GitLab

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271016 0039d316-1c4b-4281-b951-d872f2087c98

The DCHECK dereferenced the request's HttpResponseInfo.  It turns out
that when an HttpNetworkTransaction is cancelled and an
HttpResponseDrainer is used, the HttpResponseInfo is deleted with the
HttpStreamParser's knowledge.

This CL also adds a pair of tests for the situation.

BUG=368418

Review URL: https://codereview.chromium.org/284913003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271014 0039d316-1c4b-4281-b951-d872f2087c98

BUG=373748
TBR=mukai@chromium.org

Review URL: https://codereview.chromium.org/284203009

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271013 0039d316-1c4b-4281-b951-d872f2087c98

This CL:
 * filters the task of Gallery.app itself.
 * allows the gallery.app to load task icons.

BUG=358698
TEST=None

Review URL: https://codereview.chromium.org/282303006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271012 0039d316-1c4b-4281-b951-d872f2087c98

For unittest that only testing functionatility of Append DrawQuads and
SharedQuadState, only creates RenderPass for it.

This CL is first part of clean up QuadSink. Next step would be to use
test occlusion tracker for MockQuadSink.

BUG=
TEST=cc_unittests

Review URL: https://codereview.chromium.org/288273003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271011 0039d316-1c4b-4281-b951-d872f2087c98

BUG=372564

Review URL: https://codereview.chromium.org/287003006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271008 0039d316-1c4b-4281-b951-d872f2087c98

Two bugs, and general state-machine grotesqueness, prevented auto-reload
histograms from working properly in production:

1. Success could only be reported in OnCommitLoad() if can_auto_reload_page_ was
true, but can_auto_reload_page_ would be set to false in OnStartLoad(), since
OnStartLoad() had no way of knowing that the non-error page load it saw starting
was actually part of an attempt to auto-reload the page. This has been patched
around by tracking is_auto_reloading in committed_error_page_info_; when the
auto reload timer fires, it sets is_auto_reloading to indicate to OnCommitLoad
that auto-reload was running. OnCommitLoad can then use is_auto_reloading and
the loading URL to decide whether the committing load represents a success or
failure for auto-reload.

2. Whenever a non-error load started, OnStartLoad() would call
CancelPendingFetches(), intending to prevent auto-reload from replacing the
starting load with its own; as a side-effect, CancelPendingFetches() would
assume that if there was an existing error page which looked auto-reloadable
that the cancel represented a failure of auto-reload and log a histogram entry.
Unfortunately, this logic would trigger at each reload attempt by auto-reload,
inflating the failure count with nonexistent failures. To prevent this
misbehavior, CancelPendingFetches() is now no longer responsible for logging
auto-reload failures; instead, failures are detected and logged in
OnCommitLoad() (for "user navigated elsewhere" failures), OnStop (for "user
pressed Stop to abort auto-reload" failures), and the NetErrorHelperCore
destructor (for "user closed the tab" failures).

To help clean up the code, statistics are now reported through
ReportAutoReloadSuccess() and ReportAutoReloadFailure().

There are new unit tests covering these cases:
1. Auto-reload succeeding on the first attempt;
2. Auto-reload succeeding on the second attempt;
3. The user manually stopping an auto-reload;
4. The user manually stopping a non-auto-reload load;
5. The user navigating to a different URL;

As part of this cleanup, simplify the state machine; the ambiguous "can_auto_reload_page_" flag is gone, replaced by an implicit state bit in whether the timer is running (or paused) or not.

BUG=

Review URL: https://codereview.chromium.org/259613003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271007 0039d316-1c4b-4281-b951-d872f2087c98

BUG=346210

Review URL: https://codereview.chromium.org/283563002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271005 0039d316-1c4b-4281-b951-d872f2087c98

Re-landing https://codereview.chromium.org/272183005
with leak fix.

BUG=372673
TBR=michaeln@chromium.org,horo@chromium.org

Review URL: https://codereview.chromium.org/282263008

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271004 0039d316-1c4b-4281-b951-d872f2087c98

Remove ArrayBuffer ctor that is not defined.
Add explicit assignment operator for ArrayBufferView.

