GitLab

Accumulating large draw command backlogs can result in janky
performance, and in extreme cases, a GPU denial of service. This
change fixes that problem by disabling deferred rendering when
high overdraw is detected. Cases where the overdraw overwrites
the entire canvas are not affected by this change and will
stay in deferred rendering mode because overwriting the entire
canvas resets the backlog.

BUG=522764, 522972

Review URL: https://codereview.chromium.org/1308853003

git-svn-id: svn://svn.chromium.org/blink/trunk@201036 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=522614

Review URL: https://codereview.chromium.org/1305373002

git-svn-id: svn://svn.chromium.org/blink/trunk@201034 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch moves |{next,previous}PositionOf()| to "EditingUtilities.{cpp,h}"
from "Position.{cpp,h}" to simplify "Position.{cpp,h}" for improving code
health since |{next,previous}PositionOf()| don't depend of implementation of
|PositionAlgorithm| template class.

BUG=518738
TEST=n/a; no behavior changes

Review URL: https://codereview.chromium.org/1310533002

git-svn-id: svn://svn.chromium.org/blink/trunk@201033 bbb929c8-8fbe-4397-9dbb-9b2b20218538

These unit tests verify that QuotaExceededErr is thrown when appending
data to an MSE SourceBuffer that is full. This new behavior was
ntroduced by the recent CLs:
https://codereview.chromium.org/1008463002/
https://codereview.chromium.org/1013923002/

BUG=421694

Review URL: https://codereview.chromium.org/1304953002

git-svn-id: svn://svn.chromium.org/blink/trunk@201032 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Manually rebaselining a pair of tests that fail on Debug due
to limitations of specifying Release/Debug for NeedsRebaseline.

BUG=509025
TBR=joelo

Review URL: https://codereview.chromium.org/1310023002

git-svn-id: svn://svn.chromium.org/blink/trunk@201031 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is a followup CL of https://codereview.chromium.org/976543002/. In
that CL, sub-pixel accumulation is considered to calculate layerBounds,
foregroundBounds and backgroundBounds if renderLayer doesn't paint into
its own backing layer. But in that CL, sub-pixel accumulation didn't
apply to foregroundBounds and backgroundBounds for unpaginated layers.
This CL fixes above issue.

BUG=515395

Review URL: https://codereview.chromium.org/1278113003

git-svn-id: svn://svn.chromium.org/blink/trunk@201030 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Test is still timing out, but since the test doesn't log anything, it's
hard to tell if the test actually starts running. Adding some logging
so that on failures we can tell how far the test gets.

BUG=522646
TEST=test passes

Review URL: https://codereview.chromium.org/1308853005

git-svn-id: svn://svn.chromium.org/blink/trunk@201029 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We made LayoutObject::previousPaintInvalidationRect() private but
DeprecatedPaintLayerScrollArea is a friend of LayoutObject, so missed
the change. This caused incorrect fake mouse move event on composited
scroll.

Will fix the friend access issue in
https://codereview.chromium.org/1306993002/.

BUG=516016
TEST=fast/events/overflow-composited-scroll-fake-mouse-move.html

Review URL: https://codereview.chromium.org/1309983002

git-svn-id: svn://svn.chromium.org/blink/trunk@201028 bbb929c8-8fbe-4397-9dbb-9b2b20218538

m_documentLoader is null. Here's the trace:

html_viewer.mojo!blink::DocumentLoader::isCommittedButEmpty() Line 105
html_viewer.mojo!blink::FrameLoader::finishedParsing() Line 485
html_viewer.mojo!blink::FrameLoader::stopAllLoaders() Line 981
html_viewer.mojo!blink::LocalFrame::detach(blink::FrameDetachType type) Line 281
html_viewer.mojo!blink::WebFrame::swap(blink::WebFrame * frame) Line 114

BUG=521663
TEST=none
R=japhet@chromium.org

Review URL: https://codereview.chromium.org/1298643003

git-svn-id: svn://svn.chromium.org/blink/trunk@201027 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=509025
TBR=joelo

Review URL: https://codereview.chromium.org/1312583002

git-svn-id: svn://svn.chromium.org/blink/trunk@201026 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Using dirtyLinesFromChangedChild() was the wrong choice in
https://codereview.chromium.org/1167543008 as it can dirty the
lineboxes of descendants and ancestors. We're only interested
in dirtying the lineboxes of the current parent as it is
only to one of them we will possibly attach the float.

I have more to do here to understand why I'm not getting a
height on the span in my new test, but this successfully
addresses the assert in bug 510703 and seems obviously more
correct than what I currently have.


BUG=510703

Review URL: https://codereview.chromium.org/1308433003

git-svn-id: svn://svn.chromium.org/blink/trunk@201025 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is a preparation for a series of paint invalidation changes
for slimming paint v2.

- Combine layoutDidGetCalledSinceLastFrame into
  mayNeedPaintInvalidation because we handle them in the same way;

- Separate shouldCheckForChildPaintInvalidation from
  mayNeedPaintInvalidation to meet future slimming paint v2 needs;

- Mark some bit fields for future removal for slimming paint v2;

- Use direct m_bitfields access for private and simple setting and
  clearing of paint invalidation flags.

