GitLab

These things landed in the wrong place during the sync componentization.
ProtocolEvent implementations are moved to engine_impl, and the
ProtocolEventObserver is moved from driver to engine.

engine/events/
  protocol_event_observer

engine_impl/events/
  clear_server_data_request_event
  clear_server_data_response_event
  commit_request_event
  commit_response_event
  configure_get_updates_request_event
  get_updates_response_event
  normal_get_updates_request_event
  poll_get_updates_request_event

BUG=631271
TBR=bauerb

Review-Url: https://chromiumcodereview.appspot.com/2436863005
Cr-Commit-Position: refs/heads/master@{#426866}

BUG=504347

Review-Url: https://chromiumcodereview.appspot.com/2442783002
Cr-Commit-Position: refs/heads/master@{#426865}

Skia is interested in replacing all uses of SkAutoTUnref with sk_sp and
then hiding SkAutoTUnref with the intention of removing it in the
future. This appears to be the only use of this class in Chromium
outside of Skia itself.

Review-Url: https://chromiumcodereview.appspot.com/2438643005
Cr-Commit-Position: refs/heads/master@{#426864}

When content is overlayed(eg. reader mode), enabling download is quite troublesome.
The main issue is that download may cause infobar to show up.
As a result, this CL disables download for overlay content.
Currently it only affects the download button on media controls.
Will disable other elements in later CLs if I found that they can
trigger download.
Because the media control download button uses an internal css, cannot
use distilledpage.css to hide it.

BUG=650738

Review-Url: https://chromiumcodereview.appspot.com/2394653002
Cr-Commit-Position: refs/heads/master@{#426863}

This is a follow up CL, which merges RequestImpl and Job
together.

Prev CL is here - https://chromiumcodereview.appspot.com/2299963002/

BUG=478934

Review-Url: https://chromiumcodereview.appspot.com/2430953004
Cr-Commit-Position: refs/heads/master@{#426862}

Most if the changes are due to sort order of the rules, which has now been more fully specified. The rules used to be in order of decreasing length, with unspecified order for rules of the same length. Rules are now sorted on order of decreasing length and then in lexicographical order.

BUG=620414, 658303

Review-Url: https://chromiumcodereview.appspot.com/2434003003
Cr-Commit-Position: refs/heads/master@{#426861}

U_LB_COUNT is assumed to be 40 in Blink line breaking code, but it's
43 in ICU 58/Unicode 9.

Three new classes (Emoji Base, Emoji Modifier, and ZWJ) should behave
identically whether or not 'word-break: break-all' is in effect.

BUG=637001
TEST=TextBreakIterator.cpp is compiled without an error with ICU 58.
R=kojii@chromium.org
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel

Review-Url: https://chromiumcodereview.appspot.com/2440923002
Cr-Commit-Position: refs/heads/master@{#426860}

The Physical Web scanner was initially designed to run for a relatively
short duration (less than a minute) and would report any URLs seen during
this time as nearby. With this change, the scanner will keep track of how
recently a URL was seen so that URLs that are no longer nearby can be
removed from the list.

BUG=630769

Review-Url: https://chromiumcodereview.appspot.com/2413923002
Cr-Commit-Position: refs/heads/master@{#426859}

Note on matching color spaces:
*** Roughly 90% of image profiles specify sRGB.  Most
    dsts are also sRGB (and so is our default when we
    don't know the dst).  In the sRGB->sRGB case (and in
    other cases where the src and dst match), we should
    skip the xform.  QCMS checks for matching profiles,
    but the check is very weak (only works if profiles have
    the same text description).  SkColorSpace has
    a robust check for color space equality.

Performance Notes:
***Test Srcs: Averaged across 97 images pulled from the top
   10k skps.
***Test Dsts: sRGB transfer function, 2.2, other.

HP z620 Average Xform Speed-Up (QCMS / SkColorSpaceXform)
sRGB  1.51x
2.2   1.61x
Other 1.44x
(SkColorSpaceXform is actually fastest for sRGB - I just
had to drop a bunch of images from the sRGB average,
including a few that tripped up QCMS.  SkColorSpaceXform
was recognizing sRGB->sRGB was a no-op, so including these
images skewed the results in its favor - I saw like 17x)

Nexus 6P Average Xform Speed-Up (QCMS / SkColorSpaceXform)
sRGB  2.08x
2.2   1.42x
Other 3.33x
(interesting that the fallback code is actually faster than
the special cases on Arm - likely this is because the
sqrt instructions that we're using to model the curves are
slow on this platform - would be interesting to investigate
further to see if we can do better)
*** Note that turning this on for Android is left to a
    later CL.

