GitLab

Summary of changes available at:
https://chromium.googlesource.com/v8/v8/+log/26c72b54..4e3834b0

Please follow these instructions for assigning/CC'ing issues:
https://github.com/v8/v8/wiki/Triaging%20issues

Please close rolling in case of a roll revert:
https://v8-roll.appspot.com/
This only works with a Google account.

TBR=hablich@chromium.org,machenbach@chromium.org,yangguo@chromium.org,vogelheim@chromium.org

Review URL: https://codereview.chromium.org/1685193002

Cr-Commit-Position: refs/heads/master@{#374767}

This change renames WebRtcLocalAudioRenderer to TrackAudioRenderer and
re-purposes it for rendering all local MediaStream audio tracks, and
non-PeerConnection/WebRTC-sourced remote audio tracks (e.g., Cast
Streaming).  In addition, a number of small functional changes were made
to fix audio glitching, especially surrounding Start/Stop/Mute
operations:

1. Improved accuracy of GetCurrentRenderTime(), based on audio sample
counts rather than using the system clock.

2. Removed the stopping of audio render on mute.  This would cause the
video player to freeze the video unintentionally.

3. Drop/re-create AudioShifter on any pause-then-play, audio format
change, or output device change.  The existing Flush() mechanic was
causing weird time-shifting effects when the audio data flow was
restarted due to retaining the old timing data history.

4. Corrected a timing calculation in the playout_time passed to
AudioShifter::Pull().  Since the playout_time is supposed to be "in the
future," the correct timestamp should be "now plus audio delay" and not
"now minus audio delay."

5. Removed legacy call to GetOptimalBufferSize() for the audio output
buffer size.  After reviewing the history on this code, I believe it
was probably just masked issues that were recently resolved by
AudioShifter.  Testing demonstrated the extra buffering was not
needed.

Other changes: Clarified and refreshed documentative code comments.  Code
clean-ups in media_stream_renderer_factory_impl.cc (e.g., removed output
argument in favor of a return value).

Testing:

1. Connected a chrome.tabCapture.capture() MediaStream to a local VIDEO
element and confirmed reliable playback with good AV sync.

2. Using https://webrtc.github.io/samples/src/content/peerconnection/pc1/
to confirm the WebRTC use case was not broken.

3. Cast Streaming loopback testing (using a simple test extension).

BUG=577881, 577874

Review URL: https://codereview.chromium.org/1633423002

Cr-Commit-Position: refs/heads/master@{#374766}

Review URL: https://codereview.chromium.org/1685233002

Cr-Commit-Position: refs/heads/master@{#374765}

chrome::GetTotalBrowserCount() and ChromeWebUsbPermissionBubble::Bind are not available on Android, and their invocation cause a link error on Android GN build.
The caller is also not used on Android, and seems to be GCed on GYP build.

This CL ifdefs out the problematic code to fix it.

Review URL: https://codereview.chromium.org/1684043002

Cr-Commit-Position: refs/heads/master@{#374764}

In order to make the code changes easier to reason about, this is now part 1
of a two-part commit. In this commit, the existing code is moved from
content/common and content/child into content/renderer (it isn't used anywhere
else). In part 2, the keys are moved out of content into the embedder.

Part 2: https://codereview.chromium.org/1653263005/

BUG=543220

Review URL: https://codereview.chromium.org/1680873002

Cr-Commit-Position: refs/heads/master@{#374763}

Just use v2 instead ;) (compiled_resources2.gyp)

R=maruel@chromium.org,cpu@chromium.org
BUG=585553

Review URL: https://codereview.chromium.org/1686663002

Cr-Commit-Position: refs/heads/master@{#374762}

Remove the Android, Linux, and Windows Perf bot entries.
These were abandoned in: https://codereview.chromium.org/1681033003/
FYI Waterfall: https://build.chromium.org/p/chromium.mojo/console

BUG=585610
TEST=FYI Waterfall no longer lists these unused bots.
R=dpranke@chromium.org

Review URL: https://codereview.chromium.org/1684843003

Cr-Commit-Position: refs/heads/master@{#374761}

size_t is not sufficient when the browser is 32 bit and the renderers are 64 bits.

