By default, enabling the HID Device service is only possible
during disabling the HID Host service. If, on the other hand,
the HID Host service was never initialized, then a request to
enable the HID Device service should just be executed
immediately.

Test: manual, on a device without the HID Host profile.
Change-Id: I2ab4b8aec413b66d57eb322d38660f1b4c48ba32

* Codec config internal abstraction:
 - Add new classes A2dpCodecConfig and A2dpCodecs that (will)
   encapsulate all codec-related APIs
 - Add unit tests for the above two classes
 - Add method A2dpCodecConfig.buildCodecConfig(), and use it when
   creating the codec configuration instead of A2DP_InitSource2SinkCodec().
   The new method can build the codec config by taking into account
   optional user codec-related configuration preferences.
 - Use the A2DP codec config API from the hardware/libhardware bt_av.h API
 - Replace enum tA2DP_CODEC_SEP_INDEX with btav_a2dp_codec_index_t
   from the bt_av.h API
 - Move codec-specific functions from stack/include/a2dp_api.h
   and stack/a2dp/a2dp_api.cc to stack/include/a2dp_codec_api.h
   and stack/a2dp/a2dp_codec_config.cc
 - Create a new base class A2dpCodecConfig() to hold some of the
   codec-related state, and implement the corresponding A2dpCodecConfigSbc
   and A2dpCodecConfigSbcSink derived classes.
 - Move A2DP spec-related constants from stack/include/a2dp_api.h
   to stack/include/a2dp_constants.h
 - Move A2DP-related error codes from stack/include/a2dp_api.h
   to stack/include/a2dp_error_codes.h
 - Move A2DP SBC spec-related constants from stack/include/a2dp_sbc.h to
   stack/include/a2dp_sbc_constants.h

* Implement the backend mechanism for handling user (re)configuration of
  A2DP Source codecs as requested via the JNI API calls.
  Also, any codec changes are reported back via JNI API callbacks.
  The current audio parameter selection (sample rate, bits per
  sample, channel mode - mono/stereo) is as follows:
  - If the user selected parameters are acceptable (based on
    local codec capability and the remote Sink capability),
    those parameters are used.
  - Else if the Audio HAL's requested parameters are acceptable,
    those are used.
  - Else if the default settings are acceptable, those are used.
  - Else use the best match among the local and the remote device's
    capabilities.

* Update the mechanism for handling OTA configuration requests from the
  remote Sink device.
  - The OTA prefered codec configuration is ignored if the current
  codec configuration contains explicit user configuration, or if the
  codec configuration for the same codec contains explicit user
  configuration.

* Refactor the Audio HAL <-> Bluetooth audio parameter negotiation
  mechanism:
  The new mechanism gives some flexibility to the Media Framework to
  choose the appropriate audio format (sample rate, bits per sample,
  and channel mode - mono/stereo), and at the same time allows
  the Bluetooth stack to double-check / overwrite the choice.
 - out_get_parameters() on the Audio HAL side asks the Bluetooth stack
   for all currently supported formats (for the current codec),
   and returns them to the Media Framework: sample rate, bits per sample,
   and channel mode (mono/stereo).
 - The first time adev_open_output_stream() is called on the Audio HAL,
   it asks the Bluetooth stack about the audio format currently selected
   by the Bluetooth stack (based on codec negotiation with the Sink device,
   and User Configuration).
 - The second time adev_open_output_stream() is called on the Audio HAL,
   its "config" will eventually contain the audio format selected
   internally by the Media Framework. That audio format is sent to the
   Bluetooth stack.
   If that format is acceptable to the Bluetooth stack, the Bluetooth
   stack will reconfigure itself internally, and will respond back with
   those values. Otherwise, it will respond back with the values that
   should be used instead.

* Misc other fixes and refactoring:
 - Fix the BTA handling of A2DP codec reconfiguration
 - Fix a bug in the implementation of A2DP_BitsSet(), and add the
   approriate unit test. Also, fix the code that was using this function
   incorrectly.
 - The SBC encoder is compiled as a separate library
 - Replace leftover usage of "false" with "FALSE" for macros, and
   vice-versa for variable values.

Test: A2DP streaming to headsets, TestPlans/71390
Bug: 30958229
Change-Id: I440b6126e2250e33b0075f9789dd93154c007c2b

BTA closure doesn't keep the HDR messages any more, test have keep track
of HDR pointers for itself.

Bug: 33947176
Test: test/run_unit_tests.sh  net_test_bta
Change-Id: I32994f248dc2e082bf8bed426a32acf677c58ed5

With this change, when the app requests a connection to a
specific device, it is always "plugged" first.
Unplugging the previous device is left for the app to do.