Correctness Notes:
***QCMS (and the Skia fallback code path) use a LUT to
   convert linear floats to the dst transfer function.
   This results in a loss of accuracy since the float
   must be rounded to an int before being used as an
   index into the table.  Both Skia and QCMS do not
   interpolate the table for performance reasons.
   Skia avoids table-based implementations for two
   common cases (sRGB, 2.2), guaranteeing a result
   within 1 (on a 0-255 scale) of the "true" value.
   QCMS may be off by as much as 5 on the steep 2.2
   curve.
***There are several fairly common sRGB profiles that
   represent the transfer function as a small LUT.
   QCMS does not recognize that it is sRGB, and
   interpolates this LUT to build a larger LUT that
   it then uses.  This results in the conversion being
   off from the "true" value by as much as 9 (on a
   0-255 scale).

Other Behavior Changes:
*** More lenient approval of ICC profiles.  QCMS rejects
    xform matrices that are not "close enough to D50" by
    some arbitrary measure.  We don't have enough data to
    know that this is or isn't necessary.  QCMS also
    rejects profiles that it thinks are "too big".  Again,
    this seems reasonable, but we'll wait until we
    understand the motivation better.

What ICC profiles are supported?
***SkColorSpaceXform supports the same subset of profiles
   as QCMS.  Adding support for A2B and Lab profiles is
   a WIP (We had support for a few of these, but we've
   dropped it as part of a refactor.  The plan is to add
   it back as a part of a more logical, complete design).

How does this relate to color correct rendering for Chrome?
***This modifies the legacy path, not the color correct
   path.
***It's still somewhat interestingly related to color
   correct rendering, since the color correct renderer
   will also use SkColorSpaceXform in some cases
   (likely just to xform scary profiles into linear F16).
***It looks like the color correct rendering path will
   rely on tagged images being handled inside Skia, but
   it's worth mentioning that this change makes it simpler
   to do flexible xforms (ex: to F16) at decode time.

NO_DEPENDENCY_CHECKS=true

BUG=

Committed: https://crrev.com/8602b7fdc54d18aff914b5ddf0bba2a406078797
Review-Url: https://chromiumcodereview.appspot.com/2426723005
Cr-Original-Commit-Position: refs/heads/master@{#426588}
Cr-Commit-Position: refs/heads/master@{#426858}

on data heavy pages via the Chrome-Proxy-Accept-Transform
header.

BUG=657658

Review-Url: https://chromiumcodereview.appspot.com/2424263002
Cr-Commit-Position: refs/heads/master@{#426857}

BUG=658053

Review-Url: https://chromiumcodereview.appspot.com/2437873006
Cr-Commit-Position: refs/heads/master@{#426856}

These checks were originally introduced in r419019 and r422954.  This
was merged back to M54 and M55.

They were later moved to the UI thread (into
ExtensionNavigationThrottle) in r424937 to make them compatible with
PlzNavigate.  However, that only exists in M56.

Per linked bug, we need to tighten the security check currently in M54
and M55 for apps with a "webview" permission where the problematic
blob/filesystem URL requests aren't made from the guest process.

This is the plan for fixing this:
1.  <This CL> Reintroduce these IO thread checks on M56.
2.  https://codereview.chromium.org/2437753003: Tighten the checks for
    the scenario above using ChildProcessSecurityPolicy.
3.  Merge the CL from step 2 back to M55 and M54.
4.  Remove these checks from M56 and tighten the checks in
    ExtensionNavigationThrottle.

BUG=656752

Review-Url: https://chromiumcodereview.appspot.com/2435593007
Cr-Commit-Position: refs/heads/master@{#426855}

This allows me to skip refactoring it for mustash. It also avoids one place
where chrome calls back into ash that I would have to refactor as well.

The method just looks up a chromeos command-line switch, so inline that
lookup where needed.

In addition, clean up some of the splash screen init logic to make it more
clear how we handle primary vs. secondary displays.

BUG=none
TEST=run chrome with and without --first-exec-after-boot, watch background
colors as displays are initialized

Review-Url: https://chromiumcodereview.appspot.com/2437393002
Cr-Commit-Position: refs/heads/master@{#426854}

This is a re-land of:
https://codereview.chromium.org/2435813002/

The reland keeps a TimesHistogramSample class in LaunchMetrics
since it's still referenced in internal code. Once this lands, the
clank/ code can be updated and this class can be cleaned up.