BUG=581409,585382

Review URL: https://codereview.chromium.org/1685883004

Cr-Commit-Position: refs/heads/master@{#374760}

In the old interface, a static factory method returns a scoped_ptr, and the
caller had to manage the lifetime. Since this is a global object with minimal
memory footprint, and is required to outlive every IPC::Channel, it's much
easier for the global to never be destroyed. This also matches the interface for
AttachmentBrokerPrivileged.

BUG=584297

Review URL: https://codereview.chromium.org/1679763002

Cr-Commit-Position: refs/heads/master@{#374759}

BUG=585268,568130

Review URL: https://codereview.chromium.org/1684953002

Cr-Commit-Position: refs/heads/master@{#374758}

Changes:
https://chromium.googlesource.com/chromiumos/platform/system_api/+log/d9ecb3bd..ac14fd16

ac14fd1 system_api: Add kGetKioskAppRequiredPlatforVersion for update_engine

BUG=577783

Review URL: https://codereview.chromium.org/1684213002

Cr-Commit-Position: refs/heads/master@{#374757}

- create application shortcuts dialog
- permissions bubble view

BUG=585254, 585014

Review URL: https://codereview.chromium.org/1673403002

Cr-Commit-Position: refs/heads/master@{#374756}

Also blacklist a few more bots in manage.py.

This mostly affects the Clang ToT and the clang/win FYI bots:

$  testing/buildbot/manage.py --convert content_unittests
Convert content_unittests to run exclusively on Swarming
4 used to run locally and were converted:
- chromium.fyi.json: ClangToTLinux tester
- chromium.fyi.json: ClangToTMac tester
- chromium.fyi.json: ClangToTMacASan tester
- chromium.fyi.json: CrWinClang64(dbg) tester

$  testing/buildbot/manage.py --convert components_unittests
Convert components_unittests to run exclusively on Swarming
18 used to run locally and were converted:
- chromium.fyi.json: Chromium Win 10
- chromium.fyi.json: ClangToTLinux tester
- chromium.fyi.json: ClangToTLinuxASan tester
- chromium.fyi.json: ClangToTMac tester
- chromium.fyi.json: ClangToTMacASan tester
- chromium.fyi.json: ClangToTWin tester
- chromium.fyi.json: ClangToTWin(dbg) tester
- chromium.fyi.json: ClangToTWin(dll) tester
- chromium.fyi.json: ClangToTWin64 tester
- chromium.fyi.json: ClangToTWin64(dbg) tester
- chromium.fyi.json: ClangToTWin64(dll) tester
- chromium.fyi.json: CrWinClang tester
- chromium.fyi.json: CrWinClang(dbg) tester
- chromium.fyi.json: CrWinClang(shared) tester
- chromium.fyi.json: CrWinClang64 tester
- chromium.fyi.json: CrWinClang64(dbg) tester
- chromium.fyi.json: CrWinClang64(dll) tester
- chromium.fyi.json: CrWinClangLLD tester

$  testing/buildbot/manage.py --convert content_browsertests
Convert content_browsertests to run exclusively on Swarming
2 used to run locally and were converted:
- chromium.fyi.json: ClangToTLinux tester
- chromium.fyi.json: CrWinClang64(dbg) tester

BUG=98637,82385

Review URL: https://codereview.chromium.org/1683253003

Cr-Commit-Position: refs/heads/master@{#374755}

Previously there were cases where MediaCodec#stop would be called while
the MediaCodec was in a broken state, which results in hitting a CHECK.
In AVDA we don't actually need to call stop() because it's always called
right before deleting the MediaCodec. This replaces those calls with
explicit deletion.

BUG=582275
TEST=Trying to play videos with unsupported profiles; no crashes

Review URL: https://codereview.chromium.org/1687143002

Cr-Commit-Position: refs/heads/master@{#374754}

When creating a WebGL canvas, and when allocating a 2D canvas backing,
the implementation was triggering an immediate style calculation
to evaluate the image-redering CSS property.  This CL gets rid of of
these calculations by relying on the fact that if the style is dirty,
it will be updated before the next paint, which will allow for the state
to be corrected after the fact, if necessary.

