1. 03 Aug, 2017 33 commits
  2. 02 Aug, 2017 7 commits
    • Jeremy Roman's avatar
      Use V8AtomicString for strings used as property keys. · 5d6d0234
      Jeremy Roman authored
      V8 always internalizes strings when doing property lookup, so this can only
      save the work of heap-allocating a new string that will just be looked up
      later, and likely already exists in the internalized string table (e.g.
      because it is the key of a property, or because it occurs as a string literal).
      
      Found mainly by searching for string literals passed to V8String. Other
      similar cases may yet remain.
      
      No perf measurement was done, but this seems like it's easily the nicer
      thing to do in these cases.
      
      Change-Id: Idd645a83495f5ec1323f3ce87548d0f2a16aca19
      Reviewed-on: https://chromium-review.googlesource.com/598540Reviewed-by: default avatarKentaro Hara <haraken@chromium.org>
      Commit-Queue: Jeremy Roman <jbroman@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491561}
      5d6d0234
    • Sunny Sachanandani's avatar
      cc: Skip generating video release sync token if not needed. · 390899a3
      Sunny Sachanandani authored
      I removed the optimizations in SyncPointClientImpl where it'd skip
      generating sync tokens if there was no existing release sync token.
      That caused a 5% cpu time regression in the video benchmarks.
      
      I also added some tests to verify that sync tokens are being generated
      at the right places.
      
      R=jbauman
      BUG=750872
      
      Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel
      Change-Id: If739e3fb6705cf8b0739fc029ed3bd75dc43eabf
      Reviewed-on: https://chromium-review.googlesource.com/596424Reviewed-by: default avatarJohn Bauman <jbauman@chromium.org>
      Commit-Queue: Sunny Sachanandani <sunnyps@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491560}
      390899a3
    • Tsuyoshi Horo's avatar
      [DevTools] Enable network capability in worker unconditionally · ae8e4ee9
      Tsuyoshi Horo authored
      We are planning to start the field trial experiment of
      off-main-thread-fetch optimization.
      
      But in the current implementation, web developers need to enable
      'Network in workers' experiment of DevTools to show the network requests
      from workers in DevTools if the optimization is enabled in their Chrome.
      It is very troublesome for them while debugging PWA.
      
      To avoid this problem, this CL enables the network capability in worker
      unconditionally.
      
      Bug: 443374
      Change-Id: I5ed24996cd0ff30a5bd22ce1067f653a5c68ae9b
      Reviewed-on: https://chromium-review.googlesource.com/595294Reviewed-by: default avatarDmitry Gozman <dgozman@chromium.org>
      Commit-Queue: Tsuyoshi Horo <horo@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491559}
      ae8e4ee9
    • Jeremy Roman's avatar
      Use find-only GetEventListener for removeEventListener. · 41f021e6
      Jeremy Roman authored
      This was erroneously picked up by the Blink reformat; the method name is still
      lowercase at this point, so we were generating a find-and-create call when a
      find-only call suffices.
      
      https://chromium.googlesource.com/chromium/src/+/1c4d759e
      
      Tested by manually inspecting generator output.
      
      Change-Id: Ic7b6a4936b8bb7558e0ca21eddee23ca84348647
      Reviewed-on: https://chromium-review.googlesource.com/599000Reviewed-by: default avatarKentaro Hara <haraken@chromium.org>
      Reviewed-by: default avatarAdithya Srinivasan <adithyas@chromium.org>
      Commit-Queue: Jeremy Roman <jbroman@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491558}
      41f021e6
    • Tsuyoshi Horo's avatar
      Update fieldtrial_testing_config.json for OffMainThreadFetch · f429b926
      Tsuyoshi Horo authored
      Bug: 443374
      Change-Id: I001c3981e17430f8786eec2a7d475c0c994e2f5e
      Reviewed-on: https://chromium-review.googlesource.com/597573Reviewed-by: default avatarJesse Doherty <jwd@chromium.org>
      Commit-Queue: Tsuyoshi Horo <horo@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491557}
      f429b926
    • Miao-chen Chou's avatar
      arc/bluetooth: Fix the powered state synchronization between Chrome and Android · bb12724f
      Miao-chen Chou authored
      Since the observation of the local powered state change and the receiver of
      enable/disable requests from Android is in arc_bluetooth_service are in
      arc_setting_service and arc_bluetooth_bridge separately, the power state can
      get out of sync or restore to the wrong state due to the lack of coordination
      between these two pieces of information.
      
      Issues:
      - The powered state can get out-of-sync between Android and Chrome if the power
        state of adapter is repeatedly toggled from Chrome setting.
      - If the powered state is ON before deep suspend/resume, the state will
        turn to OFF after resume and turn immediately back to ON on Chrome. However
        without waiting for the previous disable intent to finish, arc_setting_service
        sends the enable intent right after while Android Bluetooth statck is still
        processing the disable intent. Also there is a 3~4 seconds delay for bringing
        down to complete on Android Bluetooth stack. Besides, the enable intent
        is treated as a no op if the ongoing intent is disable. So after 3~4 seconds
        Android sends a disable request to Chrome while the powered state is ON on
        Chrome, so the powered state fall to the wrong one, disable, instead.
      
      The changes includes:
       - moving the observation of Bluetooth adapter from arc_settings_service
         to arc_bluetooth_bridge,
       - changing the cycle of observing Bluetooth adapter from living along with ARC
         Bluetooth instance to living along with ARC Bluetooth bridge,
       - adding IntentHelperObserver to listen to the OnInstanceReady event and fire
         the initial powered state of Bluetooth to Android via intent,
       - adding two queues and a timer to track the powered state changes initiated by
         either Chrome or Android and to track the completion of the power
         changes initiated by Chrome,
       - and adding helper functions to enqueue/dequeue the power change intents and
         to compress the toggling of state to reduce the intents sent to Android.
      
      BUG=b:62578573
      TEST=(1) On Android start-up, verify that the powered state is synchronized
           between Android and Chrome.
           (2) After Repeatedly toggling Bluetooth powered state, the state
           remains synchronized between Android and Chrome.
           (3) Put device into deep suspend and resume, the state restores to
           the previous state.
      
      Change-Id: I753839e2bb9228703492ebd23703cf8b67494998
      Reviewed-on: https://chromium-review.googlesource.com/580314
      Commit-Queue: Miao-chen Chou <mcchou@chromium.org>
      Reviewed-by: default avatarLuis Hector Chavez <lhchavez@chromium.org>
      Reviewed-by: default avatarRahul Chaturvedi <rkc@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491556}
      bb12724f
    • Tom Sepez's avatar
      Fix gn gen for ipc_message_lib under use_ozone=true · 77be83bc
      Tom Sepez authored
      Broken at 4ec75161
      
      The busted FYI bot build is:
      https://build.chromium.org/p/chromium.gpu.fyi/builders/Linux%20ChromiumOS%20Ozone%20Builder/builds/27173
      
      TBR=mbarbella@chromium.org
      
      Change-Id: If4cad6ae90bab1c6ae12d65d568fbda8975ef50c
      Reviewed-on: https://chromium-review.googlesource.com/598453
      Commit-Queue: Tom Sepez <tsepez@chromium.org>
      Reviewed-by: default avatarTom Sepez <tsepez@chromium.org>
      Cr-Commit-Position: refs/heads/master@{#491555}
      77be83bc