BUG=none
R=dcarney@chromium.org,mnaganov@chromium.org

Review URL: https://codereview.chromium.org/289003007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271003 0039d316-1c4b-4281-b951-d872f2087c98

When browser navigates to another page, make sure we de-associate
NPObjects of injected Java objects from their old V8 wrappers.

Not doing this leads to potential leaks of JS objects from one
context to another.

BUG=372914
R=jochen@chromium.org

Review URL: https://codereview.chromium.org/286823002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271002 0039d316-1c4b-4281-b951-d872f2087c98

Revert of [Mac] Unselect generated password when mouse leaves password bounds. (https://codereview.chromium.org/282093006/)

Reason for revert:
Speculative revert, FormFillLatencyAfterSubmit is failing

Original issue's description:
> [Mac] Unselect generated password when mouse leaves password bounds.
> 
> In the Mac implementation of the password generation poup, the password
> row is not unselected when the mouse is moved elsewhere in the popup
> (e.g. over top of the help text).
> 
> The fix is that PasswordSelected() should always be called in
> SetSelectionAtPoint, not just when the mouse is within the password
> bounds. The Views implementation happens to work because it has a
> subview (PasswordRow) which triggers OnMouseExited() when the mouse
> moves outside of the password bounds.
> 
> Also added an optimization that PasswordSelectionUpdated() is only
> called when the value is actually changed, not every time
> PasswordSelected() is called.
> 
> BUG=114092
> R=isherman@chromium.org
> 
> Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=270975

TBR=gcasto@chromium.org,isherman@chromium.org,dubroy@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=114092

Review URL: https://codereview.chromium.org/287063007

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271001 0039d316-1c4b-4281-b951-d872f2087c98

Currently, if there is an access to the pipe, whose other side is
already closed, SIGPIPE will be raised in some platform such as linux.
It is unexpected behavior, and causes some test failure.
This is what trusted plugin does in content/app/content_main_runner.cc.

TEST=Ran PPAPINaClPNaClNonSfiTest.Audio locally many times and the flakiness
was gone. Also run trybots.
BUG=371662

Review URL: https://codereview.chromium.org/288063004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@271000 0039d316-1c4b-4281-b951-d872f2087c98

BUG=371845
TBR=jochen

Review URL: https://codereview.chromium.org/289993003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270999 0039d316-1c4b-4281-b951-d872f2087c98

We now only call this method for the main world (world_id == 0)

BUG=none
R=bauerb@chromium.org

Review URL: https://codereview.chromium.org/292493002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270998 0039d316-1c4b-4281-b951-d872f2087c98

When the next sync point overflow occurs, it returns 1, not 0.

BUG=365904, 373452

Review URL: https://codereview.chromium.org/288153004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270997 0039d316-1c4b-4281-b951-d872f2087c98

When a user do non-Tap (e.g. LongPress for drag, LongTap for context menu) operations on touch panel, 'mousedown' is not dispatched immediately.
In these cases, context menu should be hided because user doesn't focus on the context menu anymore.
This CL hide the menu on 'touchstart' in the same manner as 'mousedown'.

BUG=267211

Review URL: https://codereview.chromium.org/283363002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270996 0039d316-1c4b-4281-b951-d872f2087c98

BUG=358657

Review URL: https://codereview.chromium.org/288683002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270995 0039d316-1c4b-4281-b951-d872f2087c98

BUG=367114

Review URL: https://codereview.chromium.org/272503004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270994 0039d316-1c4b-4281-b951-d872f2087c98

malloc_usable_size has been changed in android ndk header.
(https://code.google.com/p/android/issues/detail?id=55725)
so to override the function properly, this header should be changed too.

