GitLab

PLY yacc parse table creation is (obviously) expensive.
Pre-caching this in a separate build step substantially improves build time.

Improves build time (user time) by 45% on my Linux box:
1m 50s => 59s

BUG=341748
R=haraken

Review URL: https://codereview.chromium.org/184233005

git-svn-id: svn://svn.chromium.org/blink/trunk@168611 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is the compute_interfaces_info.py side of the post-split cleanup.
This finishes the build factoring.
(Still need to add more caching for speed and some more use
of Jinja for clarity, but mostly done with build.)

Two notable changes:
* We only need to store *inherited* extended attributes, not all extended
　attributes (in auxiliary variable; needed more for event interfaces, gone now).
　This simplifies the inheritance resolution, and reduces duplication.

* Always include the same keys in the info dict, even if the value is empty.
　This simplifies both setting (since no test) and use:
　instead of having "if key in dict and dict[key]..."
　we can just have "if dict[key]".
　It also helps consistency, since we were doing it 2 ways before.

BUG=341748
R=haraken

Review URL: https://codereview.chromium.org/183853024

git-svn-id: svn://svn.chromium.org/blink/trunk@168610 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The failure happened because live-edit-no-reveal.html live edited the same script and this somehow
interfered with live-edit.html execution.

BUG=282082
TBR=pfeldman

Review URL: https://codereview.chromium.org/185683008

git-svn-id: svn://svn.chromium.org/blink/trunk@168609 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=348887
R=dgozman

Review URL: https://codereview.chromium.org/185943004

git-svn-id: svn://svn.chromium.org/blink/trunk@168608 bbb929c8-8fbe-4397-9dbb-9b2b20218538

In order to do so, make the IDL compiler able to emit code that uses
heap vectors if the types involved are of the garbage collected kind.
(For the most part done by adopting the changes from
 https://codereview.chromium.org/173363002 )

R=
BUG=340522

Review URL: https://codereview.chromium.org/185413023

git-svn-id: svn://svn.chromium.org/blink/trunk@168607 bbb929c8-8fbe-4397-9dbb-9b2b20218538

According to the specification, certain media queries accept a CSS length as their value (e.g. http://www.w3.org/TR/css3-mediaqueries/#width).
A CSS length (http://www.w3.org/TR/css3-values/#lengths) must have a known unit, unless the value is zero.
This was not properly represented in the MediaQueryExp code, where any number was accepted.
So such media queries (e.g. "(max-width: 1)") were considered valid, but evaluated to false.
This CL fixes that, making such MQs invalid.

BUG=

Review URL: https://codereview.chromium.org/186603002

git-svn-id: svn://svn.chromium.org/blink/trunk@168606 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This reduces the amount of time spent in compositing update for Silk's
toggle_drawer test case by 20%. Previously, we would update the geometry for
every composited layer mapping during every compositing update, regardless of
whether any of the inputs to updateGraphicsLayerGeometry had changed. This CL
introduces dirty bits on CompositedLayerMapping to track which RenderLayers
underwent a change that could have resulted in changes during
updateGraphicsLayerGeometry.

This CL takes a basic approach, which is sufficient for toggle_drawer.
Specifically, if a RenderLayer changes style, we mark its
CompositedLayerMapping (and any ancestors and decendants) as needing a geometry
update. If we go through layout or scroll, we trigger a forced update, which
updates the geometry of all graphics layers.

Over time, we can improve this system to track dirtiness during layout. We
might also be able to avoid marking every ancestor dirty if we removed the
downward tree walks during updateGraphicsLayerGeometry. Instead, we could
compute the descendant-dependant information in an earlier phase and detect
whether it changed directly.

R=esprehn@chromium.org, ojan@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=168554

Review URL: https://codereview.chromium.org/183763016

git-svn-id: svn://svn.chromium.org/blink/trunk@168604 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The purpose of the funciton was to describe support differences between desktop
platforms and Android. Android Chrome now supports same input types for
<datalist>. We don't need the function any longer.

