GitLab

http://src.chromium.org/viewvc/blink?view=revision&revision=181499

BUG=409708
TBR=ccameron@chromium.org

Review URL: https://codereview.chromium.org/551593002

git-svn-id: svn://svn.chromium.org/blink/trunk@181506 bbb929c8-8fbe-4397-9dbb-9b2b20218538

These forward references are not needed anymore.

Review URL: https://codereview.chromium.org/541383003

git-svn-id: svn://svn.chromium.org/blink/trunk@181505 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Import these polygon shapes reftests and fix them.
Note that these tests should be imported at some point.

This is based on http://trac.webkit.org/changeset/172973 by Zoltan Horvath <zoltan@webkit.org>.

From the WebKit patch commit log:
We didn't respect the shape-margin before a positioned polygon, since we checked
the shape-margin for only 1 vertex for a polygon edge. This patch fixes the behavior,
removes the tests from the skipped list and updates 1 test, which was incorrect.

Review URL: https://codereview.chromium.org/510513002

git-svn-id: svn://svn.chromium.org/blink/trunk@181504 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This depends on HTMLDocumentParser fix: https://codereview.chromium.org/532343002/

XMLHttpRequest and HTMLImport need callback when document has finished
parsing. This CL introduces an interface DocumentParserClient for receiving the
"parser stopped" notification, and migrates HTMLImportLoader to use the
interface.

BUG=409461

Review URL: https://codereview.chromium.org/535403002

git-svn-id: svn://svn.chromium.org/blink/trunk@181503 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Add clipping.
Add transform.
Change recording to be driven by GraphicsContext.
Update dependent code.
Move a method used only by SVGFEImage from AffineTransform to SVGFEImage

R=pdr,senorblanco
BUG=410019

Review URL: https://codereview.chromium.org/536573002

git-svn-id: svn://svn.chromium.org/blink/trunk@181502 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=181494

BUG=410541
TBR=chrishtr@chromium.org

Review URL: https://codereview.chromium.org/553613002

git-svn-id: svn://svn.chromium.org/blink/trunk@181501 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Instead of needing to update both loops any time we change anything we
should just put the Element specific things in Element::recalcStyle and
then put the shared method into ContainerNode. I also removed an old
comment about ShadowRoot::recalcStyle not supporting adjacent selectors
properly (I fixed that bug a long time ago).

There should be no change in behavior, I just moved the code around.

BUG=399816

Review URL: https://codereview.chromium.org/548693002

git-svn-id: svn://svn.chromium.org/blink/trunk@181500 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The function ScrollbarThemeMacNonOverlayAPI::paint calls into
HIThemeDrawTrack to draw the track alone, then calls into
WebThemeEngineImpl::paintScrollbarThumb (a la Mac), to make
the exact same HIThemeDrawTrack call, just with a thumb added.
Coalesce these into a single call to HIThemeDrawTrack, which
makes the thumb actually appear.

The reason that the thumb wasn't appearing in the first place is
that the nested gfx::SkiaBitLocker instances weren't playing nice
with each other. In non-impl-side painting, the gfx::SkiaBitLocker
instances would draw to the underlying canvas. With impl-side
painting, they were drawn offscreen, and this evidently doesn't
work (tested by passing the CGContextRef directly to the
WebThemeEngineImpl::paintScrollbarThumb call rather than
creating a separate gfx::SkiaBitLocker, not root-caused).

So, the solution is a bit roundabout in the sense that it just
side-steps the gfx::SkiaBitLocker problem, but this is better
than fixing that problem and continuing to do this in the
double-identical-draw-for-no-reason manner.

BUG=409708

Review URL: https://codereview.chromium.org/544913002

git-svn-id: svn://svn.chromium.org/blink/trunk@181499 bbb929c8-8fbe-4397-9dbb-9b2b20218538

ToT currently doesn't pass lint-test-expectations; address.

TBR=eae
BUG=
NOTRY=true

Review URL: https://codereview.chromium.org/547023003

git-svn-id: svn://svn.chromium.org/blink/trunk@181498 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This avoids crashing in call sites that happen during document setup and shutdown
of the document. An example of the former appears to be in the attached bug.

An example of the latter is easily reproducible: load a document with hyperlinks
in it, select some text via the text finding widget (ctrl-f), then click on the 
hyperlink. The code to close the find box will call into Blink and check
if the elements are focusable. However, the elements no longer have renderers
since the document has been deactivated to prepare for navigation.

TBR=abarth@chromium.org
BUG=251163

Review URL: https://codereview.chromium.org/533963002

git-svn-id: svn://svn.chromium.org/blink/trunk@181497 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=409727

Review URL: https://codereview.chromium.org/547493002

git-svn-id: svn://svn.chromium.org/blink/trunk@181496 bbb929c8-8fbe-4397-9dbb-9b2b20218538

overflow clip.

Previously it checked for clip but not overflow clip.