BUG=585135

Review URL: https://codereview.chromium.org/1674123003

Cr-Commit-Position: refs/heads/master@{#374753}

Missed this one first time around.

BUG=none

Review URL: https://codereview.chromium.org/1684193002

Cr-Commit-Position: refs/heads/master@{#374752}

This CL adds BluetoothRemoteGattServiceWin to BluetoothDeviceWin. The upcoming CL will implement BluetoothRemoteGattServiceWin.

BUG=579202

Review URL: https://codereview.chromium.org/1681853003

Cr-Commit-Position: refs/heads/master@{#374751}

Leftover from the previous cleanups.
Previously, when base did pretend to not know anything about allocator,
tc_set_new_mode was invoked by the content layer (in content_main).
But this required this base unittest to mimic the same content behavior.

These days, instead, tc_set_new_mode() is called by base's
EnableTerminationOnOutOfMemory(). This test is already invoking
EnableTerminationOnOutOfMemory() in its initialization, so there is
no need to perform a 2nd call to tc_set_new_mode.

Also this makes the code in this test conceptually cleaner and smaller.

BUG=564618
TEST=base_unittests

Review URL: https://codereview.chromium.org/1683153002

Cr-Commit-Position: refs/heads/master@{#374750}

We tried dropping this, but there are still uninitialized bytes
somewhere. Zero them again for now.

BUG=585804
TBR=amistry@chromium.org

Review URL: https://codereview.chromium.org/1684733004

Cr-Commit-Position: refs/heads/master@{#374749}

BUG=584828

Review URL: https://codereview.chromium.org/1691533002

Cr-Commit-Position: refs/heads/master@{#374748}

Now GetResponseHeaders won't check the navigation state, instead it will just return a nullptr if there are no response headers available.
In some cases the response headers are not available, but in the previous code there's no safe way to know that.
This removes the CHECK for the navigation state that was causing this problem.

BUG=526291
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_site_isolation

Review URL: https://codereview.chromium.org/1681703002

Cr-Commit-Position: refs/heads/master@{#374747}

BUG=575808

Review URL: https://codereview.chromium.org/1682473002

Cr-Commit-Position: refs/heads/master@{#374746}

If munmap is called during ki_init() in a kernel_wrap_test, it will print a
message saying the call was unexpected. This will cause a deadlock since it is
trying to print a message inside the memory allocator.

BUG=378596
R=sbc@chromium.org
CQ_EXTRA_TRYBOTS=tryserver.chromium.linux:linux_nacl_sdk;tryserver.chromium.mac:mac_nacl_sdk;tryserver.chromium.win:win_nacl_sdk

Review URL: https://codereview.chromium.org/1681223004

Cr-Commit-Position: refs/heads/master@{#374745}

BUG=496258

TBR=dtrainor@chromium.org

Review URL: https://codereview.chromium.org/1676023002

Cr-Commit-Position: refs/heads/master@{#374744}

On some Android devices, the device will restart right
after the browser is launched. This is due to a kernel bug on
these devices when using the Linux GetWifiSSID implementation,
which causes a kernel panic and device restart.

To avoid this, a Java-based implementation is used for Android
devices, which avoids this. Examples of affected devices include:
HTC One S
Lenovo A820
Asus ZenFone 2
MediaPad 10 FHD

R=bengr@chromium.org,rch@chromium.org
BUG=555067
TEST=Take the problem device, install browser and start it, wait for
10 seconds. Make 10 experiments, cleaning data before each run.
Device should not reboot.

Review URL: https://codereview.chromium.org/1633733005

Cr-Commit-Position: refs/heads/master@{#374743}

Notably, this includes browser_tests, which when run locally takes 3h to run
and (since recently) also doesn't pass all tests when not run on swarming.

BUG=498033
R=dpranke@chromium.org

Review URL: https://codereview.chromium.org/1685223002 .