Test: make
Change-Id: I2e27c12e2ee96d0c311615a3853d8c430f3b2c31

BTA layer for scanner is very complicated. This patch simplifies it:
* get rid of type redeclarations for BTM types
* get rid of trivial *_act methods and call BTM API directly where possible

Bug: 30622771
Test: slra FilterTest
Change-Id: I3899b30074b2abc4a3945c5cc14f1bb40a504876

Attach posted_task to the HDR message, instead of using a queue.

Bug: 33947176
Test: re-run sl4a FilteringTest 3 times with no flakiness
Change-Id: If0b69e82d11eeae52ac8e591bebff634073251f1

This patch adds support for HID Device role in bluedroid stack allowing
DUT to be used as Keyboard or Mouse.

Bug: 33011576
Change-Id: I45b581a54f6c7bbc1f25226715a7ea23e34255c0

This patch improves btu_hcif_send_cmd_with_cb, which will accept
base::Callback instead of function pointer as argument. It will also
carry information about where was packet send from, which can be
displayed when command status is received, greatly improving logs.

The improved method is also used for handling VSC advertising.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertiserTest
Change-Id: I55c3e32f9231c00bfa85f971265809d6f1fecdfb

When BLE_VND_INCLUDED is not set, advertising was not properly
initialized, which resulted in stack crashes. This patch makes sure that
LE advertising is always properly initialized

Bug: 33168939
Test: enable LE advertising with BLE_VND_INCLUDED set to false
Change-Id: I5fa0b3da39da99e750e1f8b6f0452a628094bfc4

p_cond_param->srvc_uuid is of type tBTA_DM_BLE_PF_UUID_COND

Bug: 33910711
Test: sl4a FilteringTest
Change-Id: I6b30e1e441f8256d8c8dbf593f746a3d7810eda6

Using the pointer to the beginning of a union in a member of the union,
which will then be over-written, is a bad idea(TM).

Bug: 33910711
Test: manual
Change-Id: I0b979e493688bf8c02119a2ef6707d6c8e730dcb

Test: mma -j32
Change-Id: Ia1aae4cd5618816b529449844b3a0724e4eb3200

Test: compilation test
Change-Id: I0834431378b015eafbb26c48af342af4a650972b

BTM layer checks if LE is avaliable, and if proper VSC are
avaliable, no need to do additional checks in BTA.

Test: sl4a FilteringTest
Bug: 30622771
Change-Id: I594b152d18f08a46b65930547a991dbb6f188c4b

All calls to filter related code is guarded by check if HCI VSC are
available. There is no need to keep double check.

Test: sl4a FilteringTest
Bug: 30622771
Change-Id: Ib4c53593d536a2985926ef2b725bc69050bdb035

Get rid of BTA state machine states for BLE scanning.

Test: sl4a FilteringTest
Bug: 30622771
Change-Id: I8a879604f413c9980cf64983d8ef6b9a94669295

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: I6185e6926e7363824aa573c7d65f7ab66f954e52

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: Id85504922c21f15bc36ac8bb5e4ab962ee356e3d

When LE Advertising Extension is available on the controller we must use
both extended advertising and scanning commands. Otherwise, according to
the spec, controller will return error.

This patch makes sure that proper avaliable HCI calls are made.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest as scanner
Change-Id: I6906c381f0f758bd24b592390a4ef46e7fae36ed

This patch updates BleAdvertiserHciInterface to be on pair with
Bluetooth 5.0 LE Advertising Extension feature.

Bug: 30622771
Test: net_test_stack_multi_adv native test
Change-Id: Ie0ec8c7f8ed0d7a874e392123b3486c767121140

Make BleAdvertiserHciExtendedImpl report when advertising set is
disabled because of new connection.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: Ic13a31fe4bb92f121a29d540274d13893775a450

This patch implements current BleAdvertiserHciInterface using LE
Advertising extension HCI commands.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertisingTest
Change-Id: Icaaf24d6d02d8a887b3bb57d5b858b929f11c413

Bug: 30622771
Test: sl4a BleScanTest
Change-Id: I55e6c201849a057995c0c6fda1c52af826749922

-- Also fix a bug introduced by change of data types in AVRCP Open state
and AVRCP Browse Open state. Due to change in data type and field
mismatch the status value was being read incorrectly

Bug: b/33750053
Test: Manual test of browse connect
Test: PTS test for UUIDs changed
Change-Id: Ifecf9480a3fabeee8ad8302ebb7cd48f5322dea7

* btif uses AudioTrack from frameworks/av which is under libaudioclient
* Current dependency on libmedia is incorrect as it is an indirect
  dependency to libaudioclient

Bug: 33718447
Test: code compilation, no user visible effect
Change-Id: I7c44095cd282fabf8595015910670193077c1514

This patch converts btgatt_scanner_interface_t struct into
BleScannerInterface class. It also refactors three most important
methods from this interface: RegisterAdvertiser, Scan, and Unregister.
Rest of this interface will be updated in following patches.

