d19ffd4f2ca4130c41dd5f6baddad1830598ecbc
This reverts commit c83d3e4b72.
While the above patch does help deal with the issue of "lost" kernel
alarms, experiments found it to also cause the unnecessary duplicate
setting of non-lost alarms in other devices.
This leads to a situation where the CPU constantly tries to sleep, but
occasionally keeps failing to do so, because of a collection of these
duplicate alarms constantly activating the "alarm" kernel wakeup source
, presumably due to them being triggered.
These might have accumulated over several iterations when the kernel
alarms were set again (via rescheduleKernelAlarmsLocked()).
To prevent the CPU thrashing (which heavily drains power on more
battery-constrained devices, and increases the number of alarms being
set), the aforementioned commit is backed out.
Bug: 20136729
Change-Id: I2a189cb303deb78a2faf476b0e789445407f3e11
am
2adefefa: am b6e3f930: am 2bbcc253: am b36ace74: Merge "Check that the parcel contained the expected amount of region data. DO NOT MERGE" into lmp-dev
am
5fed49f7: am 4d9fcae2: Merge "Allow for a NoOp implementation of DrmManagerClientImpl" into klp-modular-dev
am
33e54563: am 4b125836: Merge "Added function to render a drawable in all available states" into lmp-mr1-dev
…
Description
No description provided
Languages
Java
73.7%
Kotlin
14%
PowerBuilder
5.8%
C++
5.2%
AIDL
1%