BUG= 162208

Review URL: https://codereview.chromium.org/284843002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270993 0039d316-1c4b-4281-b951-d872f2087c98

Mac OS X AssertMacros.h defining macros that collide with common names, like
'check', 'require', and 'verify'.
This CL turns this behavior off by specifying
define __ASSERT_MACROS_DEFINE_VERSIONS_WITHOUT_UNDERSCORE 0
...allowing us to use these words normally.

Refs:
http://opensource.apple.com/source/CarbonHeaders/CarbonHeaders-18.1/AssertMacros.h

[blink-dev] Fix collision with Apple macros?
https://groups.google.com/a/chromium.org/d/topic/blink-dev/MIjO2Gwc3qc/discussion

Review URL: https://codereview.chromium.org/277273002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270992 0039d316-1c4b-4281-b951-d872f2087c98

With the commit "gpu: Reuse transfer buffers more aggresively" mapped
memory is free:ed on lost context. This means that we can reinstate the
DCHECKs in the FencedAllocator destructor that checks that all blocks
have been free:ed by their owner.

Tests that didn't explicitly free mapped memory now do so, and one test
that tested that leaking was OK has been removed.

BUG=328808

Review URL: https://codereview.chromium.org/282513003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270991 0039d316-1c4b-4281-b951-d872f2087c98

BUG=148383
TBR=thakis@chromium.org

Review URL: https://codereview.chromium.org/288743002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270990 0039d316-1c4b-4281-b951-d872f2087c98

- use-after-free in media::AlsaWrapper::PcmWritei (issue 374135)
- lock inversion in media::FakeAudioConsumer::Worker::DoRead (issue 374151)

BUG=374135,374151
TBR=benwells@chromium.org
NOTRY=true

Review URL: https://codereview.chromium.org/287153004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270989 0039d316-1c4b-4281-b951-d872f2087c98

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270988 0039d316-1c4b-4281-b951-d872f2087c98

Please reply to the V8 sheriff danno@chromium.org in case of problems.

TBR=danno@chromium.org

Review URL: https://codereview.chromium.org/288193003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270987 0039d316-1c4b-4281-b951-d872f2087c98

BUG=373385
R=jochen@chromium.org
NOTRY=true

Review URL: https://codereview.chromium.org/289033006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270986 0039d316-1c4b-4281-b951-d872f2087c98

This patch adds the read file operation, which is however not bound yet to
fileapi.

In the upcoming patch, support for FileStreamReader will be added. After that,
OpenFile, CloseFile and ReadFile will be bound to async file utils.

TEST=unit_tests: *FileSystemProvider*ReadFile*
BUG=248427

Review URL: https://codereview.chromium.org/287673004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270985 0039d316-1c4b-4281-b951-d872f2087c98

BUG=373594

Review URL: https://codereview.chromium.org/289963002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270984 0039d316-1c4b-4281-b951-d872f2087c98

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270983 0039d316-1c4b-4281-b951-d872f2087c98

http://build.chromium.org/f/chromium/perf/dashboard/ui/changelog_blink.html?url=/trunk&range=174125:174137&mode=html
TBR=
BUG=

Review URL: https://codereview.chromium.org/287163004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270982 0039d316-1c4b-4281-b951-d872f2087c98

Re-enable org.chromium.android_webview.test.ExternalVideoSurfaceContainerTest#testEnableVideoOverlayForEmbeddedVideo

The fix was landed: https://codereview.chromium.org/289653002

BUG=372748,372174

Review URL: https://codereview.chromium.org/292453004

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270979 0039d316-1c4b-4281-b951-d872f2087c98

The separated version of Gallery does not need the button.

BUG=358698
TEST=manually

Review URL: https://codereview.chromium.org/291483006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270978 0039d316-1c4b-4281-b951-d872f2087c98

PeerConnectionDependencyFactory should only deal with MediaStream tracks that are sent on a PeerConnection and thus is not a general MediaStream factory.
This cl also moves PeerconnectionDependencyFactory to media/webrtc since it depends on libjingle/webrtc files.

Note that PeerConnectionDependencyFactory still have some methods related to general audio input that should be further refactored.

