Refactor some static fields into a holder. This allows to statically
initialize the class in the boot image. It also has some minimal
positive impact on the class size, a net benefit when tracing is
not used.
Bug: 27248115
Change-Id: I0cdfb6acf916ad8d9a7d82bba32929dba15d4200
Apps making calls into the system server may end up persisting
internal state or making security decisions based on the perceived
success or failure of a call, or the default values returned.
The reality is that if the system process just died, init will be
along shortly to kill all running apps, so we should have no problem
rethrowing the RemoteException as a RuntimeException.
Bug: 27364859
Change-Id: If632cc36f68cd399a34de513b2d979abaa36fcaa
Encapsulating the logic to toggle multiwindow mode from recents,
and plumbing it through to accessibility global actions. Sending
accessibility events when windows bounds change. Exposing the
dock divider window type to accessibility services.
Bug: 27250995
Change-Id: Ib7491f1f853dc7f01bf5c5a4ac1f914f55d0608a
With this change ApplicationLoaders.getClassLoader()
creates linker-namespace for the classloader at the
construction time. Before this change the namespace
was created on first load of a jni library.
With this change we ensure that every classloader
has initialized namespace associated with it.
As an additional advantage we now can avoid storing
namespace-specific fields in the classloaders.
Bug: http://b/27189432
Bug: http://b/22548808
Change-Id: I3b160bd478a55171008682c40b2ebc13bdbd9882
This reverts most of commit 21d361806c, which
broke too many expectations in both apps and framework widgets. We need to
find a safer way to handle the exit transition callback.
Bug: 27359366
Bug: 27353218
Change-Id: I769ac7a25cb900c50e857839ca0563fbc5a3cb16
We now have a fixed array of job service contexts, which doesn't
vary by build configuration. Instead, we keep track of the maximum
number of concurrent jobs we want to allow to run, and don't
make use of a context if it would put us over that limit.
The available contexts is now 8 (the largest used to be 6), although
the maximum we will normally schedule is still 6. We have the other
two around only for use by the current foreground app, to allow it
to schedule work while the user is in it, even if we have reached
our normal limit on the number of concurrent jobs.
The maximum number of concurrent jobs varies based on the memory
state of the device, from 6 (if memory is normal) down to 1
(if memory is critical). We aren't yet trying to stop all jobs
if memory gets lower than critical.
Instead of just keeping track of whether a uid is in the foreground,
we now track whether it is the top as well. Only the top uid
can schedule additional jobs above the current limit.
Also improved some of the dumpsys output.
Change-Id: Icc95e42231a806f0bfa3e2f99ccc2b85cefac320
