GitLab

The current implementation sets default values in
V8T::toNative(). However, there are some code paths in which
we don't call V8T::toNative(). For example, if an IDL dictionary
is an optional argument and it isn't given, the binding layer
creates impl class by calling T::create() and passes it to
blink without calling V8T::toNative(). Since all dictionary impl
class instances should have default values, we should sets
default values in impl constructor.

BUG=321462

Review URL: https://codereview.chromium.org/479563003

git-svn-id: svn://svn.chromium.org/blink/trunk@180384 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=

Review URL: https://codereview.chromium.org/469213003

git-svn-id: svn://svn.chromium.org/blink/trunk@180383 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch adds code for removing spelling markers under words on 
blink's side.

BUG=3506

Review URL: https://codereview.chromium.org/419563003

git-svn-id: svn://svn.chromium.org/blink/trunk@180382 bbb929c8-8fbe-4397-9dbb-9b2b20218538

R=yurys, pfeldman

Review URL: https://codereview.chromium.org/476153002

git-svn-id: svn://svn.chromium.org/blink/trunk@180381 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Do not fire a DOMSubtreeModified event when Attr.value attribute is set. This
event is deprecated and both Firefox 31 and IE 11 do not fire it in this case.

Firing a DOMSubtreeModified event is problematic because it is synchronous and
it can lead to calling Attr::setValueInternal() recursively if the JS callback
sets Attr.value again. This can lead to serious problems as setValueInternal()
calls:
1. Element::willModifyAttribute(oldValue, newValue)
2. Attr::setValue(newValue);
3. Element::didModifyAttribute(newValue)

Due to recursivity, we can end up with the following call stack:
1. Element::willModifyAttribute("old-id", "new-id")
2. Attr::setValue("new-id");
3. Element::willModifyAttribute("new-id", "id-from-callback")
4. Attr::setValue("id-from-callback");
5. Element::didModifyAttribute("id-from-callback")
6. Element::didModifyAttribute("new-id")

After this, the Element's id is "new-id" because the id attribute is updated
in Element::didModifyAttribute(). However, the id in the DocumentOrderedMap
is "id-from-callback" because the id DocumentOrderedMap is updated in
Element::willModifyAttribute(). This mismatch between the DocumentOrderedMap
and the actual Element id can cause getElementById() to return incorrect
result. Even worse, it can result in a use-after-free (as in Bug 402255)
because the DocumentOrderedMap does not hold a reference to the StringImpl
used as key, assuming the StringImpl will be kept alive as it is used as 'id'
for an Element.

BUG=402255
TEST=fast/dom/Attr/set-attr-value-no-DOMSubtreeModified.html

Review URL: https://codereview.chromium.org/452093003

git-svn-id: svn://svn.chromium.org/blink/trunk@180380 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch simplifies our handling of watched selectors, allowing us to
remove the remaining pieces of user stylesheets. Currently a special
-internal-callback property is added via a user stylesheet, which depends
on the StyleBuilder applying all values and not just the cascaded values
to function correctly. The approach in this patch is to have this logic
separate to applyMatchedProperties, which means we won't need to have a
separate property and we can still use the matched properties cache in
these cases. This also allows us to remove the Vector of StyleRule in
MatchResult.

BUG=316960

Review URL: https://codereview.chromium.org/418163003

git-svn-id: svn://svn.chromium.org/blink/trunk@180379 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/475233003

git-svn-id: svn://svn.chromium.org/blink/trunk@180378 bbb929c8-8fbe-4397-9dbb-9b2b20218538

These don't do anything, since all it does is update the fill color, which is
not even used in drawHighlightForText (it takes its own color argument).

After this, nearly all updateGraphicsContext calls correspond to a
paintTextWithShadows call, which opens some refactoring options.

Review URL: https://codereview.chromium.org/477143002

git-svn-id: svn://svn.chromium.org/blink/trunk@180376 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The expectation lines have been moved to test_expectations_w3c.txt, so
they are safe to delete from TestExpectations.

This is patch 3 of 6 in the roll.

TBR=phoglund@chromium.org
BUG=404022

Review URL: https://codereview.chromium.org/477043002

git-svn-id: svn://svn.chromium.org/blink/trunk@180375 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Drop unnecessary RadioNodeList::m_onlyMatchImgElements member and replace it by
an inline function that does a CollectionType comparison.

The generated assembly for RadioNodeList.o is about the same size (1 line
shorter).

R=adamk@chromium.org

Review URL: https://codereview.chromium.org/480473002

git-svn-id: svn://svn.chromium.org/blink/trunk@180374 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/469203002

git-svn-id: svn://svn.chromium.org/blink/trunk@180373 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Rename CollectionIndexCache API to not use 'Element':
- traverseToFirstElement() -> traverseToFirst()
- traverseToLastElement() -> traverseToLast()

Using Element is confusing because some CollectionIndexCache users like
ChildNodeList deal with pure Nodes. The Node type is a template parameter
of the CollectionIndexCache class so the API should be seem to be Element
specific.

