually use the apps, in comparison with Koli 4.0.1, but that is hard to measure. I get 0.5-2 days of use from a charge now.
It seems apps are slower to start and more sluggish to use, in comparison with Koli 4.0.1, which also hard to measure.
POSSIBLY RELATED:
(But please note that my testing is with WLAN off.)
For me, this has been going for a few releases now. Using the original top tool (before the now more useless top got merged in), it was clear that systemd-logind was using the most CPU time, followed by voicecall-ui if I had been calling.
systemdatascope (or rather collectd) has recently gained per-process monitoring.
I suggest adding all of the suspected processes to this and leave it collecting for some time.
Then plot it with systemdatascope.
Note that it scaled graphs down to ‘millipercent’ so be sure to read them correctly.
Also note that you may want to use collectd’s ProcessMatch rather that Process directive for apps that apeear more than once in ps (like all the invoker-started apps).
I tried monitoring my top-used apps one by one, but didn’t get very far. It looked like there was suspiciously little sleeping going on, though.
Then, after receiving some updates, things now look pretty normal powerwise since several days. App change logs show that Battery Buddy got a relevant update - an app I had forgotten about, because I don’t start or interact with it. (There’s a background service that starts automatically.)
So, battery consumption is good again and I think I know why. Big thanks to @direc85 for your work on Battery Buddy!
Sometimes. But it is also in part due to community apps not sleeping very well.
Still, I find the systemd-logind at the top of the list in CPU time suspicious. I haven’t gotten around to investigating it yet though.