GitLab

This removes ~242 KB of *-expected.png and leaves only a few
platform-specific expectations left in
LayoutTests/platform/linux/virtual/android/fullscreen/

Some tests were simultaneously pixel tests and text tests, which is
not great for a reftest. To avoid having the (formated) log output in
*-expected.html, instead make those tests time out if any check fails.

Review URL: https://codereview.chromium.org/457663002

git-svn-id: svn://svn.chromium.org/blink/trunk@180036 bbb929c8-8fbe-4397-9dbb-9b2b20218538

MediaSource.isTypeSupported() should not return true without making sure
that the HTMLMediaElement supports the content type as well.

BUG=402386
R=acolwell@chromium.org

Review URL: https://codereview.chromium.org/456313003

git-svn-id: svn://svn.chromium.org/blink/trunk@180035 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch is part 2 of merging the CSS .in files. Having a single .in
file for listing CSS properties makes it both easier to add new properties
and add new property flags.

BUG=396992
TBR=pfeldman

Review URL: https://codereview.chromium.org/415613002

git-svn-id: svn://svn.chromium.org/blink/trunk@180034 bbb929c8-8fbe-4397-9dbb-9b2b20218538

In order to debug private scripts, JS errors thrown in the private scripts
(e.g., ReferenceError thrown when we make a typo in private scripts)
need to be reported to stderr. This CL makes the change.

The fprintf(stderr, ...) is enabled only in Debug builds.

This CL also renames throwDOMExceptionInPrivateScriptIfNeeded to rethrowExceptionInPrivateScript.

BUG=341031

Review URL: https://codereview.chromium.org/460573002

git-svn-id: svn://svn.chromium.org/blink/trunk@180033 bbb929c8-8fbe-4397-9dbb-9b2b20218538

AudioNode::isMarkedForDeletion is not needed any more in oilpan builds
for the following (non-trivial) reason:

0 isMarkedForDeletion is set to true in AudioNode::dispose().
In oilpan builds, AudioNode::dispose() is called after the AudioNode
becomes unreachable. In other words, isMarkedForDeletion is set to true
after the AudioNode becomes unreachable.

- The only user of isMarkedForDeletion is AudioNodeInput::canUpdateState().
AudioNodeInput::canUpdateState() returns false if isMarkedForDeletion is true.

- AudioNodeInput has a strong reference to the AudioNode.

These facts indicate that AudioNodeInput::canUpdateState() always returns true. Thus this CL removes the canUpdateState() checks from the code base. This also enables us to remove AudioNode::isMarkedForDeletion.

BUG=340522
NOTRY=true

Review URL: https://codereview.chromium.org/458923003

git-svn-id: svn://svn.chromium.org/blink/trunk@180032 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Revert of Fix leaks when the shared worker is requested to be terminated while loading the main script. (https://codereview.chromium.org/451603002/)

Reason for revert:
UAF bug in WebSharedWorkerImpl::stopWorkerThread().

Original issue's description:
> Fix leaks when the shared worker is requested to be terminated while loading the main script.
> 
> BUG=398742
> 
> Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=180013

TBR=jochen@chromium.org,sigbjornf@opera.com
NOTREECHECKS=true
NOTRY=true
BUG=398742

Review URL: https://codereview.chromium.org/465863002

git-svn-id: svn://svn.chromium.org/blink/trunk@180031 bbb929c8-8fbe-4397-9dbb-9b2b20218538

A reftest cannot be marked as NeedsRebaseline/NeedsManualRebaseline.
This caused lint errors on all bots.

> Mark test as needing rebaseline after Skia change improves it.
> 
> virtual/gpu/fast/canvas/canvas-createImageBitmap-colorClamping.html is less dirty after Skia change.
> 

TBR=bungeman@chromium.org,zerny@chromium.org

Review URL: https://codereview.chromium.org/463703004

git-svn-id: svn://svn.chromium.org/blink/trunk@180030 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL fixes flaky crashes in tables/mozilla/bugs/.

RenderImageResource::shutdown unregisters the owner RenderImage from an
ImageResource. We should call it in RenderImage::destroy because
RenderObject::willRenderImage, which is a callback of ImageResourceClient,
should not be called after RenderImage::destroy.

BUG=398342

Review URL: https://codereview.chromium.org/458383003

git-svn-id: svn://svn.chromium.org/blink/trunk@180029 bbb929c8-8fbe-4397-9dbb-9b2b20218538

r179771 fixed the crash bug but it didn't fix the leak bug.