BUG=323223

Review URL: https://codereview.chromium.org/272043003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270977 0039d316-1c4b-4281-b951-d872f2087c98

Revert of Stop ServiceWorker context when no controllee is associated (and when all refs are dropped) (https://codereview.chromium.org/272183005/)

Reason for revert:
Has memory leaks in
EmbeddedWorkerInstanceTest.InstanceDestroyedBeforeStartFinishes

Original issue's description:
> Stop ServiceWorker context when no controllee is associated (and when all refs are dropped)
> 
> Currently we stop the ServiceWorker thread/context when:
> 1. After all documents that are controlled by the SW are closed, with a delay (5 min)
> 2. All references to the version are dropped (i.e. in dtor)
> 
> In the current impl/codebase a worker is immediately stopped by 2
> in most cases.
> 
> The same stop timer (ScheduleStopWorker()) could be probably called
> each time handling event is finished to stop the worker more 
> aggressively, though this patch doesn't do so.
> 
> BUG=372673
> TEST=tested manually with chrome://serviceworker-internals/
> 
> Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=270971

TBR=michaeln@chromium.org,horo@chromium.org,kinuko@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=372673

Review URL: https://codereview.chromium.org/286203008

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270976 0039d316-1c4b-4281-b951-d872f2087c98

In the Mac implementation of the password generation poup, the password
row is not unselected when the mouse is moved elsewhere in the popup
(e.g. over top of the help text).

The fix is that PasswordSelected() should always be called in
SetSelectionAtPoint, not just when the mouse is within the password
bounds. The Views implementation happens to work because it has a
subview (PasswordRow) which triggers OnMouseExited() when the mouse
moves outside of the password bounds.

Also added an optimization that PasswordSelectionUpdated() is only
called when the value is actually changed, not every time
PasswordSelected() is called.

BUG=114092
R=isherman@chromium.org

Review URL: https://codereview.chromium.org/282093006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270975 0039d316-1c4b-4281-b951-d872f2087c98

This reverts changes user_manager_impl.cc that were lost in previous
revert (https://codereview.chromium.org/284333002) because file got moved.

BUG=331530
TBR=antrim

Review URL: https://codereview.chromium.org/288923010

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270974 0039d316-1c4b-4281-b951-d872f2087c98

Revert of Re-enable key silk case benchmarks on Android (https://codereview.chromium.org/279853002/)

Reason for revert:
Still failing on N5. e.g. 

http://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5%20Perf/builds/393/steps/smoothness.fast_path.key_silk_cases/logs/stdio

BUG=355952

Original issue's description:
> Re-enable key silk case benchmarks on Android
> 
> All the known issues with these telemetry benchmarks have been fixed.
> 
> BUG=355952
> 
> Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=270830

TBR=ernstm@chromium.org,skyostil@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=355952

Review URL: https://codereview.chromium.org/292453005

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270973 0039d316-1c4b-4281-b951-d872f2087c98

Currently we stop the ServiceWorker thread/context when:
1. After all documents that are controlled by the SW are closed, with a delay (5 min)
2. All references to the version are dropped (i.e. in dtor)

In the current impl/codebase a worker is immediately stopped by 2
in most cases.

The same stop timer (ScheduleStopWorker()) could be probably called
each time handling event is finished to stop the worker more 
aggressively, though this patch doesn't do so.

BUG=372673
TEST=tested manually with chrome://serviceworker-internals/

Review URL: https://codereview.chromium.org/272183005

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270971 0039d316-1c4b-4281-b951-d872f2087c98

The existing test code returns an empty ImageSkiaRep where
bytes are allocated but nothing is initialized. This was okay
because ImageSkiaTest doesn't touch the contents of the image.

Now ImageSkia can rescale the image data and such cases are
tested. So using a dummy ImageSkiaRep wouldn't be helthy.
Rather some data should be drawn.

BUG=373748
R=oshima@chromium.org
TEST=

Review URL: https://codereview.chromium.org/285393002

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@270970 0039d316-1c4b-4281-b951-d872f2087c98