BUG=510908

Review URL: https://codereview.chromium.org/1302993003

git-svn-id: svn://svn.chromium.org/blink/trunk@201024 bbb929c8-8fbe-4397-9dbb-9b2b20218538

EventHandler currently have two state vars for tracking
node under mouse: m_nodeUnderMouse and m_lastNodeUnderMouse.
These two pointers are used primarily for mouse entry/exit
events, but they point to the same node until we are about
to send mouseenter/leave/over/out events. In the few cases
these pointers become unequal otherwise (happens only when
dispathMouseEvent is called with setUnder=false), it is not
clear why we didn't update the old state. My experiments
suggest that updating the old state doesn't hurt in these
cases. So it makes sense to nuke the m_lastNodeUnderMouse
state.

BUG=

Review URL: https://codereview.chromium.org/1293973004

git-svn-id: svn://svn.chromium.org/blink/trunk@201023 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Take care of a FIXME in SecurityOrigin.cpp to check the validity of any innerURL during construction and to handle invalidity as a 'unique' origin.

TBR=mkwst
BUG=514076

Review URL: https://codereview.chromium.org/1294933004

git-svn-id: svn://svn.chromium.org/blink/trunk@201022 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This test has been flaky all along because the sequence of events it
tries to observe is not fully observable. There's still some coverage of
the reset-animation-when-no-longer-running case elsewhere.

BUG=522652

Review URL: https://codereview.chromium.org/1292703004

git-svn-id: svn://svn.chromium.org/blink/trunk@201021 bbb929c8-8fbe-4397-9dbb-9b2b20218538

It still references LayoutState.

Review URL: https://codereview.chromium.org/1312533002

git-svn-id: svn://svn.chromium.org/blink/trunk@201020 bbb929c8-8fbe-4397-9dbb-9b2b20218538

CascadeOrder is always ignoreCascadeOrder after [1] and can be removed.

[1] https://codereview.chromium.org/1298173004

BUG=487125

Review URL: https://codereview.chromium.org/1291873005

git-svn-id: svn://svn.chromium.org/blink/trunk@201019 bbb929c8-8fbe-4397-9dbb-9b2b20218538

No change in behavior, just a minor cleanup.

NOTRY=true
R=leviw

Review URL: https://codereview.chromium.org/1308093002

git-svn-id: svn://svn.chromium.org/blink/trunk@201018 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Both dispatchDidLoadResourceFromMemoryCache() and
sendRemainingDelegateMessages() are only called when a cache
hit occurs. Combine their logic into dispatchDidLoadResourceFromMemoryCache().

BUG=

Review URL: https://codereview.chromium.org/1309533002

git-svn-id: svn://svn.chromium.org/blink/trunk@201017 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch gets rid of |PositionAlgorithm<Strategy>::{upstream,downstream}| by
replacing them by |most{Forward,Backward}CaretPosition()| to simplify
|PositionAlgorithm<Strategy>| template class for improving code health, since
|PositionAlgorithm<Strategy>::{upstream,downstream}| are simple wrapper
functions.

BUG=518738
TEST=n/a; no behavior changes

Review URL: https://codereview.chromium.org/1302353002

git-svn-id: svn://svn.chromium.org/blink/trunk@201016 bbb929c8-8fbe-4397-9dbb-9b2b20218538

And add a layout test.

BUG=520717

Review URL: https://codereview.chromium.org/1301393002

git-svn-id: svn://svn.chromium.org/blink/trunk@201015 bbb929c8-8fbe-4397-9dbb-9b2b20218538

NOTRY=true
TBR=

Review URL: https://codereview.chromium.org/1312513002

git-svn-id: svn://svn.chromium.org/blink/trunk@201014 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL adds two necessary methods to the line layout API and updates callers
to use them.

1. LineLayoutBoxModel::boxShadowShouldBeAppliedToBackground
2. LineLayoutItem::enclosingBoxModelObject (implemented in LineLayoutBoxModel.h
   to avoid pulling LayoutBoxModel into LineLayoutItem)

BUG=499321

Review URL: https://codereview.chromium.org/1303303002

git-svn-id: svn://svn.chromium.org/blink/trunk@201013 bbb929c8-8fbe-4397-9dbb-9b2b20218538

A previous commit removed support for webkitGetUserMedia from insecure
origins. This removes support from the still-behind-a-flag, promise
based mediaDevices.getUserMedia from insecure origins as well.

BUG=520765

Review URL: https://codereview.chromium.org/1311473003

git-svn-id: svn://svn.chromium.org/blink/trunk@201012 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=estark
NOTRY=true
BUG=523500

Review URL: https://codereview.chromium.org/1311503003

git-svn-id: svn://svn.chromium.org/blink/trunk@201011 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=523164

Review URL: https://codereview.chromium.org/1304023002

git-svn-id: svn://svn.chromium.org/blink/trunk@201010 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=523165

Review URL: https://codereview.chromium.org/1304943002

git-svn-id: svn://svn.chromium.org/blink/trunk@201009 bbb929c8-8fbe-4397-9dbb-9b2b20218538

NOTRY=true

Review URL: https://codereview.chromium.org/1308563002

git-svn-id: svn://svn.chromium.org/blink/trunk@201008 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Background:
To disable scrolling (and pinch-zoom) during <video>-
fullscreen ScrollingCoordinator makes both viewports
unscrollable in updateAfterCompositingChangeIfNeeded().