Note for BaseMultipleFieldsDateAndTimeInputType::createShadowSubtree:
|shouldAddPickerIndicator| is always true. We can remove it.

BUG=none
TEST=none; no behaivor changes.

Review URL: https://codereview.chromium.org/181013012

git-svn-id: svn://svn.chromium.org/blink/trunk@168603 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The file is not being referenced from anywhere. Its functionality was moved to SVGAnimatedEnumeration.h

BUG=349370
TBR=haraken,pdr,fs

Review URL: https://codereview.chromium.org/188413002

git-svn-id: svn://svn.chromium.org/blink/trunk@168602 bbb929c8-8fbe-4397-9dbb-9b2b20218538

RenderLayerScrollableArea::updateScrollCornerStyle shouldn't do work if we're using overlay scrollbars

This function showed up on a profile of updateStyle in Silk's toggle_drawer
test case, but there no reason to update the scroll corner style if we're using
overlay scrollbars. If the author has created a custom scrollbar, we'll realize
that we don't have overlay scrollbars.

R=esprehn@chromium.org
BUG=335703

Review URL: https://codereview.chromium.org/178903004

git-svn-id: svn://svn.chromium.org/blink/trunk@168601 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Follow-up to the split CL.
Just clean-up, no changes to generated EventInterfaces.in.

* Rename to generate_event_interfaces (from "*_names")
This is b/c:
- the output file is called EventInterfaces.in
- the output file is used for both EventFactory and EventNames

* Remove EventInterfaces.in from run-bindings-tests
This file is useless, and adds unnecessary complexity to r-b-t.
The actual output file is trivial:
namespace="Event"

Source/bindings/tests/idls/TestInterfaceEventConstructor
...and I've never looked at it during CG changes.

With it, we need to have 2 passes of compute_interfaces_info,
and a separate temporary file.
Without it, once Terry's CL lands, we'll have no individual files,
and we can just use one temporary directory, removing lots of
temp file complexity.

* Compute source_dir robustly (not using getcwd)

* Refactoring and renaming

Follow-up to:
Split generate_event_interfaces.py from compute_interfaces_info.py
https://codereview.chromium.org/185303008/

BUG=341748
R=haraken

Review URL: https://codereview.chromium.org/186433003

git-svn-id: svn://svn.chromium.org/blink/trunk@168599 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL changes GraphicsLayer::removeAllChildren to be linear in the number of
children instead of n^2. Previously, we were removing the zero-th element of
the vector in a loop, which causes a shift of all the elements of the vector.
Now we walk the vector from the end, which means remove is constant cost.

This code is called during a compositing update if anything changes about the
GraphicsLayer's children. We didn't actually see this code on our profile
because we're not actually changing the GraphicsLayer tree, but it seemed worth
fixing.

R=ojan@chromium.org

Review URL: https://codereview.chromium.org/187603011

git-svn-id: svn://svn.chromium.org/blink/trunk@168598 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch makes Animations get serviced in the order specified by the
spec[1]. This means that once custom effect are implemented, they will
be queued up in the correct order. This also gets us closer to having
the animation stack in a correct order (I'm thinking of emptying the
stack after getting out the compositable values, as we traverse the
animations in the correct order).

[1] http://dev.w3.org/fxtf/web-animations/#the-animation-stack

BUG=334936

Review URL: https://codereview.chromium.org/185593010

git-svn-id: svn://svn.chromium.org/blink/trunk@168597 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This function turned up in one of our profiles. I'm not sure this change makes
much of a difference, but we might as well not have silly code on the profile.

R=esprehn@chromium.org

Review URL: https://codereview.chromium.org/187793005

git-svn-id: svn://svn.chromium.org/blink/trunk@168596 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Add tests for sticky and various display values, tests should show up all green.

BUG=231754

Review URL: https://codereview.chromium.org/178453002

git-svn-id: svn://svn.chromium.org/blink/trunk@168595 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This code is not in use anymore, mostly because of bindings changes.

