GitLab

I was very confused by the ScriptLoader code which called
CSP::allowScriptNonce and the corresponding variable
which was called isValidScriptNonce.  I searched around
for nonce functions and found that they only checked
the nonce (not all policies) and I originally thought
this was an error!  However after further investigation
I found that allowScriptNonce actually checks all active
CSPs and returns true only if they all pass, thus making
the following line (which disables further CSP checks
on the load) OK.

I renamed these allowScriptNonce and allowStyleNonce
functions to have a 'With' in their name and added
a comment in the header explaining their behavior.

While making this change, I also reduced the verbosity
of several repeated uses of the string ContentSecurityPolicy.
I think most blink hackers would be able to look up that CSP
refers to ContentSecurityPolicy.

I removed several redundant calls to Document::contentSecurityPolicy
(again just caused the lines to be needlessly verbose).

Finally I made ScriptLoader::executeScript use the preferred
early-return pattern instead of a long indented block
for the !frame case.  I suspect that the !frame check
can actually be moved much earlier in the function
or even turned into an ASSERT.

This is still my long yack-shave to actually make the
preloader correct so I can fix the bugs which my
patch to make the HTML parser yield more agressively
(and thus not starve the preloader) possible to land:
https://codereview.chromium.org/258013009/

BUG=356292

Review URL: https://codereview.chromium.org/353873003

git-svn-id: svn://svn.chromium.org/blink/trunk@177065 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Currently CSSValueList::item does a bounds check and will return null if
the given index is out of range. This patch makes the interface slightly
more sane by removing the check, which is currently otherwise skipped by
calling itemWithoutBoundsCheck. Note that the vector access will perform
a bounds check (with release assert) anyway. The only caller where the
bounds check behaviour is useful is for the IDL binding so a separate
function itemWithBoundsCheck is created for this (already tested by
value-list-out-of-bounds-crash.html).

Review URL: https://codereview.chromium.org/333163004

git-svn-id: svn://svn.chromium.org/blink/trunk@177064 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/359473008

git-svn-id: svn://svn.chromium.org/blink/trunk@177063 bbb929c8-8fbe-4397-9dbb-9b2b20218538

- Custom bindings for modules are moved into bindings/modules/v8/custom, and
- Custom bindings for core are moved into bindings/core/v8/custom.
- Update gn and gypi files.

BUG=358074

Review URL: https://codereview.chromium.org/331373002

git-svn-id: svn://svn.chromium.org/blink/trunk@177062 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This replaces flags + FullScreenCheckType, for a few reasons:

 1. An UnprefixedRequest will be added, for which neither
    ALLOW_KEYBOARD_INPUT nor FullScreenCheckType is possible. Not all
    previously possible combinations are possible, either.

 2. An enum is clearer than flags at the call site. In particular,
    0 was passed to HTMLMediaElement::enterFullscreen() which would
    disallow keyboard input, which was not obivous and possibly not
    even the original intention. The behavior was preserved but should
    probably be changed to allow keyboard input in a future patch.

 3. LEGACY_MOZILLA_REQUEST as a public part of Element.h is weird.

BUG=383813

Review URL: https://codereview.chromium.org/356933002

git-svn-id: svn://svn.chromium.org/blink/trunk@177061 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=369451
TEST=no new tests because of just refactoring.

Review URL: https://codereview.chromium.org/350863002

git-svn-id: svn://svn.chromium.org/blink/trunk@177060 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Add another test for empty (but rendered) paths.

BUG=388676

Review URL: https://codereview.chromium.org/357943002

git-svn-id: svn://svn.chromium.org/blink/trunk@177059 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/352603003

git-svn-id: svn://svn.chromium.org/blink/trunk@177058 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=

Review URL: https://codereview.chromium.org/350153004

git-svn-id: svn://svn.chromium.org/blink/trunk@177057 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Review URL: https://codereview.chromium.org/355033003

Review URL: https://codereview.chromium.org/355033003

git-svn-id: svn://svn.chromium.org/blink/trunk@177056 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Calling them is basically a layering violation, so it should not be encouraged.

Also made one method of RenderObject private that is already used as such,
and updated RenderLayerClipper::clippingRootForPainting to use an already-available method on RenderLayer that is also used by RenderLayer::enclosingCompositingLayerForPaintInvalidation. This method does almost the same thing as RenderLayerClipper::clippingRootForPainting is intended to do; the latter just throws in an if(transform()).

BUG=384520

Review URL: https://codereview.chromium.org/342203004

git-svn-id: svn://svn.chromium.org/blink/trunk@177055 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Unregister doesn't affect existing control over a document, but prevents the
worker from controlling on future document loads.

