GitLab

We're going to change to use more than one background windows to enclose panels in the stack because not all panels are fully aligned when some of them are being dragged. The 'Window' suffix is added to the naming to make it clear.

BUG=223155
TEST=none due to no functionality change
TBR=thakis@chromium.org

Review URL: https://codereview.chromium.org/14188014

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194666 0039d316-1c4b-4281-b951-d872f2087c98

A recent change to accessibility on Chrome OS is triggering a call to
MenuItemView::GetAccessibleState on a MenuItemView that has a delegate
but no command id, which asserts in MenuModelAdapter when we try to
look up a command that's not there.

The fix is to not call MenuModelAdapter::GetAccelerator to look up a
command when the command is zero.

BUG=230107,232079

Review URL: https://codereview.chromium.org/14161006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194665 0039d316-1c4b-4281-b951-d872f2087c98

This function was used in only one place and that place was wrong.

The implementation was long and complicated on Windows, and long and
complicated and copied verbatim from elsewhere with 3 lines different on Posix.

The place it was used was in the proxy launcher when copying the profile. It's
not clear to me why we wouldn't want the profile files in the filesystem cache
when running tests. Quite the opposite, we want the tests to run as fast as
possible.

The only place this should matter is in the startup tests. And the startup
tests do things to the profile after it gets copied that should page some files
back in! This adds another step to the startup tests to evict the profile files
for cold startup tests only.

This is a reland of r192940 which crashed on Mac. The previous perf startup
test used the profile directory before it was initialized. That patch was in
turn a reland r191854 which broke the startup test. The previous patch
replaced the CopyRecursive call with CopyDirectory which does something
slightly different. This new patch implements a CopyDirectoryContents function
to do what's required here.

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194664 0039d316-1c4b-4281-b951-d872f2087c98

On some platforms, listening on "localhost" binds to the IPv6 loopback
(::1), even if there's no global IPv6 connectivity. In such situations,
HostResolverImpl restricts lookups to IPv4 (AF_INET) for performance
reasons, but then navigating to http://localhost results in "connection
refused".

HostResolverProc has a workaround for this situation. This CL adds a
test for this behavior and replicates the workaround in
HostResolverImpl::ServeFromHosts (for the built-in asynchronous DNS).

BUG=224215

Review URL: https://codereview.chromium.org/14208010

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194663 0039d316-1c4b-4281-b951-d872f2087c98

BUG=none
TEST=none

Review URL: https://codereview.chromium.org/14178004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194662 0039d316-1c4b-4281-b951-d872f2087c98

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194661 0039d316-1c4b-4281-b951-d872f2087c98

http://build.chromium.org/f/chromium/perf/dashboard/ui/changelog_blink.html?url=%2Ftrunk&range=148536%3A148558&mode=html

TBR=tkent@chromium.org

Review URL: https://codereview.chromium.org/13870008

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194660 0039d316-1c4b-4281-b951-d872f2087c98

Fix race condition if a pulsing bookmark button is deleted while the bookmark bubble controller is showing.

BUG=180102
TEST=ActionBoxTest.BookmarkAPageTest reliably works

Review URL: https://codereview.chromium.org/14130008

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194658 0039d316-1c4b-4281-b951-d872f2087c98

Two small tweaks to the audio indicator:
  - change the equalizer levels to look better with fewer bars
  - make animation slightly faster 300 ms per frame to 250 ms per frame.
This doesn't noticeably increase CPU usage on my machine.

Also fixed a small bug where the indicator would be drawn incorrectly
when an equalizer level was 0.

BUG=3541
TEST=Played a youtube video. Verified that the audio indicator in the
     favicon was updated.

Review URL: https://codereview.chromium.org/14139022

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194657 0039d316-1c4b-4281-b951-d872f2087c98

BUG=222681
For BrowserProcessImpl change:
TBR=sky@chromium.org

Review URL: https://codereview.chromium.org/14200028

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194656 0039d316-1c4b-4281-b951-d872f2087c98

Add a way to add ConsumeCallbacks to a SpdyBuffer in order to be notified
when Consume() is called. Use that to ensure that flow control receive
windows are updated appropriately regardless of what the SpdyStream's
delegate does.

Make IncreaseRevWindowSize private in both SpdyStream
and SpdySession.

BUG=176592

Review URL: https://codereview.chromium.org/14311002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194655 0039d316-1c4b-4281-b951-d872f2087c98

> Remove scoped_array usage from RegistryKeyBackup.
> 
> BUG=171111
> 
> Review URL: https://codereview.chromium.org/14265020

TBR=dcheng@chromium.org

Review URL: https://codereview.chromium.org/14330007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194654 0039d316-1c4b-4281-b951-d872f2087c98

> Remove scoped_array from Chromium.
> 
> C++11 provides unique_ptr<T[]>, and Chromium has implemented
> scoped_ptr<T[]> to match its behavior during the transition period. As a
> result, scoped_array<T> is now redundant and is being removed.
> 
> BUG=171111
> 
> Review URL: https://codereview.chromium.org/14081006

TBR=dcheng@chromium.org

Review URL: https://codereview.chromium.org/14225009

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194653 0039d316-1c4b-4281-b951-d872f2087c98

Style-only change.

