Historically RemoteInputConnectionImpl has been calling on

  InputMethodManager#isActive()

to see if the InputConnection is considered to be active or inactive
when actually handling each incoming InputConnection operation.

However, InputMethodManager#isActive() is also known to be tricky
because it internally calls InputMethodManager#checkFocus(), which can
have non-trivial side effects.

With this CL, RemoteInputConnectionImpl keeps maintains its own
boolean state on whether #deactivate() is already called or not so
that it does not need to rely on IMM#isActive() any more.

For 99% cases there must be no observable behavior change, and for the
remaining 1% cases the new behavior should be more easily
understandable.

Bug: 291826769
Test: atest CtsInputMethodTestCases:InputConnectionHandlerTest#testInputConnectionSideEffect
Merged-In: I2fb9c549da19ff01e7cc3fd8bfc1f9c19aa0f0a8
Change-Id: I2fb9c549da19ff01e7cc3fd8bfc1f9c19aa0f0a8

With my previous CLs [1][2], we can generally assume the following
things.

 - InputConnection#closeConnection() always gets called when it
   becomes invalidated, including the case when the IME has switched
   to another IME client.

 - Each implementation of InputConnection#closeConnection() is
   expected to do appropriate clean up including finishing composing
   text. {Base,Editable}InputConnection actually do this correctly.

With that it should be safe for #finishComposingText*() to follow the
same pattern about isActive().

 [1]: I234309c5880c9fe0b299b8bd0f8862796d4dda0d
      9f9afe52
 [2]: If2a03bc84d318775fd4a197fa43acde086eda442
      aaa38c9f

Bug: 35301295
Bug: 291826769
Test: presubmit
Merged-In: If913701bf8555f5d76fceb272e38a99f5e243bbe
Change-Id: If913701bf8555f5d76fceb272e38a99f5e243bbe

It is guaranteed that the following two boolean expressions have
always the same value observed outside from
RemoteInputConnectionImpl#mLock.

  A. RemoteInputConnectionImpl#mFinished
  B. RemoteInputConnectionImpl#mInputConnection != null

With that we should be able to simply merge them into an atomic
reference object

  AtomicReference<InputConnection>

without requiring RemoteInputConnectionImpl#mLock as a lock object.

This CL does so as a preparation to clean up RemoteInputConnectionImpl
for Bug 291826769.

There should be no observable behavior change, except for the fact
that

  RemoteInputConnectionImpl#dumpDebug()

no longer blocks other operations that required mLock, which is kind
of out of our original intention.

Bug: 291826769
Test: presubmit
Merged-In: I69fa2c81670f84be3dd4a808262758a803f69dfc
Change-Id: I69fa2c81670f84be3dd4a808262758a803f69dfc

To make the embedded window of SplashScreenView non-focusable, needs
to add the window flag separately from SurfaceControlViewHost#setView

Bug: 319369153
Test: verify the embedded window of splash screen won't grant focus
Change-Id: Ic9b48dcc4de25e681405b3dd78468597a8a7c627

* changes:
  Add recovery permission to Settings app
  Add 16k option key in settings

During development the Wi-Fi service might not work, let's handle this
case so the Setting app doesn't crash when opening the about phone
page

Logic borrowed from the Bluetooth controller

Change-Id: Id8f6ba5d7cbac5780363612b50848ecfea10bc04

This CL integrates the IpSecPacketLossDetector into VCN
route selection. With this change, when a data stall is
reported, the network candidate will be penalized and
then deprioritized during network selection. A network
candidate will stop being penalized until it hitting a timeout
or passing the validation.

Bug: 282996138
Test: atest FrameworksVcnTests(new tests), CtsVcnTestCases
Change-Id: Ifabd6fdea1d5a4fea40cf929dbab7c26d37274ac

Bug: 307556883
Test: n/a
Change-Id: I5001247c342f23b07e1355db7e50ad90cce66b07

syntax from go/android-codereviews#include-an-owners-file-from-a-sub-directory-within-the-same-git-project

Change-Id: Ia02a5efdda1c2dea65753a63cbd2e78e56bac6eb

When building SDKs that include some, but not all @FlaggedApis, a few
Javadoc @see/@link references will point to objects that are
inaccessible, which will trigger Lint 101 errors.