R=rob.buis@samsung.com

Review URL: https://codereview.chromium.org/469913002

git-svn-id: svn://svn.chromium.org/blink/trunk@180371 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/479563002

git-svn-id: svn://svn.chromium.org/blink/trunk@180370 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Use tighter typing in RadioNodeList::elementMatches() to give C++11 compilers
a chance to devirtualize the call to HTMLImageElement::formOwner().

R=adamk@chromium.org

Review URL: https://codereview.chromium.org/475613003

git-svn-id: svn://svn.chromium.org/blink/trunk@180369 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Use an AtomicString as key in DocumentOrderedMap instead of a StringImpl*.
This is safer as the HashMap Traits rely on the StringImpl* to not be a
dangling pointer (it is calling StringImpl::hash()). Using an AtomicString
as key will make sure we don't experience use-after-free such as in Bug
402255.

I ran Bindings/get-element-by-id.html test and did not see any performance
regression.

R=esprehn@chromium.org, morrita@chromium.org
BUG=402255

Review URL: https://codereview.chromium.org/467973003

git-svn-id: svn://svn.chromium.org/blink/trunk@180365 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The SVG instance tree was recently removed which also removed most of the gnarly event handling in SVG. Now we can support SVG also in event.path.

BUG=402721

Review URL: https://codereview.chromium.org/477883003

git-svn-id: svn://svn.chromium.org/blink/trunk@180364 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Introduce HasHTMLTagName functor and use it with ElementTraversal to simplify
the code a bit.

R=adamk@chromium.org

Review URL: https://codereview.chromium.org/471173003

git-svn-id: svn://svn.chromium.org/blink/trunk@180363 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Now documentation context menu item appears without token selection.

BUG=391593

Review URL: https://codereview.chromium.org/471583003

git-svn-id: svn://svn.chromium.org/blink/trunk@180362 bbb929c8-8fbe-4397-9dbb-9b2b20218538

When using a "pixel-moving" filter (ie., drop-shadow or blur), we
disable the normal clip operation, since we want to want to render
the full results of the element pre-clip for filtering. Clipping is
then applied just before drawing the filtered result.

However, we shouldn't disable *all* clipping recursively, since we
want children which clip themselves to still continue to be clipped.
The fix is to pass the original (unmodified) paintingInfo to
paintChildren().

BUG=350411

Review URL: https://codereview.chromium.org/476953003

git-svn-id: svn://svn.chromium.org/blink/trunk@180361 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Before this CL, we use bool flags to indicate which value should be
returned. We can use Nullable<T> instead.

Review URL: https://codereview.chromium.org/466323002

git-svn-id: svn://svn.chromium.org/blink/trunk@180360 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch moves insert style rule button further to the left to not overlap with scrollbar. It also fixes its width.

As a drive-by, this patch sets "translateZ" rule hack on styles sidebar pane as a workaround for compositor bug. This makes scrollbar appear on MB Air.

BUG=404102
R=pfeldman

Review URL: https://codereview.chromium.org/477113002

git-svn-id: svn://svn.chromium.org/blink/trunk@180359 bbb929c8-8fbe-4397-9dbb-9b2b20218538

R=ager@chromium.org
BUG=

Review URL: https://codereview.chromium.org/477913002

git-svn-id: svn://svn.chromium.org/blink/trunk@180358 bbb929c8-8fbe-4397-9dbb-9b2b20218538

With the shutdown crashes involving InjectedScriptManager::CallbackData
having been addressed, remove some crash test expectations.

R=ager@chromium.org
BUG=398472
NOTRY=true

Review URL: https://codereview.chromium.org/478643002

git-svn-id: svn://svn.chromium.org/blink/trunk@180357 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The blob: scheme is not handled by ParsedURL, so it is
impossible to compute the human-readable worker name
for them using the common scheme.

R=pfeldman, vsevik
BUG=404098
NOTRY=true

Review URL: https://codereview.chromium.org/471413002

git-svn-id: svn://svn.chromium.org/blink/trunk@180356 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Currently, importScript function gets scripts to load which contain "../"
in their path. These are written in their sourceURL's, but they should
be actually normalized to match network resource.

R=vsevik, apavlov, pfeldman

Review URL: https://codereview.chromium.org/475073002

git-svn-id: svn://svn.chromium.org/blink/trunk@180355 bbb929c8-8fbe-4397-9dbb-9b2b20218538

ServiceWorkerContainer.unregister() is going away but the change will be
done in steps. This patch updates some select tests focused on
unregistration to use the new unregister() function.