BUG=400012
TEST=n/a
NOTRY=true

Review URL: https://codereview.chromium.org/460213002

git-svn-id: svn://svn.chromium.org/blink/trunk@180028 bbb929c8-8fbe-4397-9dbb-9b2b20218538

virtual/gpu/fast/canvas/canvas-createImageBitmap-colorClamping.html is less dirty after Skia change.


git-svn-id: svn://svn.chromium.org/blink/trunk@180027 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=400387
R=ojan@chromium.org,esprehn@chromium.org,leviw@chromium.org
NOTRY=true

Review URL: https://codereview.chromium.org/459463003

git-svn-id: svn://svn.chromium.org/blink/trunk@180026 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is a follow up CL to r178818, adding a comment why the order is important.

BUG=364377
NOTRY=true

Review URL: https://codereview.chromium.org/460053002

git-svn-id: svn://svn.chromium.org/blink/trunk@180024 bbb929c8-8fbe-4397-9dbb-9b2b20218538

PopupMenu::disconnectClient accesses some objects, and they might be destructed
when a RenderMenuList object is garbage collected. We should call
PopupMenu::disconnectClient earlier.

This fixes SelectPopupMenuStyleTest.PopupListBoxRTLRowWidth in Oilpan.

BUG=402439

Review URL: https://codereview.chromium.org/459313002

git-svn-id: svn://svn.chromium.org/blink/trunk@180023 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Have DocumentOrderedMap API deal with AtomicString type instead of StringImpl*
as it would be unsafe to use if the StringImpls were not atomized.

This basically moves the AtomicString::impl() calls from the DocumentOrderedMap
API call sites to the DocumentOrderedMap implementation. This simplifies the
call sites a bit and makes the DocumentOrderedMap code look safer.

R=morrita@chromium.org, adamk@chromium.org

Review URL: https://codereview.chromium.org/459203004

git-svn-id: svn://svn.chromium.org/blink/trunk@180022 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=kouhei@chromium.org,hajimehoshi@chromium.org
BUG=400588,402742,402743

Review URL: https://codereview.chromium.org/456863004

git-svn-id: svn://svn.chromium.org/blink/trunk@180021 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch is part 1 of merging the CSS .in files (part 2 is also up, at
https://codereview.chromium.org/415613002). Having a single .in file for
listing CSS properties makes it both easier to add new properties and
add new property flags.

I've renamed some of the StyleBuilder flags to make it more obvious what
these refer to and also re-ordered CSSProperties.in in some sort of
sensible fashion (see part 2 patch for what this will look like with the
remaining properties); the high/low priority split doesn't make too much
sense right now but will of course make sense once CSSPropertyNames.in
is merged into the same file.

BUG=396992
TBR=pfeldman

Review URL: https://codereview.chromium.org/410953002

git-svn-id: svn://svn.chromium.org/blink/trunk@180020 bbb929c8-8fbe-4397-9dbb-9b2b20218538

r179509 incorrectly converted the master map. It didn't
include the @ToT Blink group.

TBR since this is causing the flakiness dashboard to be
down for the blink bots.

TBR=szager@chromium.org

Review URL: https://codereview.chromium.org/462963003

git-svn-id: svn://svn.chromium.org/blink/trunk@180019 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=180004

BUG=377445
TBR=eseidel@chromium.org

Review URL: https://codereview.chromium.org/463763002

git-svn-id: svn://svn.chromium.org/blink/trunk@180018 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Including polymer.html in karma.conf.js makes platform.js throw
"Possible attempt to load Polymer twice" when a polymer component is imported.

I think this is because polymer.html is loaded twice: once by karma and again on
the first time that the browser encounters a <link href="polymer.html">. I believe
this might happen because of the way karma includes files by appending some kind
of hash to them.

If polymer.html is parsed twice, then the <script src="polymer.js"> also runs twice.

This CL adds karma-loader.html which imports polymer.html (so it should avoid
importing it a second time) and adds a listener for polymer-ready, which I didn't
think was necessary but seems to resolve some flakiness in karma.

BUG=400387
R=ojan@chromium.org,esprehn@chromium.org

Review URL: https://codereview.chromium.org/459223002

git-svn-id: svn://svn.chromium.org/blink/trunk@180017 bbb929c8-8fbe-4397-9dbb-9b2b20218538

If the file doesn't exist in the repo, we should return a 400
instead of throwing an error and returning a 500.