We already implement this behavior but no layout test existed yet.

BUG=388095

Review URL: https://codereview.chromium.org/358553003

git-svn-id: svn://svn.chromium.org/blink/trunk@177054 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Rather than reallocating the players vector, we can just reserve exactly the
correct capacity ahead of time. I didn't see these mallocs on a profile, but
this code is executed every frame during composited animations.

R=dstockwell@chromium.org

Review URL: https://codereview.chromium.org/353083002

git-svn-id: svn://svn.chromium.org/blink/trunk@177053 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This CL changes LoadableTextTrack::m_trackElement from a WeakMember to a Member.

In non-oilpan, LoadableTextTrack::m_trackElement is implemented as a weak pointer,
but it's just because there is no way in non-oilpan to make it a strong pointer
without creating a cycle. Ideally LoadableTextTrack::m_trackElement should be
implemented as a strong pointer. This is possible in oilpan, and this CL makes the change.

BUG=340522

Review URL: https://codereview.chromium.org/351563005

git-svn-id: svn://svn.chromium.org/blink/trunk@177052 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=387457

Review URL: https://codereview.chromium.org/345423004

git-svn-id: svn://svn.chromium.org/blink/trunk@177051 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch adds functionality to proactively add "px" suffix if editing
happens in certain CSS properties.

List of properties where only distance of length values are allowed is
taken from http://www.w3.org/TR/CSS21/propidx.html

BUG=387884

Review URL: https://codereview.chromium.org/343343003

git-svn-id: svn://svn.chromium.org/blink/trunk@177050 bbb929c8-8fbe-4397-9dbb-9b2b20218538

There are no callsites in Chromium.  This function is unsafe because it
ignores namespace URI.

BUG=

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=176895

Review URL: https://codereview.chromium.org/354763002

git-svn-id: svn://svn.chromium.org/blink/trunk@177049 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch updates the two remaining tests from the old text autosizer
for the new fast text autosizer.

cluster-list-item.html has been modified to not show the list bullets
and a comment has been added about why.

oscillation-javascript-fontsize-change.html has been updated to not
be a shrink-wrapped div. This changes the test significantly but I
think the test coverage may still be worth it.

BUG=

Review URL: https://codereview.chromium.org/356133003

git-svn-id: svn://svn.chromium.org/blink/trunk@177047 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This switches the default on Win 7 to use the native Win32 version of Apache
2.2 instead of LigHTTPd.  In theory Apache will substantially more stable
and also probably faster. I am hoping that this fixes the situation in bug
378444 where the server just doesn't work on some runs as well.

We also match more of the test expectations on other platforms.

If this change sticks and does improve flakiness, we'll try switching XP
as well in a later change.

This change also deletes all of the FlakyTest entries for http tests
on Win7, so that we can better get a real sense of the flakiness.

R=ojan@chromium.org
BUG=378444, 30536

Review URL: https://codereview.chromium.org/327423008

git-svn-id: svn://svn.chromium.org/blink/trunk@177046 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This isolates ExemptIFrameAllowFullScreenRequirement to a single call
site and makes the code more readable.

R=acolwell@chromium.org

Review URL: https://codereview.chromium.org/339903005

git-svn-id: svn://svn.chromium.org/blink/trunk@177045 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Turns out we're not really ready to do this.
The preloader has lots of bugs at least in the
re-use-existing-preload logic. Yes, we should fix
these bugs. Yes, making all tokens go through the
preloader would force us to fix these bugs. But
until we do, it seems better to sweep them back
under the rug (and just have flakiness in the wild)
rather than making preloader bugs completely
unavoidable by forcing every token through it.

This is a partial revert of:
https://codereview.chromium.org/302063002/

BUG=356292,388308,389226

Review URL: https://codereview.chromium.org/359623004

git-svn-id: svn://svn.chromium.org/blink/trunk@177044 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch adds another round of GC to make sure Document is destroyed after |WorkerMessagingProxy::workerObjectDestroyed|.

BUG=389042

Review URL: https://codereview.chromium.org/350013003

git-svn-id: svn://svn.chromium.org/blink/trunk@177043 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This class is fairly simple, but it's worth taking the rare opportunity
to add a unit test to Blink.

Review URL: https://codereview.chromium.org/343303003

git-svn-id: svn://svn.chromium.org/blink/trunk@177042 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Move attributes-related methods from ElementData to AttributeCollection.
Now that we have a dedicated class for Element attributes, it seems odd for
the attributes-related API to be in Element / ElementData.

This CL does not change the Element API for attributes so that the patch
stays small. However, the idea would be to later move callers away from the
attributes API on Element and instead call Element::attributes() and then
interact directly with the returned AttributeCollection object.