BUG=404064

Review URL: https://codereview.chromium.org/474193002

git-svn-id: svn://svn.chromium.org/blink/trunk@180354 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=dgozman

Review URL: https://codereview.chromium.org/472273002

git-svn-id: svn://svn.chromium.org/blink/trunk@180353 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Instead of listening to particular model notifications it is enough to listen for profiling lock changes.

BUG=403684
R=pfeldman@chromium.org

Review URL: https://codereview.chromium.org/471013002

git-svn-id: svn://svn.chromium.org/blink/trunk@180352 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We should pause any actions that could affect performance if any of the targets is being profiled, not only current target.

Compared to r180336 profiling lock is now initialized in Locks.js

BUG=403684
R=pfeldman@chromium.org

Review URL: https://codereview.chromium.org/472263002

git-svn-id: svn://svn.chromium.org/blink/trunk@180351 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Timeline extension API is removed from DevTools front-end.

BUG=403684
R=pfeldman@chromium.org

Review URL: https://codereview.chromium.org/477103002

git-svn-id: svn://svn.chromium.org/blink/trunk@180350 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=403684,398785
R=pfeldman@chromium.org

Review URL: https://codereview.chromium.org/476773002

git-svn-id: svn://svn.chromium.org/blink/trunk@180349 bbb929c8-8fbe-4397-9dbb-9b2b20218538

To achieve the specified goal, the patch does the following:
- StylePropertiesSection.styleRule no longer has sourceURL as it could
be always fetched from styleRule.rule
- StylePropertiesSection._createRuleOrigin is receiving CSSRule and
ruleLocation as arguments
- BlankStylePropertiesSection is taught to infer the actual rule
location of the rule that will be inserted

Layout Test change: the CSSOM rule now don't show source location (used to be "located" on first line)

R=pfeldman, vsevik, apavlov

Review URL: https://codereview.chromium.org/474433004

git-svn-id: svn://svn.chromium.org/blink/trunk@180348 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The parsing logic for clip-path and filter is identical to that for
marker-start, etc.

BUG=404023

Review URL: https://codereview.chromium.org/475303002

git-svn-id: svn://svn.chromium.org/blink/trunk@180346 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=404091
NOTRY=true

Review URL: https://codereview.chromium.org/468223005

git-svn-id: svn://svn.chromium.org/blink/trunk@180345 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch synchronizes diverged pseudoId names in Style Sidebar Pane
with RenderStyleConstants.h. Ideally, they should be generated, but so
far we don't have a solution for this.

BUG=380644
R=apavlov

Review URL: https://codereview.chromium.org/476803002

git-svn-id: svn://svn.chromium.org/blink/trunk@180344 bbb929c8-8fbe-4397-9dbb-9b2b20218538

The codemirror got frozen because selection from its
input text area was stolen via "add new property" editing functionality.

BUG=402838
R=apavlov, vsevik

Review URL: https://codereview.chromium.org/476893002

git-svn-id: svn://svn.chromium.org/blink/trunk@180343 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Revert of Make profiling lock global rather than per Target (patchset #2 of https://codereview.chromium.org/475803002/)

Reason for revert:
'common' should not be aware of profiler.

Original issue's description:
> Make profiling lock global rather than per Target
> 
> We should pause any actions that could affect performance if any of the targets is being profiled, not only current target.
> 
> BUG=403684
> R=sergeyv@chromium.org
> 
> Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=180336

TBR=sergeyv@chromium.org,apavlov@chromium.org,yurys@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=403684

Review URL: https://codereview.chromium.org/471393002

git-svn-id: svn://svn.chromium.org/blink/trunk@180342 bbb929c8-8fbe-4397-9dbb-9b2b20218538

We show a console message if:
 - An invalid string is set to 'value' IDL attribute.
 - An invalid string is set to 'value' content attribute.
 - 'date' is set to 'type' attribute and existing 'value' content attribute is
   invalid.
 - 'date' is set to 'type' attribute and existing user-specified value is
   invalid.

BUG=397952
TEST=Update expectations of automated tests.

Review URL: https://codereview.chromium.org/461323006

git-svn-id: svn://svn.chromium.org/blink/trunk@180341 bbb929c8-8fbe-4397-9dbb-9b2b20218538

These are crashing on layout tests in the oilpan build. The reason is
that with oilpan these are just raw pointers and are therefore not
default initialized to null.

R=haraken@chromium.org, oilpan-reviews@chromium.org, timloh@chromium.org

Review URL: https://codereview.chromium.org/473163002

git-svn-id: svn://svn.chromium.org/blink/trunk@180340 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Currently there is a bug on V8 APIs on Message objects, and they don't return valid values
when they are called at the first time if V8 has a pending exception.
We need to call twice in order to get valid values.

This is a temporary fix in the Blink side until we fix the V8 side.

BUG=341031

Review URL: https://codereview.chromium.org/478623002

git-svn-id: svn://svn.chromium.org/blink/trunk@180339 bbb929c8-8fbe-4397-9dbb-9b2b20218538