As a temporary solution, this CL introduces gtest as a secondary build
source. For now, gtest can be included in testonly targets as a direct
GN target dependency, by placing the gtest source under "testing/gtest"
at the root of the tree.

Moving forward, we can use solutions like gclient to pull in external
dependencies such as gtest, or we can depend on it being available on
the system.

BUG=21339022,21340407

Change-Id: I80ab3a81a53d7c29d54857a8d0cd9677fed77549

BUG=21339022

Change-Id: I946fad7bf8116054f9081e064748793e79a6f85a

BUG: 21119653
Change-Id: I8e4caf1946f525ea57fadf003ca3fd7d303976cb

BUG=20491586

This is not a comprehensive fix. More work is necessary to ensure the
correct separation of BLE_INCLUDED TRUE/FALSE as well as removing many
of the hacked-up if() statements resulting from this patch.

Change-Id: I1812594feba0bf43369096e472a8b18cde305255

git reset --hard d06cfb98^^
git reset --soft aosp/master
git commit -a

Change-Id: I2939d9bbd825110a42a71008e5255945e38a77e1

These patches will allow us to compile the whole AOSP with clang/llvm
before all the following problems are fixed.
* Suppress warnings on unused variables.
  * Remove unused static variables.
* Use only gcc to compile code that requires gcc atomic functions.
  They should be converted to standard atomic functions later.
* Suppress warnings on redefined typedefs.

BUG: 20765701
Change-Id: Ib1d6e2ef86c255ba87c1d51f9fdeabe5c8de2ada
(cherry picked from commit 134b2f45)

LE low power extended scan feature stack changes
based on Google BLE spec v0.95

Bug: 20029342
Change-Id: I1db2d9bb648295340d0e93ca89d27e1b938d33d3

Change-Id: I0d43b3922b2b5dc8c57511d908f45dfe8d8f1b9c

Change-Id: Iabbb20b111beaaaef8a465f6917b4286e1b93972

Defers the incrementing of the total services count after the allocation
of the instance ID for the service. If the increment is not deferred,
the current service is compared with its own instance in the cached
service list while allocating the instance ID. This leads to an
incorrect instance ID being allocated for the service.

Change-Id: I547888ae0f7015cfdb9b1a995373523699642a6c

- Set the service availability flag during ciev events
- Don't drop AT+COPS and AT+CNUM events on the floor if there is no
  service. Instead, send fake OK responses up the stack.

Bug: 20127723
Change-Id: Id2d9ae8c5df5b5372e50e804e44ed51657b2a4a9

Fix minor conditional check and clean up surrounding code.

Bug: 20717705
Change-Id: I581fda37708062327c86cc0548b2665c9811be25

As a side effect of the memory corruption, the AVRCP metadata
information was not visible.
Apparently, the bug was in a very old (almost unchanged) code,
and it was triggered because the rest of the code has evolved.

Bug: 20108272
Change-Id: I28466edf6a2fbfc31cfb46d74d03b6c0399f8594

The long term key (LTK) was not assigned to the correct struct element
before saving the key to NVRAM.

Change-Id: Icfce82f5b70518e725875771756edde0b9caeb88

Bug: 20717705
Change-Id: I76e76cf33d7c9f4366f921daf2230e4f681098f7

To clarify local/peer key deliniation.

Change-Id: I79573015c80456ce3287acfe5551a6881ff581b2

Two separate issues prevented the correct LTK from being storred in
non-volatile storage and thus prevented an LE device from working after
Bluetooth was restarted:

1. tBTM_SEC_BLE_KEYS contained a field "pltk" which stored the LTK, but
   downstream the field "ltk" in the same structure was used to access
   the key.

2. The structure element p_key_value was not deep-copied for
   BTA_DM_BLE_KEY_EVT events, causing data corruption and instabilities.

This patch addresses both issues.

Bug: 20091926
Change-Id: I9f9481a44bfe83cacd351f9e578451e77b573564

A small bug in btif_read_le_key() would prevent a paired LE device from
being added back into the internal device database after reboot, causing
the security keys not being loaded.

Also added a debug statement to trace keys being loaded.

Bug: 20091926
Change-Id: Id2d6c69497857edc21e2d4c6a67ad9e8dadabcaa

Change-Id: I32d8975f0f6e1ea18a361a19260ffea7a3ec8467

In order to only scan for peripherals when we have pending background
connections, we need to keep track of devices that have been added to
the whitelist.

Bug: 20091926
Change-Id: Id7b1861829d984c150d69371ebd13f7d52d3093c

Change-Id: I22d97303054eccc876c4a9c7c0a50e369ff4fa62

Change-Id: Ifa83bd342458eba33fdd11a9db38d2f9e818b759

Change-Id: Ib7b05e8907497cf3e2f06a94ff25686ccba43a81

When privacy is disabled, the local BDA is not set correctly when ACL
connections are created. This leads to SMP pairing challenges to be sent
with the wrong local BDA (00:00:00:00:00:00).

Bug: 20091926
Change-Id: I554245a3a3cbed72d79aa6c1261c813999019e2b

Change-Id: I0fe4ea75c6448f43f1638614a9a6519fa58962de

With the LE-only mode changes in place, there is a potential race
condition between the audio-gateway (AG) registration and the hands-free
service initialization as the core stack is now initialized before the
hands-free service starts.

Modified from patch submitted by Nitin Arora <niarora@codeaurora.org>.

Bug: 20481584
Change-Id: I78d62329aa5e4113c72fac2d40efa671442bae2d

Bug: 19544388
Change-Id: Ib16878c08345983b3f72b484f2d5e4eb4791f2f9

merge in m-wireless-internal-release history after reset to f5d67e1e

Fixed a problem where A2DP source was initialized twice. Once before
btif was properly initialize and then a second time when the JNI
requests A2DP to be initialized, which would then fail.

Change-Id: Ia5f10a5bda344fe3bd66818a302b6b3f9db32a20

merge in m-wireless-internal-release history after reset to f5d67e1e