TBR since this is causing 500s on the live server.

TBR=szager@chromium.org

Review URL: https://codereview.chromium.org/459243004

git-svn-id: svn://svn.chromium.org/blink/trunk@180016 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Because <img> scale its contents on resize, incremental invalidation
doesn't apply to it. Force full invalidate if the image content is
resized during layout.

BUG=397939
TEST=fast/repaint/image-resize.html
TEST=fast/repaint/percent-size-image-resize-container.html

Review URL: https://codereview.chromium.org/458373003

git-svn-id: svn://svn.chromium.org/blink/trunk@180015 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=398742

Review URL: https://codereview.chromium.org/451603002

git-svn-id: svn://svn.chromium.org/blink/trunk@180013 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=179998

BUG=369942
TBR=pdr@chromium.org

Review URL: https://codereview.chromium.org/466573002

git-svn-id: svn://svn.chromium.org/blink/trunk@180012 bbb929c8-8fbe-4397-9dbb-9b2b20218538

testlistjson should go down the same codepath as other content
that serves up the actual file content so that it gets
the right master name.

TBR since this is causing 500s on the live server.

TBR=szager@chromium.org

Revert "Get rid of testlistjson query parameter."

This reverts r180005. Turns out that the android JSONResultsGenerator
really does depend on testlistjson not having the results and times. :(

TBR=szager@chromium.org

Review URL: https://codereview.chromium.org/467453002

git-svn-id: svn://svn.chromium.org/blink/trunk@180011 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This reverts r180005. Turns out that the android JSONResultsGenerator
really does depend on testlistjson not having the results and times. :(

TBR=szager@chromium.org

Review URL: https://codereview.chromium.org/459263002

git-svn-id: svn://svn.chromium.org/blink/trunk@180010 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The RenderLayerCompositor can determine whether it needs to rebuild the tree.
We don't need to dirty this much state.

R=esprehn@chromium.org

Review URL: https://codereview.chromium.org/450403003

git-svn-id: svn://svn.chromium.org/blink/trunk@180008 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This function cleans out a single dirty bit in the RenderLayer tree without
cleaning the entire subtree, which breaks the invariant that dirty bits
represent are connected up to the root of the tree. Fortunately, this function
isn't needed any longer.

This CL also fixed a bug whereby we'd issue invalidations for the previous
location of an element that became visible. That can manifest as a bug if the
element moves after becoming visible or if the old rect exists in another
graphics layer.

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

Review URL: https://codereview.chromium.org/456963002

git-svn-id: svn://svn.chromium.org/blink/trunk@180007 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The spec (http://dev.w3.org/csswg/mediaqueries-4/#hover)  for the hover
media query has added keywords for the feature's value. This CL updates
the media query keywords to propertly parse and evaluate the feature;
however, it remains only sparsely implemented. The hover (and pointer)
media queries currently only produce non-false output for touchscreen
devices. The plumbing work to actually produce the correct results in
all cases will be done in a future CL.

BUG=136119

Review URL: https://codereview.chromium.org/451113002

git-svn-id: svn://svn.chromium.org/blink/trunk@180006 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This is an optimization to make it so that we don't have to send
the whole results.json file down since the JSONResultsGenerator
needs this to figure out which tests it can strip. The only code
left that uses this is the Android fork of this code, which needs
to go away. So, I'm not OK with keeping this special-cased code
in the generic test results server just for that.

Deleting the code means that the requests will get the full
results.json file instead of the one with the results and times
stripped out.

TBR because this is causing 500s on the live server and that
push can't be rolled back without breaking other things.

TBR=szager@chromium.org

Review URL: https://codereview.chromium.org/463683002

git-svn-id: svn://svn.chromium.org/blink/trunk@180005 bbb929c8-8fbe-4397-9dbb-9b2b20218538

There was a regression bug filed (bug 377445) shortly
after the original commit, however the author chose
not to revert, but rather instead attempt a fix.

The later of those fixes was marked not lgtm' by another
reviewer.

I'm rolling out the whole mess to unblock shipping M37.

R=eae
BUG=377445

Review URL: https://codereview.chromium.org/459153002

git-svn-id: svn://svn.chromium.org/blink/trunk@180004 bbb929c8-8fbe-4397-9dbb-9b2b20218538

deprecated_master would be referenced before assignment when
uploaded with the chromium.linux style master name.