Original CL description:

This is general-purpose functionality that allows
recording UMA metrics before the C++ native library
has been loaded and is useful to have outside of the
Chrome layer. In particular, I will be using it from
components/variations to record information about the
first run variations seed fetch.

No functional changes in this CL.

BUG=632199
TBR=bauerb@chromium.org, thestig@chromium.org

Review-Url: https://chromiumcodereview.appspot.com/2439113002
Cr-Commit-Position: refs/heads/master@{#426853}

This fixes a bug where IsIsolateExtensionsEnabled() had been called prior to the
field trials actually being initialized, so the call returned true regardless
of the finch experiment state, resulting in a possible misconfiguration of
security state.

PreCreateThreads() seems to be the earliest place in BrowserProcessImpl where we
can do this registration.

BUG=657629

Review-Url: https://chromiumcodereview.appspot.com/2438823003
Cr-Commit-Position: refs/heads/master@{#426852}

This CL takes a few steps along the way to unify compositor code
between Mus+Ash and the browser.

1. BitmapUploader uses ui::CompositorFrameSink.
2. MusBrowserCompositorOutputSurface uses
ui::CompositorFrameSink.
3. Introduces ui::Window::RequestCompositorFrameSink that replaces
RequestSurface
4. WindowSurface/ServerWindowSurface use
MojoCompositorFrameSink interface and surface.mojom is deleted.

TODO: Merge WindowSurface and WindowCompositorFrameSink.

BUG=656975
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel

Review-Url: https://chromiumcodereview.appspot.com/2430653002
Cr-Commit-Position: refs/heads/master@{#426851}

Merge internal change: 136848647

BUG=555865

Review-Url: https://chromiumcodereview.appspot.com/2417183003
Cr-Commit-Position: refs/heads/master@{#426850}

Important changes included:
AVX2 and NEON halffloat conversion.
Add F16C cpu detection.
Support for I411 removed.
Side by side UV for I420 output improves memory coherency.

HalfFloat AVX2 ported from SSE2 using same magic number method, which
is 20% faster than vcvtps2ph method and produces identical results.

HalfFloat Neon version adapted from inner loop of vectorized C, but folds shift and narrow into
one instruction and uses element multiply instead of vector to save a
register and dup instruction.  Neon version is also full performance with -Os.

This CL enables -O2 for libyuv_neon as well.

ExtractAlpha ported to AVX2.
ARGB4444ToI420 ported to MSA.
F16C cpu detection for AVX hardware that has halffloat conversion support.

Change log:
https://chromium.googlesource.com/libyuv/libyuv/+log/198bce39..550cf829
Full changes
https://chromium.googlesource.com/libyuv/libyuv/+/198bce39..550cf829

TEST=TestHalfFloatPlane_denormal
BUG=libyuv:560, libyuv:650, libyuv:572, libyuv:645, libyuv:649
R=hubbe@chromium.org
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_precise_blink_rel

Review-Url: https://chromiumcodereview.appspot.com/2425423006
Cr-Commit-Position: refs/heads/master@{#426849}

I'm going to need this for the aura-mus work so that what is the
WindowTreeClient implementation can track capture and update the
server appropriately.

I didn't update the desktop client as I don't think I'll need that one.

BUG=none
TEST=none
R=sadrul@chromium.org

Review-Url: https://chromiumcodereview.appspot.com/2436023003
Cr-Commit-Position: refs/heads/master@{#426848}

If chrome is shut down, then a user deletes a request from a
notification, the scheduler callback will not be set, and we could
crash trying to inform it.  To fix this, when we start, set an
empty scheduler callback to receive the call, so we don't try
to call through a nullptr.

BUG=64911

Review-Url: https://chromiumcodereview.appspot.com/2429213009
Cr-Commit-Position: refs/heads/master@{#426847}

BUG=658318
TBR=tkent@chromium.org
NOTRY=true
NOTREECHECKS=true

Review-Url: https://chromiumcodereview.appspot.com/2438323002
Cr-Commit-Position: refs/heads/master@{#426846}

There is a conflict in two recent CLs:
https://chromiumcodereview.appspot.com/2437963002
https://chromiumcodereview.appspot.com/2428353005,
in which WebRTC does not take audio focus both for the input stream and
the output stream. This CL lets WebRTC take audio focus on Android from
MediaSession again to fix the issue.

