Quick install skips a lot of the normal install steps in order to
dramatically reduce the installation time [eg Twitter normally takes
20s to install. But, installs in under 2.5s with quick install]
The specific optimizations [with caveats]:
1. Use the JIT. Although the oat file is technically created, it
only contains the exploded contents of the APK and does not contain
pre-compiled native binary code. While this improves install time,
it impacts app execution. [saves 17s]
2. Bypass Play verification. Play normally verifies all installs
to ensure we're not installing malware. But, it can take multiple
seconds for Play to collect and send package information to our
backend servers. [saves 2.7s]
3. Reduce JAR file verification. Due to the structure of the JAR
certs, we cannot completely bypass JAR processing. However we skip
the step of verifying every manifest entry. [saves 1.3s]
NOTE: #2 and #3 will only occur on eng/user-debug builds.
Bug: 22848361
Change-Id: I48e77595ad5c13a9534fdb06da67ba8dae2797fb
This mode turns on after the screen has been off for 15 minutes,
and then cycles through 15 minutes of idle and 1 minute of
maintenance, ragardless of whether the device is moving around.
It currently only impacts network access and sync/job scheduling.
It does not remove access to wake locks or alarms for any apps.
It also doesn't report in the public API that the device is in
idle mode (since it isn't modifying the behavior of the power
manager) -- this is probably what we desire, since we don't want
stuff like GCM to be reporting these frequent changes.
We'll probably at least want to have the alarm manager do some
kind of more aggressive batching of alarms in this most (not allowing
more than one wakeup every minute?). That's for the future.
Also updated batterystats to include this new information, which
means the format of some of the data has changed -- device_idle
is no longer a flag, but an enum of (off, light, full), and there
is no information about time spent in light modes.
Also added new data about the maximum duration spent in both light
and full idle modes, to get a better understanding of how those
are behaving.
And did a little cleanup of DeviceIdleController, removing the
sensing alarm which was redundant with the regular alarm.
Change-Id: Ibeea6659577dc02deff58f048f97fcd9b0223307
An old optimization in ViewRoot prevents updating a window surface
while a window animation is playing. SystemUI and other small system
components that blend these animations disable this for a smoother
experience. Disable it in ResolverActivity as well.
Bug 24989381
Change-Id: Iac7d1c7b1101ed8d2bc4c3557277a773ce871beb
In rare cases, we might have created a network policy before an IMSI
was available. Because this policy is persisted, and we incorrectly
think that it always applies, we end up annoying the user when data
usage goes over the 2GB default warning threshold.
This patch fixes the network matching logic to ignore these empty
network policies when present.
Bug: 24972775
Change-Id: Id26499b6716121dddf0f2c05b848b0bed5995e72
Bug 24993183
The bounds for the drawable weren't being set, so the
drawable was never drawn to the correct size.
Change-Id: I992951f28ee6fac2ce8752d2dc0e23f51d3cf584
(cherry picked from commit f2a3703696)
Remove the partial fix [it did not work for child fragment managers]
and replace with a more general fix that works with all fragments.
Bug: 23838271
Change-Id: I88b465f6a06a6ad627b9651b9e2eea41fae08972
EXIT event does not have to udpate the pointer icon shape, because
when it immediately moves to another window, its HOVER_ENTER will
handle the icon shape.
Also HOVER_EXIT can happen after the HOVER_ENTER of the new window,
and in that case updating pointer icon at HOVER_EXIT will overwrite
the pointer icon for the new window.
Bug: 24415739
Change-Id: I08fc72cc69bbc3a6eef36d501d93e8e9ad36df85