BUG=409680

Review URL: https://codereview.chromium.org/543123003

git-svn-id: svn://svn.chromium.org/blink/trunk@181495 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Previously, there was some special logic in RenderText::clippedOverflowRectForPaintInvalidation
to detect a special scenario when the paint invalidation container of the
RenderText was a child of the containing block/layer of the RenderText.
This had two flaws: one, it was checking for a condition different than that, and
two the descendant check is quite expensive if the paint invalidation container
is far away in the render tree from the RenderText.

The intent of the code is to delegate invalidation to the containing block/layer.
The new implementation does that directly, by overriding containerForPaintInvalidation as well as
clippedOverflowRectForPaintInvalidation
to delegate directly.

This also fixes a bug in which RenderText objects invalidate much too widely
if the RenderText is being reparented accross a frame boundary. This started
happening because now we let paint invalidation containers cross frame boundaries.

BUG=410541

Review URL: https://codereview.chromium.org/544743003

git-svn-id: svn://svn.chromium.org/blink/trunk@181494 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Add support for font family names with a font-weight suffix (such as
Light, Semibold, Black  etc). If the full name fails to match an 
installed font check for a font-weight suffix and if so strip it and set
the weight according to the prefix.

The current implementation matches IE when it comes to picking or
overriding the weight based on the suffix.

TEST=fast/text/font-weight-variant.html
BUG=335323, 410333, 410865, 408260
R=cpu@chromium.org, dglazkov@chromium.org

Review URL: https://codereview.chromium.org/542603004

git-svn-id: svn://svn.chromium.org/blink/trunk@181493 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Conditionally disable subpixel text for certain older fonts at sizes
below 16px in order to improve legibility and rendering quality. This is
limited to a specific set of fonts and does not apply to webfonts.
This list of fonts in question includes all fonts in the "Microsoft Core
fonts for the Web" collection.

Specifically the following fonts: arial, courier new, times new roman,
andale mono, comic sans, impact, georgia, trebuchet, verdana, webdings,
tahoma, and lucida console.

R=cpu@chromium.org, scottmg@chromium.org
BUG=405445,395425

Review URL: https://codereview.chromium.org/544843004

git-svn-id: svn://svn.chromium.org/blink/trunk@181492 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Fix to keep the selection of the text field in input element after setSelectionRange is called by dom events related to mouse

The problem is that the selection is cleared when handling mouse release.
The selection should be kept if the selection is set by setSelectionRange during handling mouse event.

BUG=32865
R=yosin@chromium.org
R=rbyers@chromium.org
TEST=LayoutTests/fast/forms/setrangetext-within-events.html

Review URL: https://codereview.chromium.org/507533002

git-svn-id: svn://svn.chromium.org/blink/trunk@181490 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=181485

TBR=jchaffraix@chromium.org

Review URL: https://codereview.chromium.org/547873002

git-svn-id: svn://svn.chromium.org/blink/trunk@181489 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=181484

TBR=vollick@chromium.org

Review URL: https://codereview.chromium.org/547053003

git-svn-id: svn://svn.chromium.org/blink/trunk@181488 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The current spec says Selection.collapse(null) raises exception and the 
implementation follows.
However, many web apps assume that chrome clears selection when 
Selection.collapse(null) calls.
Thus we revert the behavior.

BUG=395318

Review URL: https://codereview.chromium.org/546703003

git-svn-id: svn://svn.chromium.org/blink/trunk@181487 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This reverts commit 9cd85241c98a725b9fdd2d9923137a77cf4aeb8a.

The revert did not fix the mac pixel tests

BUG=
TBR=keishi, jchaffraix@chromium.org

Review URL: https://codereview.chromium.org/544203004

git-svn-id: svn://svn.chromium.org/blink/trunk@181486 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The setScrollOffset code was either generating a synchronous
invalidation or using repaint-after-compositing-update. This
change removes the former code path.

While the overall test invalidation area is the same in all
tests, we generate more invalidations. The extra invalidations
are caused by forcing a subtree walk of the shifted scrollable
areas and thus forcing an invalidation for some of the descendants.

Review URL: https://codereview.chromium.org/541273002

git-svn-id: svn://svn.chromium.org/blink/trunk@181485 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This currently causes correctness and memory issues due to the way the
compositor handles masking out the corners. Let's punt on this for now.

BUG=409984,365828

Review URL: https://codereview.chromium.org/544183002

git-svn-id: svn://svn.chromium.org/blink/trunk@181484 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This reverts commit 11871f14944c99fa718abfc12226ddc8d1da71a8.

Revert "Auto-rebaseline for r181475"

This reverts commit ccabe1281427cc3be9c2708bbdbfaba8ab7bd087.

These CLs mayb have broken a LOT of Mac pixel tests. I will watch the
bots...

TBR=jchaffraix@chromium.org, keishi

Review URL: https://codereview.chromium.org/543203004

git-svn-id: svn://svn.chromium.org/blink/trunk@181483 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=haraken,oilpan-reviews
BUG=404622
NOTRY=true

Review URL: https://codereview.chromium.org/547743002

git-svn-id: svn://svn.chromium.org/blink/trunk@181482 bbb929c8-8fbe-4397-9dbb-9b2b20218538

http://src.chromium.org/viewvc/blink?view=revision&revision=181475

TBR=jchaffraix@chromium.org

Review URL: https://codereview.chromium.org/544233003

git-svn-id: svn://svn.chromium.org/blink/trunk@181481 bbb929c8-8fbe-4397-9dbb-9b2b20218538

RenderStyle::visualInvalidationDiff() is already returning
'more than enough' paint invalidation flag.
The setShouldDoFullPaintInvalidation() call in
RenderLayerModelObject::styleDidChange() is just redundant.