BUG=596516

Review-Url: https://chromiumcodereview.appspot.com/2437343002
Cr-Commit-Position: refs/heads/master@{#426845}

BUG=chromedriver:1503,chromedriver:1478

Review-Url: https://chromiumcodereview.appspot.com/2431863002
Cr-Commit-Position: refs/heads/master@{#426844}

Make PaymentOption store a Drawable instead of id for the icon.
This will allow to use icons not contained in the default resources.

Original patch http://crrev.com/2436883002 from rob.buis@samsung.com.

BUG=620173

Review-Url: https://chromiumcodereview.appspot.com/2436363002
Cr-Commit-Position: refs/heads/master@{#426843}

BUG=658311
TBR=majidvp@chromium.org
NOTRY=true
NOTREECHECKS=true

Review-Url: https://chromiumcodereview.appspot.com/2441063002
Cr-Commit-Position: refs/heads/master@{#426842}

This is an automated CL created by the recipe roller. This CL rolls recipe
changes from upstream projects (e.g. depot_tools) into downstream projects
(e.g. tools/build).

Please review the expectation changes, and LGTM as normal. The recipe roller
will *NOT* CQ the change itself, so you must CQ the change manually.

More info is at https://goo.gl/zkKdpD. Use https://goo.gl/noib3a to file a bug
(or complain)

build:
  https://crrev.com/75023e2d624be7d2535efdb683e05f1697713a1e Change archive_retry_summary step to happen also when there are new failures. (qyearsley@chromium.org)

R=qyearsley@chromium.org

Recipe-Tryjob-Bypass-Reason: Autoroller
Bugdroid-Send-Email: False
Review-Url: https://chromiumcodereview.appspot.com/2435803007
Cr-Commit-Position: refs/heads/master@{#426841}

Previously we would just crash due to CHECK when we detected media
streams with duplicate bytestream ids. This CL adds explicit checks
for duplicate stream ids in FFmpegDemuxer and skips those streams.

BUG=657437

Review-Url: https://chromiumcodereview.appspot.com/2434383003
Cr-Commit-Position: refs/heads/master@{#426840}

BUG=640947
TBR=falken@chromium.org
NOTRY=true
NOTREECHECKS=true

Review-Url: https://chromiumcodereview.appspot.com/2442043002
Cr-Commit-Position: refs/heads/master@{#426839}

Currently on Blimp IME, hitting Enter only sets the text in the web
page. So user has to hit Enter twice in order to see search results.
This patch solves this by generating a synthetic Enter key press event
and sending it to the renderer.

BUG=611132

Review-Url: https://chromiumcodereview.appspot.com/2393043004
Cr-Commit-Position: refs/heads/master@{#426838}

In the preliminary CL http://crrev.com/2406153003, I changed Rietveld.latest_try_jobs to Rietveld.latest_try_job_results and made it return a dict of Builds (named tuple objects with builder_name and build_number) to result info dicts.

The purpose of this was to make it so that rebaseline-cl could look through the result info dicts to see whether each build is completed or just started, to avoid starting extra builds.

After looking at it some more, I see that in the Rietveld response, builds that are just started but not completed have no build number information.

To avoid having to pass around the result dict information from Rietveld, it's possible to just have a list of Build objects with build_number set to None when the build is just pending or started but not completed.

So, in summary, this CL makes it so that when one runs rebaseline-cl twice in a row, it now prints out:

$ ./webkit-patch rebaseline-cl
 Triggering try jobs for:
   android_blink_rel
   linux_precise_blink_rel
   linux_trusty_blink_rel
   mac10.10_blink_rel
   mac10.11_blink_rel
   mac10.11_retina_blink_rel
   mac10.9_blink_rel
   win10_blink_rel
   win7_blink_rel
 Please re-run webkit-patch rebaseline-cl once all pending try jobs have finished.
 $ ./webkit-patch rebaseline-cl
 There are existing pending builds for:
   android_blink_rel
   linux_precise_blink_rel
   linux_trusty_blink_rel
   mac10.10_blink_rel
   mac10.11_blink_rel
   mac10.11_retina_blink_rel
   mac10.9_blink_rel
   win10_blink_rel
   win7_blink_rel
 Please re-run webkit-patch rebaseline-cl once all pending try jobs have finished.

And this CL undoes some of the changes made in http://crrev.com/2406153003 in order to simplify the code, because now I think passing full result info dicts is unnecessary.