Cr-Commit-Position: refs/heads/master@{#374742}

TBR=simonhatch@chromium.org
BUG=585872

Review URL: https://codereview.chromium.org/1682363004

Cr-Commit-Position: refs/heads/master@{#374741}

Creating a select element dropdown in an out-of-process iframe is
crashing the browser process because the RenderWidgetHostView for the
new widget becomes associated with the top-level frame's renderer
process, creating a routing ID conflict. This CL initializes
the new RenderWidgetHostView with the correct RenderProcessHost.

BUG=582511

Review URL: https://codereview.chromium.org/1685133003

Cr-Commit-Position: refs/heads/master@{#374740}

In crrev.com/14046012 we started to iterate over all RenderViews and
navigating them to about:blank to trigger their unload event handlers
while the main window is still present.  This code is broken when OOPIFs
are enabled (i.e. we cannot navigate a *remote* RenderView / RenderFrame).

This CL basically reverts crrev.com/14046012 and instead puts the burden
on test authors, to make sure that their unload handlers can robustly
handle being executed when the main window is gone.  This makes the
test harness work better in presence of OOPIFs.

BUG=583089

Review URL: https://codereview.chromium.org/1656033004

Cr-Commit-Position: refs/heads/master@{#374739}

Previously, any image animating would trigger low-quality filters for all
images during the animation timer window. Now only do so for images that are
actually animating.

BUG=583238

Review URL: https://codereview.chromium.org/1685933002

Cr-Commit-Position: refs/heads/master@{#374738}

Writing to disk via data()/used() creates a file of the minimum
possible size which exposed an off-by-one bug in a corruption
check.

BUG=

Review URL: https://codereview.chromium.org/1654053002

Cr-Commit-Position: refs/heads/master@{#374737}

This patch will have to be landed manually because line ending
changes are mangled by the CQ. Local presubmit tests pass.

BUG=None
R=creis@chromium.org

Review URL: https://codereview.chromium.org/1680333003 .

Cr-Commit-Position: refs/heads/master@{#374736}

this is a revert of a change I made here: https://codereview.chromium.org/1637203002/diff/60001/chrome/browser/ui/aura/chrome_browser_main_extra_parts_aura.cc

BUG=585522

Review URL: https://codereview.chromium.org/1690443002

Cr-Commit-Position: refs/heads/master@{#374735}

MediaCodecAudioDecoder is an audio decoder that uses Android
MediaCodec API and implements the AudioDecoder interface, i.e.
it can be attached to the DecoderStream and used in the Spitzer
pipeline directly.

This CL only defines its interface and have notes about
possible implementation.

BUG=542910

Review URL: https://codereview.chromium.org/1685483002

Cr-Commit-Position: refs/heads/master@{#374734}

base::Callback will have extra template parameters with default values. But
manual forward declaration causes a compile error on the parameter addition.
This CL unifies the forward decls for its preparation.

BUG=554299

Review URL: https://codereview.chromium.org/1658813003

Cr-Commit-Position: refs/heads/master@{#374733}

BUG=None

Review URL: https://codereview.chromium.org/1688783004

Cr-Commit-Position: refs/heads/master@{#374732}

The message removes alignment of search results to the right of
status bar.

BUG=585853
R=pfeldman, dgozman

Review URL: https://codereview.chromium.org/1689743002

Cr-Commit-Position: refs/heads/master@{#374731}

R=sky@chromium.org
BUG=558054

Review URL: https://codereview.chromium.org/1684843002

Cr-Commit-Position: refs/heads/master@{#374730}

Previously this was enabled through the (lock screen)
RemoteControlClient, but this was removed in
https://codereview.chromium.org/1652163002/. Reenable
it through the notification's MediaSessionCompat.

BUG=585393

Review URL: https://codereview.chromium.org/1679923005

Cr-Commit-Position: refs/heads/master@{#374729}

TEST=reproduction in bug no longer occurs.
BUG=576607

Review URL: https://codereview.chromium.org/1668973002

Cr-Commit-Position: refs/heads/master@{#374728}