R=danakj
BUG=

Review URL: https://codereview.chromium.org/14205009

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194650 0039d316-1c4b-4281-b951-d872f2087c98

C++11 provides unique_ptr<T[]>, and Chromium has implemented
scoped_ptr<T[]> to match its behavior during the transition period. As a
result, scoped_array<T> is now redundant and is being removed.

BUG=171111

Review URL: https://codereview.chromium.org/14081006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194649 0039d316-1c4b-4281-b951-d872f2087c98

This CL is the 4th in a series of changes that straddle webkit vs chromium repositories.
1) WK: Declare new virtual createSnapshotFile/didCreateSnapshotFile public api methods.
2) CR: Implement the new create method such that the didCreate method is invoked in response.
3) WK: Use the new create method and implement the new didCreate method.
4) CR: Cleanup the obsolete/deprecated blocks of code.

BUG=174200

Review URL: https://codereview.chromium.org/13891009

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194648 0039d316-1c4b-4281-b951-d872f2087c98

BUG=170229
TEST=by hand as per repro on bug

Review URL: https://codereview.chromium.org/13972011

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194647 0039d316-1c4b-4281-b951-d872f2087c98

This fixes a bug when reading from a HTTP mount, if the Content-Length header
is not available.

It occurs because the code was caching a pointer to a std::vector's buffer,
after it had resized it.

BUG=none
R=noelallen@chromium.org

Review URL: https://codereview.chromium.org/13831005

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194646 0039d316-1c4b-4281-b951-d872f2087c98

BUG=

Review URL: https://codereview.chromium.org/14017004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194645 0039d316-1c4b-4281-b951-d872f2087c98

Hopefully this is the last one before the scoped_array removal can land.

BUG=171111

Review URL: https://codereview.chromium.org/14161010

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194644 0039d316-1c4b-4281-b951-d872f2087c98

BUG=171111

Review URL: https://codereview.chromium.org/14265020

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194643 0039d316-1c4b-4281-b951-d872f2087c98

AppCacheExecutableHandlers - parse manifest file and store/retrieve the 'bit' indicating executable.

BUG=101800

Review URL: https://codereview.chromium.org/13881003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194642 0039d316-1c4b-4281-b951-d872f2087c98

Local environment:
$ cat ~/.gyp/include.gypi
{
  'variables': {
    'debug_devtools': 1,
  },
}

The build fails with the following error:

../../chrome/browser/ui/webui/devtools_ui.cc: In constructor 'DevToolsUI::DevToolsUI(content::WebUI*)':
../../chrome/browser/ui/webui/devtools_ui.cc:239:59: error: cannot allocate an object of abstract type '{anonymous}::LocalhostDataSource'
../../chrome/browser/ui/webui/devtools_ui.cc:172:7: note:   because the following virtual functions are pure within '{anonymous}::LocalhostDataSource':
../../content/public/browser/url_data_source.h:40:23: note: 	virtual std::string content::URLDataSource::GetSource() const

Review URL: https://codereview.chromium.org/14328007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194641 0039d316-1c4b-4281-b951-d872f2087c98

BUG=none
TEST=bisect something where the regression range includes a blink roll.
Outputs working link for blink regression range.

Review URL: https://codereview.chromium.org/13851009

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194640 0039d316-1c4b-4281-b951-d872f2087c98

If a layer has a replica, which has a mask, then the layer itself does not
need to clip to its bounding box. This dcheck verified that it would.

Add some new unit tests and pixel tests that verify behaviour of the mask
texture with masked replicas. The current behaviour is correct. The
replica is masked at the origin of the replicated layer.

Tests:
LayerTreeHostImplTest.ReflectionMaskLayerWithDifferentBounds
LayerTreeHostImplTest.ReflectionMaskLayerForSurfaceWithUnclippedChild
LayerTreeHostImplTest.MaskLayerForSurfaceWithClippedLayer
LayerTreeHostMasksPixelTest.MaskOfLayer
LayerTreeHostMasksPixelTest.ImageMaskOfLayer
LayerTreeHostMasksPixelTest.MaskOfClippedLayer
LayerTreeHostMasksPixelTest.MaskWithReplica
LayerTreeHostMasksPixelTest.MaskWithReplicaOfClippedLayer
LayerTreeHostMasksPixelTest.MaskOfReplica
LayerTreeHostMasksPixelTest.MaskOfReplicaOfClippedLayer

R=enne
BUG=171734

Review URL: https://codereview.chromium.org/14244021

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194639 0039d316-1c4b-4281-b951-d872f2087c98

> Add a virtual keyboard webui at chrome://keyboard/
> 
> This is just the old virtual keyboard that was removed a couple of years
> ago, being used as a temporary stand-in.
> 
> BUG=222801
> 
> Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194579
> 
> Review URL: https://codereview.chromium.org/13652010

TBR=bryeung@chromium.org

Review URL: https://codereview.chromium.org/14329004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194638 0039d316-1c4b-4281-b951-d872f2087c98

This is just the old virtual keyboard that was removed a couple of years
ago, being used as a temporary stand-in.