TBR since I had to push this already to fix the broken push
and rolling back wasn't an option.

TBR=szager@chromium.org
NOTRY=true

Review URL: https://codereview.chromium.org/462883002

git-svn-id: svn://svn.chromium.org/blink/trunk@180003 bbb929c8-8fbe-4397-9dbb-9b2b20218538

When using overtype mode in content editable content we usually replace the
caret by a block cursor that covers the next character to be replaced.
However at the end of lines we should show the 'normal' blinking caret. In
those cases we were incorrectly adding a selection block till the end of the
container of the editable content.

That's because we were using a complicated (and wrong) way to detect the end
of lines based on modifying the internal selection used by the overtype
mode. We now use the proper tools to detect the end of line detection which
also supports bidi text.

BUG=385003

Review URL: https://codereview.chromium.org/437843002

git-svn-id: svn://svn.chromium.org/blink/trunk@180002 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Code to set the flag was removed as part of debugging crbug.com/378444
but this causes ports to report as "in use" for some time after the
test server stops, meaning that re-running layout tests with only a
short delay may fail.

Restore setting the flag, but scoped to non-Windows systems.

BUG=378444
R=dpranke@chromium.org

Review URL: https://codereview.chromium.org/464543003

git-svn-id: svn://svn.chromium.org/blink/trunk@180001 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/459083002

git-svn-id: svn://svn.chromium.org/blink/trunk@180000 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We did not have any test coverage on :host-context and targeted style
recalc. Added a few tests.

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

Review URL: https://codereview.chromium.org/458943002

git-svn-id: svn://svn.chromium.org/blink/trunk@179999 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Reason for revert:
Caused 10-50% performance regressions on svg, see http://crbug.com/402133.

Original issue's description:
> [SVG2] Make transform, gradientTransform and patternTransform presentation attributes.
>
> Spec:
> http://www.w3.org/TR/css3-transforms/#svg-transform
> http://www.w3.org/TR/css3-transforms/#svg-syntax
> http://www.w3.org/TR/css3-transforms/#svg-gradient-transform-pattern-transform
>
> BUG=369942
>
> Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=179702

TBR=dschulze@chromium.org,ed@opera.com
NOTREECHECKS=true
BUG=369942

Review URL: https://codereview.chromium.org/459053002

git-svn-id: svn://svn.chromium.org/blink/trunk@179998 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Move remaining ui.js functionality into existing classes and elements.

BUG=400393
NOTRY=true

Review URL: https://codereview.chromium.org/459983002

git-svn-id: svn://svn.chromium.org/blink/trunk@179993 bbb929c8-8fbe-4397-9dbb-9b2b20218538

In PageAnimator, it's possible frame tree has been mutated by
Javascript. To avoid this, we iterator through documents collection
instead, and check whether they are still present in their frames.

This was part of CL 214953004, which got reverted in 335103004 when it
caused crbug.com/383946. So I am adding back the reverted CL in parts.
This part should be only related to crbug.com/359028.

BUG=359028

Review URL: https://codereview.chromium.org/463493004

git-svn-id: svn://svn.chromium.org/blink/trunk@179992 bbb929c8-8fbe-4397-9dbb-9b2b20218538

When we stopped storing the mediaType as a string on m_medium on StyleResolver, 
we started calling FrameView::mediaType() every time that the mediaType as required.
In the related bug's test case, mediaType() was being called a lot, 
in order to compare it to "print" to see which default style sheet should be used.
Since mediaType() calls also call InspectorInstrumentation::applyEmulatedMedia(), 
they have a certain overhead, that impacted the benchmark's results.

I've changed it so that the mediaType() call and the comparison to "print"
would be done once, and the result cached.

BUG=394275

Review URL: https://codereview.chromium.org/405953002

git-svn-id: svn://svn.chromium.org/blink/trunk@179991 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL makes RenderLayer::hasVisibleContent ASSERT that the state it reads
isn't dirty. This CL also fixes a couple callers that were reading the state
while it was dirty.

R=skobes@chromium.org

Review URL: https://codereview.chromium.org/453113002

git-svn-id: svn://svn.chromium.org/blink/trunk@179990 bbb929c8-8fbe-4397-9dbb-9b2b20218538