Skip to content
Snippets Groups Projects
  1. Jun 21, 2020
    • jonerlin's avatar
      Fine tune btif config mechanism to limit the number of section in bt config... · b9e10460
      jonerlin authored
      Fine tune btif config mechanism to limit the number of section in bt config cache and improve the bt config clone performance
      
      * Bt config cache will grow bigger and bigger without limitation and
      cause many Bluetooth stability issues
      * Do not clone the whole config cache before writing config cache
      into file, clone paired devices section instead.
      * Implement BtifConfigCache class to handle bt devices config add,
      remove, paired, unpaired,..etc.
      
      Bug: 143515989
      Test: 1. Add BtifConfigCacheTest unit tests to test bt config cache
      setup, paire with a new devices, unpaire with a devices, remove the
      sections with specific key.
      2. Checking the paired devices information still existed in settings
      after Bluetooth disable or device reboot.
      3. Checking the paired devices information still saved in btif_config
      file after Bluetooth disable and device reboot.
      4. Remove btif_confing files and enable Bluetooth, checking the
      btif_config file can be created and included adapter information
      normally.
      5. Checking the paired devices information be clean in btif_config
      file after device factory reset
      6. keep BLE scanning, then pair with a bluetooth device, connect a
      bluetooth device, disable Bluetooth, checking the function work
      normally.
      Tag: #stability
      
      Change-Id: Ide1ba1b6daaf9d102ab96976cbf28e65031423d0
      b9e10460
  2. Jun 19, 2020
  3. Jun 18, 2020
    • Martin Brabham's avatar
      Cert: Add new functionality to cert/run · 375e693d
      Martin Brabham authored
      Place the venv in ashmem by default.
      This decreases the --clean build runtime
      by ~40%.  This also decreases the regular build runtime
      by ~10%.
      
      The build time that is measured is the time for the
      python/cert changes to happen.  This isn't accounting
      for building C++ files and objects into intermediates
      and target.
      
      Please note, if you lose power or reboot you will lose
      data in ashmem.
      
      Added a flag '--gotta-go-slow' for those that want the venv
      to remain on disk.
      
      Added a flag '--speed-hax' as a bucket to add non default
      speed improvements. Currently, this will move HostOnlyCert
      to ashmem which decreases clean builds by 54% and
      non clean builds by 16% when coupled
      with default gotta go fast feature.
      
      Bug: 159358844
      Test: cert/run --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Test: cert/run --clean --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Test: cert/run --gotta-go-slow --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Test: cert/run --clean --gotta-go-slow --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Test: cert/run --speed-hax --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Test: cert/run --gotta-go-slow --speed-hax --host --verbose --test_filter=SecurityTest:test_dut_initiated_no_input_no_output_no_input_no_output
      Tag: #gd-refactor
      Change-Id: Ibdf42f0cf7c2e9a32e3360f2f071443722796a4f
      375e693d
    • Sal Savage's avatar
      Don't query for metadata on playing status · 2ffe6420
      Sal Savage authored
      These removed lines of code were for a problem that no longer seems to
      exist with other phones. It was a band-aid at best, having us pull
      metadaa on 'playing status' when a phone really should have just been
      sending the update to us anyway. Pulling metadata here causes a bad user
      experience with cover artwork where the new metadata is interpreted as a
      track change in Java land and cover artwork can be redownloaded. This
      happens primarily with non-database aware players after a BIP reconnect
      where we're guarenteed nothing about the handle we receive.
      
      Tag: #compatibility
      Bug: 158232844
      Test: atest BluetoothInstrumentationTests
      Change-Id: Idb88e7c1e1553319cb26eba2c18e3cd2bbd552cc
      2ffe6420
    • Chienyuan's avatar
      gd HCI: Handle LE white list when cancel connecion · 6e1085f9
      Chienyuan authored
      Tag: #gd-refactor
      Bug: 152346341
      Test: gd/cert/run --host
      Test: atest bluetooth_test_gd
      Change-Id: I1067695121c8e6f0336450b58c82cc5700a01cfc
      6e1085f9
    • Treehugger Robot's avatar
    • Treehugger Robot's avatar
      141ae2d1
    • Jack He's avatar
      LruCache: Improve efficieny and ease of use of APIs · ab58405b
      Jack He authored
      * Add a Find() function that returns the pointer to the value associated
        with a key, further changes to the value using that poitner does not
        warm up the cache
      * Remove eviction callback, but instead return an optional evicted node
        when Put() evicts a cold node. This prevents potential deadlock when
        calling LruCache methods in the callback
      * HasKey() is not zero-copy
      * Get() calls Find()
      * Add unit tests for these new features
      * Modify MetricIdAllocator to use these new features
      
      Bug: 143515989
      Test: atest --host bluetooth_test_common
      Change-Id: I9071c86a9041e5c95b349824889ccedf9f9c18dc
      ab58405b
    • Treehugger Robot's avatar
      1aa0ff37
  4. Jun 17, 2020
  5. Jun 16, 2020
  6. Jun 15, 2020
Loading