The correct fix is to update the Javadoc. However, to unblock the SDK
build for now, temporarily hide Lint 101 errors.

Bug: 319541666
Test: m sdk dist
Change-Id: Icbd1d99d5fcd6cee37bf588ef45443e7e946f742

After checking, the modules under these directories can not be built
because the upper-level directories lack Android.mk.
Even if these Android.mk are included, they cannot be tested correctly
with atest.

Bug: 319568429
Test: N/A
Change-Id: I797d2327e340fcd38cde03dd0f520ec756be5fc3

After checking, the modules under these directories can not be built
because the upper-level directories lack Android.mk.
In addition, the format of the Android.mk in these directories is also
incorrect, and if it is included, it will cause the build to break.

Bug: 319565408
Test: N/A
Change-Id: I1a7b2fc2ad78bcbcd6cd3f65c0a5043656711dae

The two major roles of the full api surface stubs (i.e.
android_*_stubs_current) are:
- Appended to classpath when a module sets sdk_version
- Input to generating hiddenapi flags and determine the accessibility of
  the api at runtime

With Trunk Stable, all flagged apis that are not enabled should not be
accessible at runtime. In other words, the currently existing full api
surface stubs that include both non flagged and flagged apis should not
be used to generate the hiddenapi flags.

To resolve this problem, this change introduces full api surface stubs
of the "exportable" stubs (i.e. non flagged apis + flagged apis enabled
by release configuration).

The new modules are named similarly to the currently existing full api
surface stubs, but appends the "_exportable" prefix to differentiate
with the currently existing modules. These modules are solely used for
hiddenapi flags generation, and the currently existing modules continue
to serve the first aforementioned role.

The new modules will toggle between the from-text stubs module and the
from-source stubs module when the exportable stubs can be generated from
the from-text stubs generating `java_api_library` module. For now, these
modules are always built from source.

Test: m android_stubs_current_exportable
Bug: 317426356
Change-Id: I1f6f738ae9c5ebf151a1554aa1e357ef0092c89c

This CL creates an abstract class NetworkMetricMonitor and a
subclass IpSecPacketLossDetector. IpSecPacketLossDetector
is responsible for continuously monitoring IPsec packet loss
and reporting to the caller when the data loss rate surpass
a threshold. This detector allows the VCN to switch from
the underlying network that has bad quality.

This CL also defines new carrier configs that specify the
data loss threshold and detecting frequency

Bug: 282996138
Test: atest FrameworksVcnTests(new tests), CtsVcnTestCases
Change-Id: I5f4b8e1821bdbb37f0a6de6e1584f8e3c87273f3

Bug: 307556883
Test: n/a
Change-Id: I2836908a00981d54e884a495a662bc293219c0e4

Allow specifying default combined api properties. This would be currently used to control the enablement of new mainline modules

Bug:b/289203818
Test: verified locally
Change-Id: I9415140453ce04efd5461cb8684c70e987810ecc

"CameraTooTests" requires the prior installation of "CameraToo" to run
correctly.
Set "CameraToo" as `data` for "CameraTooTests" and add
"AndroidTest.xml" to it.

Bug: 319184564
Test: atest CameraTooTests
Change-Id: I105fc12c7e63e534b0744184876e3526f69021d0

* changes:
  Adjust the NetworkRequest configuration used by SecurityController
  Design of the VPN icon reflects the invalidated status
  Update the VPN icon to reflect the validation status of the VPN network

The package name was changed. Change the use-sites to reflect that.

Test: TreeHugger
Change-Id: I039d02f599692efd51ca8a3316338914eb30a3b9