BUG=654919

Review-Url: https://chromiumcodereview.appspot.com/2439693003
Cr-Commit-Position: refs/heads/master@{#426837}

This makes WebRTCConfiguration into a data struct instead of wrapping a
RTCConfiguration. The motivation for this is to introduce
RTCConfiguration as an IDL dictionary, at which point the extra layer
starts to look unnecessary.

BUG=649343

Review-Url: https://chromiumcodereview.appspot.com/2415673003
Cr-Commit-Position: refs/heads/master@{#426836}

This change makes sure that we gracefully handle the case where a
pending pointer lock request is still active while destroying the
RenderWidgetHost.

BUG=655592

Review-Url: https://chromiumcodereview.appspot.com/2437213002
Cr-Commit-Position: refs/heads/master@{#426835}

BUG=658305
TBR=senorblanco@chromium.org
NOTRY=true
NOTREECHECKS=true

Review-Url: https://chromiumcodereview.appspot.com/2439143002
Cr-Commit-Position: refs/heads/master@{#426834}

This cl makes div output the same way as classic:
- when a div receives either ChromeVox or page focus, read its text content or its aria label
- this cleans up output and retains the previously hardened behavior in Classic.

TEST=browse on a variety of sites like Gmail, Groups. Verify that despite bad markup, ChromeVox performs well and provides a smooth experience.
BUG=619279
TBR=dtseng@chromium.org
NOTRY=true
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:closure_compilation

Committed: https://crrev.com/0026818d58dc22ebea46fe8ff31456aa0ac7bf65
Cr-Commit-Position: refs/heads/master@{#426497}

patch from issue 2433823003 at patchset 20001 (http://crrev.com/2433823003#ps20001)

Review-Url: https://chromiumcodereview.appspot.com/2441033002
Cr-Commit-Position: refs/heads/master@{#426833}

BUG=555865

Review-Url: https://chromiumcodereview.appspot.com/2435863004
Cr-Commit-Position: refs/heads/master@{#426832}

THROTTLED is used when the consumer is expecting that low priority
requests will soon be followed by higher priority ones.  It directs the
network stack to reserve resources to handle the coming requests.  In
practice this will mean that only a very small number of THROTTLED requests
may be outstanding at any given time.

BUG=600839

Review-Url: https://chromiumcodereview.appspot.com/1866483002
Cr-Commit-Position: refs/heads/master@{#426831}

Big CL that fixes many small things related to display information being
propagated when a display is modified. The primary change is that
ws::DisplayManager now handles OnDisplayModified() calls. This entails
the following steps:

1. The PlatformDisplay corresponding to the display is updated.
2. Mus sends an IPC to WM clients telling them the display has changed.
   This requires a new message in mojom::WindowManager and new method in
   WindowManagerDelegate.
3. Update the root ServerWindow for the display.
4. Mus sends an IPC to any other clients telling them the display has
   changed.

The ViewportMetrics class is moved from src/services/ui/ws/ to
src/services/ui/display so that it can be used in
PlatformScreenDelegate.

PlatformScreen now implements ui::DisplayController::StateController so
it can pick the display size.  A message is added to
mojom::TestDisplayController to trigger changing the display size. This
allows developers to try the functionality. An unused ash developer
keyboard shortcut is used to trigger changing the display size. This
shortcut is only temporary until settings are working.

Another small part of this change is the display insets are sent from
ash to PlatformScreen. This is tangentially related as it relies on
DisplayManager::OnDisplayModified() working.

BUG=641012

Review-Url: https://chromiumcodereview.appspot.com/2434923002
Cr-Commit-Position: refs/heads/master@{#426830}

This element is no longer used, and was included in this file by
mistake. Removing it reduces the size of the element bundle by about
7KB.

BUG=425625
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:closure_compilation

Review-Url: https://chromiumcodereview.appspot.com/2436323002
Cr-Commit-Position: refs/heads/master@{#426829}

the "show all" button.

BUG=657935

Review-Url: https://chromiumcodereview.appspot.com/2439873002
Cr-Commit-Position: refs/heads/master@{#426828}

This CL adds support on the Blimp to propagate loading status and page
load status to Tab. This enables the UI to display the loading spinner
on the tablets and page load progress bar on phones.

The page load progress was set to 50 percent for now.

BUG=651473

Review-Url: https://chromiumcodereview.appspot.com/2394663002
Cr-Commit-Position: refs/heads/master@{#426827}