BUG=222801

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194579

Review URL: https://codereview.chromium.org/13652010

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194635 0039d316-1c4b-4281-b951-d872f2087c98

QUIC crypto: move config objects.

Currently the client and server configs are setup and torn-down for each
connection. Since they are supposed to be per-client and per-server objects,
this change makes them parameters that are passed into the connection

Merge internal change: 44269387

QUIC crypto steps 6 and 7: per-server strike register.

This change adds a per-server strike-register that allows the server to
complete 0-RTT connections if the client has enough information cached.

Due to the fact that the per-server and per-client objects
(QuicCryptoServerConfig and QuicCryptoClientConfig) are currently setup and
torn down for each connection, there's no tests in this change for a 0-RTT
handshake because we can't do one yet. The next change will move these objects
into the right place so that 0-RTT handshakes can be tested.

This change also reminded me why I had a server nonce: without it the server
cannot terminate any connections if the strike-register fails. So the server
nonce is firmly back.

Merge internal change: 44228897

R=rch@chromium.org

Review URL: https://codereview.chromium.org/13976007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194634 0039d316-1c4b-4281-b951-d872f2087c98

BUG=231725
TEST=none

Review URL: https://codereview.chromium.org/14063011

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194633 0039d316-1c4b-4281-b951-d872f2087c98

windows.h defines BEGIN_INTERFACE -- we need to undefine it before we redefine
it.

BUG=none
R=sbc@chromium.org

Review URL: https://codereview.chromium.org/13998024

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194632 0039d316-1c4b-4281-b951-d872f2087c98

It is flaky on linux too. The failure tends to reproduce more often on linux
swarm slaves than linux_rel but this is really race-condition like.

TBR=jeremya@chromium.org
BUG=223467

Review URL: https://codereview.chromium.org/14329003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194631 0039d316-1c4b-4281-b951-d872f2087c98

My http://crrev.com/192493 caused a regression:
FindBarView doesn't select-all on tab-switch focus restoration.

Restore the FindBarView::SearchTextfieldView class with RestoreFocus override.
( this restores the previous select-all behavior to fix the regression )

BUG=232290
TEST=Switching back to a tab that had the find bar focused (from a tab that didn't have the find bar focused) restores find bar focus and selects the entire text content.
R=finnur@chromium.org
TBR=ben@chromium.org

Review URL: https://codereview.chromium.org/13976011

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194630 0039d316-1c4b-4281-b951-d872f2087c98

This provides a replacement for the inner class of SerializedVar for serializing a PP_Var. It achieves 2 purposes: 1) it refactors the design of SerializedVar::Inner which was growing increasingly messy (and was going to become nastier) 2) it provides support for serializing/deserializing dictionary and array PP_Vars. To serialize a PP_Var, first the transitive closure of that var is computed (all nodes referenced) and then each var in the transitive closure is written to the message. Some trickiness arises from the fact that a PP_Var can be either a primitive or a reference and references have to be maintained when transmitting over the wire. The comments in the header files provide a description of the approach in more detail.

BUG=

Review URL: https://codereview.chromium.org/13887007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194629 0039d316-1c4b-4281-b951-d872f2087c98

Previously P2PSocketHostTcp was calling Write() even when another write is
pending.

BUG=232046

Review URL: https://codereview.chromium.org/13926013

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194628 0039d316-1c4b-4281-b951-d872f2087c98

Serializing calls to getAuthToken will eliminate unnecessary calls to
GAIA and consolidate UI flows.

There is are two queues of requests for each (extension, scopes)
tuple, one for the non-interactive IssueToken flow and one for the
interactive prompt.

BUG=228908
(step #1 of the bug description)

Review URL: https://codereview.chromium.org/14270007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194627 0039d316-1c4b-4281-b951-d872f2087c98

R=enne
BUG=171734

Review URL: https://codereview.chromium.org/14299010

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194626 0039d316-1c4b-4281-b951-d872f2087c98

This change puts the final signatures containing doubles in place so that
they can be used by Blink code. After this change lands Blink will be updated
to call these methods instead of the temporary xxxFloat() methods.

BUG=227156
TEST=All existing tests still pass. No user visible behavior has changed.

Review URL: https://codereview.chromium.org/13866046

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194624 0039d316-1c4b-4281-b951-d872f2087c98

logged into (or has been logged into).
re-submission of https://codereview.chromium.org/13903018

Review URL: https://codereview.chromium.org/14065014

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194622 0039d316-1c4b-4281-b951-d872f2087c98

This short fix in the deprecated bluetooth ChromeOS implementation
shows a trusted device (but not paired) as paired. This allows testing
the Verbatim mouse in the current implementation.

BUG=232145
TEST=Connect to a Verbatim device (connected but not paired). Device shows up in the settings list.

Review URL: https://codereview.chromium.org/14316006

Patch from Alex Deymo <deymo@chromium.org>.

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194621 0039d316-1c4b-4281-b951-d872f2087c98

With fixed include path.

TBR=cpu@chromium.org
BUG=225822

Review URL: https://codereview.chromium.org/14096011

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@194620 0039d316-1c4b-4281-b951-d872f2087c98