BUG=349061
TEST=all existing tests pass

Review URL: https://codereview.chromium.org/546513002

git-svn-id: svn://svn.chromium.org/blink/trunk@181480 bbb929c8-8fbe-4397-9dbb-9b2b20218538

MQs in one section now overlay. Hovering will show the values on the ruler.

BUG=409870

Review URL: https://codereview.chromium.org/526423002

git-svn-id: svn://svn.chromium.org/blink/trunk@181479 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Failures that have only happened once, and only on one bot should be classified
differently than other failures-- it's quite possible that they're flakes.

NOTRY=true
BUG=405241

Review URL: https://codereview.chromium.org/546433003

git-svn-id: svn://svn.chromium.org/blink/trunk@181478 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Scrolling used to generate immediate invalidations that are now
postponed to the paint invalidation phase. In order to make this
works, scrolling updates needs to to happen before paint
invalidations.

The test changes are either neutral (same amount of invalidations,
except that we generate them one at a time) or
compositing/repaint/fixed-pos-inside-composited-intermediate-layer
is a progression.

Review URL: https://codereview.chromium.org/498773007

git-svn-id: svn://svn.chromium.org/blink/trunk@181475 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=None
TBR=keishi, dpranke

Review URL: https://codereview.chromium.org/545983002

git-svn-id: svn://svn.chromium.org/blink/trunk@181474 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Also do a minor refactoring of the decision logic, which hopefully
makes it a bit easier to understand.

This is a fix for the performance regression introduced by
crrev.com/432273004, which inadvertently disabled the parser cache
when compiling without any compile options.

R=marja@chromium.org
BUG=404622
BUG=399580

Review URL: https://codereview.chromium.org/528013002

git-svn-id: svn://svn.chromium.org/blink/trunk@181472 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This makes it a little clearer than when we have 12-argument functions.

BUG=388946

Review URL: https://codereview.chromium.org/538663002

git-svn-id: svn://svn.chromium.org/blink/trunk@181471 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Now that we are using atomics to read/write the mark bit we cannot rely on NO_SANITIZE_ADDRESS
since it doesn't propagate to called methods (in this case acquireLoad/releaseStore). We don't
want to add NO_SANITIZE_ADDRESS to acquireLoad/releaseStore so instead we explicitly
unpoison/poison the header.

R=ager@chromium.org, haraken@chromium.org, oilpan-reviews@chromium.org, zerny@chromium.org
BUG=

Review URL: https://codereview.chromium.org/544043003

git-svn-id: svn://svn.chromium.org/blink/trunk@181470 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Drive-by fixes:
1) Move DOMSyntaxHighlighter to ui/
2) Remove "CSS" from experiment name
3) Fix documentation for RegExp methods

BUG=391593

Review URL: https://codereview.chromium.org/530003002

git-svn-id: svn://svn.chromium.org/blink/trunk@181468 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=391593

Review URL: https://codereview.chromium.org/547613002

git-svn-id: svn://svn.chromium.org/blink/trunk@181467 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=409857
R=pfeldman@chromium.org, sergeyv@chromium.org, dgozman@chromium.org

Review URL: https://codereview.chromium.org/535143003

git-svn-id: svn://svn.chromium.org/blink/trunk@181466 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This change...
- removes unnecessary 'wait_for_update' calls because a register promise
is resolved with a registration object that has an installing worker.
- removes 't.step_func' from promise chains.
- unloads iframes gracefully.


BUG=401381
TEST=run_webkit_tests.py --debug http/tests/serviceworker/fetch-event.html

Review URL: https://codereview.chromium.org/531423002

git-svn-id: svn://svn.chromium.org/blink/trunk@181465 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Instead introduce VisibleSelection::intersectsNode(Node*) and static
version of Range::intersectsNode.

R=yosin@chromium.org
BUG=388681

Review URL: https://codereview.chromium.org/541633004

git-svn-id: svn://svn.chromium.org/blink/trunk@181464 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=

Review URL: https://codereview.chromium.org/545123002

git-svn-id: svn://svn.chromium.org/blink/trunk@181463 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Rows have role="row", and cells have role="girdcell".

We apply aria-activedescendant to the calendar because the calendar cell is not
focusable. Thus, each of cells needs to have id=.  Also, cells should have
descriptive text as aria-label.

- getLocale() needs to replace _ with - because Intl.DateTimeFormat doesn't
  accept _.

- This CL doesn't improve the calendar picker for type=month and type=week, and
  won't be harmful for them.

BUG=123896

Review URL: https://codereview.chromium.org/536903002

git-svn-id: svn://svn.chromium.org/blink/trunk@181462 bbb929c8-8fbe-4397-9dbb-9b2b20218538