Review URL: https://codereview.chromium.org/183973036

git-svn-id: svn://svn.chromium.org/blink/trunk@168594 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The recently added 'HashTraits::PeekInType' type allows us to remove
HashMap<RefPtr..> template specialization without loosing the efficiency
and without bringing any significant changes and complexity to the HashMap
class.

To get rid of 'RefPtrHashMap' a new 'RefPtrValuePeeker' type was created
and used as 'HashTraits::PeekInType' for RefPtr. The 'RefPtrValuePeeker'
class can be constructed either from RefPtr, PassRefPtr or a plain pointer,
then it behaves like a plain pointer itself within HashTable methods.

The proposed change brings the following benefits:
1) Removes tons of duplicated code
2) Allows avoiding ref-count churn also at HashSet<RefPtr>

Review URL: https://codereview.chromium.org/184233006

git-svn-id: svn://svn.chromium.org/blink/trunk@168593 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The assertion failure in HarfBuzzShaper is triggered because the TextRun
passed to paintTextWithShadows is replaced with
objectReplacementCharacterTextRun which only has a length of 1.
Previously, we did not have an implementation of emphasis mark
drawing for complex text with HarfBuzz, so instead of running into the
assertion, those paint calls were not doing anything.  The indices have
been wrong since the emphasis marks drawing for RenderCombineText
elements was fixed in 7565e3bbdccf98450271 (in 2011).  I updated the
test case to be equivalent to the original fuzz testing report and so
that it actually combines text that is marked as combined where no
font with narrow digits exists.

BUG=348682

Review URL: https://codereview.chromium.org/187783002

git-svn-id: svn://svn.chromium.org/blink/trunk@168592 bbb929c8-8fbe-4397-9dbb-9b2b20218538

See the bug and http://crrev.com/255129 for the details.

BUG=345554
R=haraken@chromium.org, jamesr@chromium.org

Review URL: https://codereview.chromium.org/177053003

git-svn-id: svn://svn.chromium.org/blink/trunk@168591 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=168578

BUG=349222
TBR=qinmin@chromium.org

Review URL: https://codereview.chromium.org/187813010

git-svn-id: svn://svn.chromium.org/blink/trunk@168590 bbb929c8-8fbe-4397-9dbb-9b2b20218538

These header files are not being referenced from anywhere. They were replaced by NewSVGAnimatedProperty.h

BUG=349370
TBR=haraken,pdr,fs

Review URL: https://codereview.chromium.org/188413003

git-svn-id: svn://svn.chromium.org/blink/trunk@168589 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Now TestRunner is being moved from CppBoundClass to gin::Wrappable,
where it is not appropriate to use WebElements directly. This CL moves
some functions which treats WebElement* from testRunner to internals.
(To be exact, this CL only copies some functions, but I'll remove those ones
in testRunner later.)

BUG=331301

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=168495

Review URL: https://codereview.chromium.org/176953009

git-svn-id: svn://svn.chromium.org/blink/trunk@168588 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL also removes all calls to |localAttributeToPropertyMap()| and |registerAnimatedPropertiesFor##OwnerType()|, the methods which were defined via the macros. This map was already empty, so no change in behavior.
|SVGElement::cleanupAnimatedProperties()| contents are moved back to |~SVGElement()| again, as they were temporary workaround for the old SVGProperty implementation bug.

BUG=308818
NOTRY=true

Review URL: https://codereview.chromium.org/177803003

git-svn-id: svn://svn.chromium.org/blink/trunk@168587 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The members of RTCIceCandidate are no longer read-only.

BUG=349437

Review URL: https://codereview.chromium.org/187673002

git-svn-id: svn://svn.chromium.org/blink/trunk@168586 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This change makes shapes implementation inline with webkit's.
Also using references is lot safe for caller method and
also removes extra null checks.

Insprired by https://bugs.webkit.org/show_bug.cgi?id=124876

Review URL: https://codereview.chromium.org/178473024

git-svn-id: svn://svn.chromium.org/blink/trunk@168585 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=255363

Review URL: https://codereview.chromium.org/187823003

git-svn-id: svn://svn.chromium.org/blink/trunk@168584 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This makes StyleSheets RefCountedGarbageCollected. Next step is
removing RefPtrs and raw pointers and dealing with finalization
order issues.