Bug: 30622771
Test: sl4a BleScanApiTest
Change-Id: Ie35356f6c3c4f5488514ef55a48a32c93fb21b83

For incoming connections we are not storing the handles in BTIF which
implies that BTIF ignores the message (and so does jni -> java). This is
fixed by sending the handle as part of BTA message on incoming accept.

Other bugs include connection from remote (incoming) and disconnecting
from host. They also depended on not storing the handle properly in
first place.

Bug: b/33555377
Bug: b/30984220

Test: Manual testing of incoming connections
Change-Id: I84950d42aba2d2c975ea86cc8217fd0129cc90e9

* Add unit test for confirm (c1), p1, and p2
* Refactor the code to reduce number of methods
* Refactor c1's code to make it testable

Bug: 32413756
Test: Code compilation, Unit Test, PTS SM Tests
Change-Id: I5cc876b7dd2b21a0780ac9a5236420223df6cddb

Simplify the code to fix a typo.  F_SETFD should have been
F_SETFL.

Test: run unit tests

out/host/linux-x86/nativetest/test-vendor_test_host/test-vendor_test_host \
        --gtest_filter=AsyncManagerSocketTest.*

Change-Id: I215381d6e130428acae7d73d80fefc97cbf64651

Test: run the unit tests
 out/host/linux-x86/nativetest/test-vendor_test_host/test-vendor_test_host \
     --gtest_filter=AsyncManagerSocketTest.*

Change-Id: I8c5e05c5506a067a6148d54dc23bbb931ad612e7

* Replace assert with CHECK
* Remove all NDEBUG definitions
* Remove hacks for BT_LIBCHROME_NDEBUG
* Removed some removed directories from Makefile such as hcis, brcm, rpc

Coccinelle-assisted:

@@
@@

- #include "base/logging.h"
+ #include <base/logging.h>

@ assert_included @
@@

@ base_logging_included @
@@

@ depends on (assert_included && !(base_logging_included)) @
@@

- #include <assert.h>
+ #include <base/logging.h>

@ depends on (assert_included && base_logging_included) @
@@

- #include <assert.h>

@@
expression E;
@@

- assert(E);
+ CHECK(E);

And a bash script:

for file in $(find . -name "*.cc"); do
  spatch --sp-file replace_assert_with_CHECK.cocci --in-place $file
done

The following files are maually edited:
	btif/src/btif_config.cc
	btif/src/btif_avrcp_audio_track.cc
	btif/src/btif_gatt_client.cc
	osi/src/data_dispatcher.cc
	osi/src/reactor.cc
	osi/src/thread.cc
	osi/src/fixed_queue.cc
	osi/src/list.cc
	osi/src/allocation_tracker.cc
	osi/src/alarm.cc
	osi/test/wakelock_test.cc

Bug: 31781465
Test: Code compilation, Unit Tests, BtStressTest, BtFunhausMetricsTest
Change-Id: I21dc10a45be31665e41441b75b0515ed87523988

This patch replace BTA state machine state associated with registration
of new application, together with BTA_GATTC_API_REG_EVT, and
tBTA_GATTC_API_REG. Instead, it uses closure to post registration task,
and callback for registration event.

Bug: 30622771
Test: sl4a ConcurrentBleAdvertiserTest GattConnectTest
Change-Id: I62d68485170ef5472237d62b285353f2c9bc8250

* This function used to take (const uint8_t*) as the second parameter
  and it causes all its calls to cast a (const char*) to this type
* This patch changes it to (const char*) so no cast needed

Coccinelle-assisted:

@ rule1 @
type T;
expression A, B, C;
@@

smp_debug_print_nbyte_little_endian(A,
- (T)
B, C);

and

for file in $(find . -name "*.cc"); do
  spatch --sp-file refactor_smp_print_func.cocci --in-place $file
done

Bug: 33663033
Test: code compilation, no user visible effect
Change-Id: Ibeed0b414514acc4a8e94e47e99117bdb3e454ff

Test: Code still compiles
Change-Id: I4a86cd3b515ad3da8f852f024952e3f7a85d1071

Bug: 30622771
Test: compilation test
Change-Id: I0e502fb9b5b2cc3ccad0aa142a0ec9fedea7d1df