Initial pass at leak-detection on the unit
tests. Due to quirks in jemalloc we need
to run the leak check on both the main thread
and the RT thread to get decent leak coverage.
Change-Id: I0dce754aef5593f1ef5c3bed86e1168437a34b74
This will let it play nicely with heads up.
- Move to QS Container.
- QS Edit is always full height (some layout hacks to do this)
- Always draw QS customizer on top when animating
- Block all panel scrolling while QS edit is open (all touches
go to editing)
- Instantaneously change the height of the QS container at
start/end of animation as needed
Bug: 26969293
Change-Id: Iedc6f5aaf659dcc6750972eae5f69cc0cd2df844
KeyChain can throw an assertion error if
is not around, don't allow that to take down
system.
Bug: 27518175
Change-Id: I99418dfb65c58d3e07cbda91860cdb493b96a836
This CL changes nothing except for deleting redundant type parameters by
using diamond operator in TextServicesManagerService (TSMS).
No behavior change is intended.
Bug: 27456430
Change-Id: I18708fe44d2d9edfef8c1cf7cffef1bb0b5c003a
This reverts commit 9e7a1c9824cffca32fd7d58bb13bd3416ab32091.
Seems that that CL causes ArrayIndexOutOfBoundsException when
initializing InputMethodManagerService, which results in an infinite
boot animation.
Bug: 27129703
Bug: 27348943
Change-Id: I474a87876670ac018c675ac7b4608e90fbb2434b
We allow the home stack to be visible right behind the fullscreen
stack if all the activities fullscreen stack are translucent and
visible. However, if we are starting an acitvity it isn't visible
yet so our check for translucent activities in the fullscreen
stack will return true since there is nothing visible. This will
cause the home stack to the visible and the visiblilty of its app
token set to true in window manager and it been factored into the
transition animation.
In addition to checking if the activity is visible, we now also
check if it is the starting activity when trying to determine if
the stack is translucent.
Bug: 27448511
Change-Id: Icbbd57c0632cc5389c7ac894800a41f75d4bc450
The motivation of this CL is enhance the default IME subtype enabling
algorithm. The new approach is done by score based algorithm. The
design of the matching score is determined as follows:
- The matching score for the each two locale is up to 3 and determined
as follows:
- Score of 3 : matches all language, script and country.
- Score of 2 : matches the language and script.
- Score of 1 : matches the language regardless of country.
- Score of 0 : doesn't match the language regardless of script and
country.
- All locales are fully expanded before matching by addLikelySubtags in
ICU.
Bug: 27129703
Change-Id: I5bb1bd8cdb9096d516d60beb9936e55bf2b757ae
If the list of providers for the application being bound is empty,
don't bother trying to register them.
Bug: 27510621
Change-Id: I921fb70a4c97433fa123fa88e170745ee013ee35
For instrumentation test apks, it should not statically include
Java libraries that are already present in the app module being
instrumented. They only need to be on classpath during compile
time.
Bug: 26689464
Change-Id: I68d4be95e30ae70bbab23d016a87eb085c10cc50
RingtoneManager is the public API that everyone should be going
through, and it now has the side-effect of caching the set ringtone
to make it available before CE storage is unlocked.
Bug: 27435331
Change-Id: I30ed4e2df2ef1e4fd47f947c70845aaa74356384
When setting default ringtones, RingtoneManager now caches the
selected media for playback before the device is unlocked. However,
this API hasn't historically required the caller to hold storage
permissions.
To keep this working, we attempt to delegate ringtone access over
through RingtonePlayer, which is what we do for playback. However,
because we're caching the real ringtone bits now, we need to be much
more careful about the PFDs we're willing to return. This change
requires that they be in external storage, and that they have the
ringtone/alarm/notification bit set.
Bug: 27366059
Change-Id: I59c2adc1d1250a3eac281f190f35a7cb3119967b
If Bluetooth was enabled at boot time, but the actual Bluetooth
package wasn't encryption aware, we need to kick off another enabled
pass once the user is unlocked.
Bug: 27326711
Change-Id: I511c4fd8567792e5bc4d49fa95929fd87edd0fbf
It's in the system process, so whatever it has to say is probably
important enough to stick around.
Change-Id: Iaeca99d6b9c4881c53ab89216b5128e345a5ff0f