Problem:
Currently only the outer viewport's scrollability-
setting is tested during fullscreen-video.

Solution:
Extend test case to also verify the inner viewport.

BUG=448738

Review URL: https://codereview.chromium.org/1287253005

git-svn-id: svn://svn.chromium.org/blink/trunk@201007 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Also, remove android's resource-request-callbacks-expected.txt,
which I missed when deleting the test in https://codereview.chromium.org/1288323006/

BUG=523091

Review URL: https://codereview.chromium.org/1309893002

git-svn-id: svn://svn.chromium.org/blink/trunk@201006 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=oilpan-reviews
BUG=514651
NOTRY=true

Review URL: https://codereview.chromium.org/1309593003

git-svn-id: svn://svn.chromium.org/blink/trunk@201005 bbb929c8-8fbe-4397-9dbb-9b2b20218538

1. Truncate trailing zeroes in doubles.
2. Ignore negative values in paddings.

BUG=501896

Review URL: https://codereview.chromium.org/1307813002

git-svn-id: svn://svn.chromium.org/blink/trunk@201004 bbb929c8-8fbe-4397-9dbb-9b2b20218538

To implement FOUC tests we need to be able to control both the
didReceiveData and the didFinishLoading callbacks to have precise
control over when resources finish loading.

This patch expands the interface of WebURLLoaderTestDelegate to delegate
all the steps called in WebURLLoaderMock::ServeAsynchronousRequest. It
also provides a default implementation of each one so tests don't need to
add each proxy method. It also allows this patch to avoid being multiple
steps to handle the Chromium changes.

The next patch in Chromium will update ServeAsynchronousRequest
to use the new methods.

BUG=521692
R=pdr@chromium.org

Review URL: https://codereview.chromium.org/1303953003

git-svn-id: svn://svn.chromium.org/blink/trunk@201003 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Previously because one table cell may paint its collapsed borders
multiple times, we enclose all collapsed border drawings in a
ScopeRecorder.

However, actually a table cell paints each border only once. When it
paint collapsed borders multiple times, it just paint different borders
in different time.

Use display item type to distinguish the different borders to avoid
the ScopeRecorder.

Also revert https://codereview.chromium.org/1234463005 to invalidate
table cells with changed collapsed borders.

BUG=510492

Review URL: https://codereview.chromium.org/1306433002

git-svn-id: svn://svn.chromium.org/blink/trunk@201002 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=353746
TBR=
NOTRY=true

Review URL: https://codereview.chromium.org/1309903002

git-svn-id: svn://svn.chromium.org/blink/trunk@201001 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is a follow-up to https://codereview.chromium.org/1291613010/ where
I forgot to implement the referrerpolicy attribute for img tags as a
runtime-enabled feature.

BUG=490608

Review URL: https://codereview.chromium.org/1302403005

git-svn-id: svn://svn.chromium.org/blink/trunk@201000 bbb929c8-8fbe-4397-9dbb-9b2b20218538

After discussions w/ Moz and spec updates, modify the experimental
(behind a flag) IDBObjectStore/IDBIndex getAll()/getAllKeys() methods
to treat 0 the same as "not passed", and return as many results as
possible.

Also adds missing tests for IDBObjectStore.getAllKeys() and tweaks
the tests to avoid the deprecated assert_object_equals() function.

R=cmumford@chromium.org

Review URL: https://codereview.chromium.org/1303983002

git-svn-id: svn://svn.chromium.org/blink/trunk@200999 bbb929c8-8fbe-4397-9dbb-9b2b20218538

In web-platform-tests there's a move away from the generic (and
fragile, buggy) assert_object_equals(). Upstream, the cache storage
tests moved to dedicated assert_response_equals() (and friends);
mirror that in Blink.

R=jkarlin@chromium.org

Review URL: https://codereview.chromium.org/1295633003

git-svn-id: svn://svn.chromium.org/blink/trunk@200998 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Introduce {next,previous}PositionOf() as replacement of {next,previous}() member functions in PositionAlgorithm

This patch introduces |{next,previous}PositionOf()| as replacement of
|{next, previous}()| member function in |PositionAlgorithm| template class
to simplify |PositionAlgorithm| class by isolating editing algorithm and
representation of a position for improving code health.

This patch also changes |PositionMoveType| as enum class and non-default
parameter to clear intention of each call sites. Though, we may want to drop
|CodePoint| movement.

Following patch will move them to "EditingUtilities.{cpp,h}".

BUG=518738
TEST=n/a; no behavior changes

Review URL: https://codereview.chromium.org/1305963003

git-svn-id: svn://svn.chromium.org/blink/trunk@200997 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=

Review URL: https://codereview.chromium.org/1309653002

git-svn-id: svn://svn.chromium.org/blink/trunk@200996 bbb929c8-8fbe-4397-9dbb-9b2b20218538