R=erik.corry@gmail.com, haraken@chromium.org, wibling@chromium.org
BUG=

Review URL: https://codereview.chromium.org/187313005

git-svn-id: svn://svn.chromium.org/blink/trunk@168583 bbb929c8-8fbe-4397-9dbb-9b2b20218538

idl_definitions_builder.py was separate from idl_definitions.py
b/c the latter had lots of Perl compatibility cruft (JSON export etc.)
Now that it's gone, there's no reason to have these separate, as
idl_definitions was almost only class definitions, and
idl_definitions_builder was basically constructors.

Keeping them separate made it harder to read (b/c had to bounce back
and forth), and the calls to the actual constructors were incredibly
long.

This merges them, deleting one file and making the code quite simple.
Net -70 lines, though more like -100 due to adding a few comments
(class diagram).

Change is just merging the functions into constructors (in a few cases
changing an alternative constructor into a class method) and adding
a few comments.

Further cleanup is possible (remove IdlEnum and IdlUnionType),
but involve CG changes, so will do in followup.

BUG=345137
TBR=haraken

Review URL: https://codereview.chromium.org/177233006

git-svn-id: svn://svn.chromium.org/blink/trunk@168582 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL broke oilpan Debug builds with clang.

>
../../third_party/WebKit/Source/core/animation/css/CSSAnimatableValueFactory.cpp
> 1.      <eof> parser at end of file
> 2.      Per-file LLVM IR generation
> 3.      ../../third_party/WebKit/Source/heap/Heap.h:904:10: Generating code
for
> declaration
> 'WebCore::GarbageCollectedFinalized<WebCore::CSSCalcExpressionNode>::finalize'
> clang: error: unable to execute command: Aborted (core dumped)
> clang: error: clang frontend command failed due to signal (use -v to see
> invocation)



> Remove unneeded destructors and add finalization support to various classes.
> 
> Issues reported by the Blink GC clang plugin.
> 
> R=haraken@chromium.org
> BUG=
> 
> Review URL: https://codereview.chromium.org/184853006

TBR=zerny@chromium.org

Review URL: https://codereview.chromium.org/183853038

git-svn-id: svn://svn.chromium.org/blink/trunk@168581 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL might broke browser_tests in Mac:
http://build.chromium.org/p/chromium.webkit/builders/Mac10.8%20Tests/builds/6374

> id of iframe incorrectly sets window name
> 
> As per the specification, the iframe's name should be an empty string if unset:
> http://www.whatwg.org/specs/web-apps/current-work/multipage/the-iframe-element.html#attr-iframe-name
> 
> Instead, Chromium was using the iframe's id as window name if the name was
> unset. This change makes us behave like Firefox 27 and IE 11, verified using:
> http://jsfiddle.net/xf5H7/9/
> 
> This CL sets the name attribute iframe in a lot of layout tests so that
> testRunner.dumpChildFramesAsText() keeps printing the same result.
> 
> R=arv@chromium.org, tkent@chromium.org, abarth, arv, tkent
> BUG=347169
> TEST=fast/frames/iframe-no-name.html
> 
> Review URL: https://codereview.chromium.org/187103002

TBR=ch.dumez@samsung.com

Review URL: https://codereview.chromium.org/184633004

git-svn-id: svn://svn.chromium.org/blink/trunk@168579 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Subtitle support for native fullscreen video is enabled in https://codereview.chromium.org/22454003/.
However, there are a couple bugs:
1. FullscreenElementStack::currentFullScreenElementFrom() returns NULL if the 
fullscreen element is in a child frame, we need to use 
FullscreenElementStack::fullscreenElementFrom().

2. When the video element is appended to the top level frame, we should exclude 
it when rebuilding the compositing layer tree.

A new layout test is added: fullscreen/full-screen-iframe-allowed-video.html

This CL also fixes a bug that some of the fullscreen pixel tests are no longer 
running due to  https://codereview.chromium.org/24438004

BUG=349222,347843

Review URL: https://codereview.chromium.org/187193002

git-svn-id: svn://svn.chromium.org/blink/trunk@168578 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The isCSSPrimitiveValue check is likely covering up parser bugs,
and toCSSPrimitiveValue has the assert to ensure we're not
doing something bad.

R=eae,ksakamoto@chromium.org,timloh@chromium.org
BUG=

Review URL: https://codereview.chromium.org/184103034

git-svn-id: svn://svn.chromium.org/blink/trunk@168577 bbb929c8-8fbe-4397-9dbb-9b2b20218538

If a constructor attribute is annotated with either [MeasureAs=..] or
[DeprecateAs=..], then emit a needed callback wrapper that performs
the required reporting before looking up the constructor.

Use it to correctly handle use counters for the "webkitURL" properties
that both Window and WorkerGlobalScope currently provide; use counters
also added for these here.

(The above change also happens to fix use counter reporting
for the window.WebKitShadowRoot constructor.)

R=nbarth@chromium.org,haraken@chromium.org
BUG=348985

Review URL: https://codereview.chromium.org/186673002

git-svn-id: svn://svn.chromium.org/blink/trunk@168576 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This change moves effect and timing conversion logic out of Animation
and into EffectInput and TimingInput respectively.

This refactor includes a minor change in behaviour:
Animations created on elements without an active document renderer will no
longer return null, instead the animation's effect will be null.
This behaviour still avoids calls to the CSS parser and style resolver if
the check fails.

Review URL: https://codereview.chromium.org/182063005

git-svn-id: svn://svn.chromium.org/blink/trunk@168575 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Some change used PassRefPtrWillBeRawPtr for AnimatableColor which is not yet on the heap.
Thus build is broken in oilpan bots. This CL fixes it.

TBR=kouhei

Review URL: https://codereview.chromium.org/183183010

git-svn-id: svn://svn.chromium.org/blink/trunk@168574 bbb929c8-8fbe-4397-9dbb-9b2b20218538

 determine shrink-to-fit behavior

In order to determine whether the RemainingSpace is 
undetermined or not (shrink-to-fit) the current algorithm 
is checking out the style for clarifying if the Render object 
is float or absolute positioned.

It's better to use the Render functions instead, since the 
style might be applied or not, depending on several factors, 
like CSS properties inheritance.

BUG=79180, 234204

Review URL: https://codereview.chromium.org/181333003

git-svn-id: svn://svn.chromium.org/blink/trunk@168573 bbb929c8-8fbe-4397-9dbb-9b2b20218538

MediaControlPanelMuteButtonElement/MediaControlPanelVolumeSliderElement
were the only classes that inherited
MediaControlMuteButtonElement/MediaControlVolumeSliderElement.

Remove the m_panelMuteButton->show() in MediaControls::reset(), because
there is nothing that ever hides the button in the first place.

BUG=341813

Review URL: https://codereview.chromium.org/185363011

git-svn-id: svn://svn.chromium.org/blink/trunk@168572 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The noshade attribute is a boolean attribute and when set on hr element
it shows a gray color. When there is color attribute the default gray
color should be replaced by the color mentioned by the color attribute.
Firefox and IE show the same behaviour but Blink is different. Making
the behaviour of Blink similiar to Firefox and IE's behaviour.

When the color attribute is present that value is applied and the
default gray color is ignored. Incase of no color attribute the default
gray color is applied.

I already landed this patch on Webkit. Just modified the layout test file.
http://trac.webkit.org/changeset/161334

R=esprehn@chromium.org,eseidel@chromium.org
BUG=175359

Review URL: https://codereview.chromium.org/162993002

git-svn-id: svn://svn.chromium.org/blink/trunk@168571 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=348581

Review URL: https://codereview.chromium.org/185593011

git-svn-id: svn://svn.chromium.org/blink/trunk@168570 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=

Review URL: https://codereview.chromium.org/186763002

git-svn-id: svn://svn.chromium.org/blink/trunk@168569 bbb929c8-8fbe-4397-9dbb-9b2b20218538