Using AttributeCollection directly has the benefit of avoiding redundant
calls to ElementData::isUnique() and the caller can store AttributeCollection
object in a variable and then any later interaction with this object won't
require checking if ElementData is unique or not.

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

Review URL: https://codereview.chromium.org/354023008

git-svn-id: svn://svn.chromium.org/blink/trunk@177041 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Its implementation is already removed from chromium in r279633.

BUG=386888

Review URL: https://codereview.chromium.org/354073003

git-svn-id: svn://svn.chromium.org/blink/trunk@177040 bbb929c8-8fbe-4397-9dbb-9b2b20218538

No change in logic, but rename and move.

I used different naming convensions based on suggestions by fs@opera.com :
"referencingElements" is renamed "outgoingReferences", and
"referencedElements" is renamed "incomingReferences".

- |SVGDocumentExtensions::setOfElementsReferencingTarget|
  moved to |SVGElement::setOfIncomingReferences|
- |SVGDocumentExtensions::addElementReferencingTarget|
  moved to |SVGElement::addOutgoingReferenceTo|
- |SVGDocumentExtensions::removeAllTargetReferencesForElement|
  moved to |SVGElement::removeAllOutgoingReferences|
- |SVGDocumentExtensions::rebuildAllElementReferencesForTarget|
  moved to |SVGElement::rebuildAllIncomingReferences|
- |SVGDocumentExtensions::removeAllElementReferencesForTarget|
  moved to |SVGElement::removeAllIncomingReferences|

Changes to files SVG\w+Element.cpp made by a sed script:
s/document()\.accessSVGExtensions()\.addElementReferencingTarget(this, \([a-Z()\*]\+\))/addOutgoingReferenceTo(\1)/g
s/document()\.accessSVGExtensions()\.removeAllTargetReferencesForElement(this)/removeAllOutgoingReferences()/g
s/document()\.accessSVGExtensions()\.setOfElementsReferencingTarget(this)/setOfIncomingReferences()/g

BUG=378670

Review URL: https://codereview.chromium.org/344883007

git-svn-id: svn://svn.chromium.org/blink/trunk@177039 bbb929c8-8fbe-4397-9dbb-9b2b20218538

MediaQueryMatcher currently runs script synchronously inside layout and
recalc style which is bad. Instead we should collect all the listeners
that have changed and notify them asynchronously. Incidentally this is
what the spec says to do too, and also what Firefox and the polyfills
do.

This is the first step where we collect all the listeners that changed
in a Vector and then notify them as a group. This also fixes our behavior
where adding a listener in the middle of notification would get that one
called too, which is not what the spec says.

This also changes the notification order such that listeners are notified
in groups based on the creation time of their MediaQueryList which also
matches Firefox (and the polyfill) instead of in the order of global
calls to addListener.

Based on https://codereview.chromium.org/214383008/ by esprehn@

BUG=356487

Review URL: https://codereview.chromium.org/348893004

git-svn-id: svn://svn.chromium.org/blink/trunk@177038 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This makes the callers less awkward, and also removes this API from the
public interface of StyleResolver which is out of hand.

Review URL: https://codereview.chromium.org/356853004

git-svn-id: svn://svn.chromium.org/blink/trunk@177037 bbb929c8-8fbe-4397-9dbb-9b2b20218538

DevTools: Use in JavaScriptSourceFrame current target's debuggerModel instead of WebInspector.debuggerModel

BUG=

Review URL: https://codereview.chromium.org/355823002

git-svn-id: svn://svn.chromium.org/blink/trunk@177036 bbb929c8-8fbe-4397-9dbb-9b2b20218538

This patch moves PositionWithAffinity class to editing/ from dom/Position.h
since it is "affinity" is editing concept rather than DOM concept.

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

Review URL: https://codereview.chromium.org/353603003

git-svn-id: svn://svn.chromium.org/blink/trunk@177035 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Previously we've attempted to have animations and transitions apply as
part of the cascade instead of always as the last step, which required
applying animation-related properties before high priority properties.
This hasn't ended up working and we have no plans to fix the cascade,
so we no longer need the separate animation properties pass.

BUG=232273

Review URL: https://codereview.chromium.org/351553006

git-svn-id: svn://svn.chromium.org/blink/trunk@177034 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Note: Test expectations change for the testBreakpointInCollectedReload test
are not caused by the change, but by the change in test harness.
These changes reveal an error in BreakpointManager implementation that will be fixed in the following patch.

BUG=388131
R=sergeyv

Review URL: https://codereview.chromium.org/350403004

git-svn-id: svn://svn.chromium.org/blink/trunk@177033 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Miniature cleanup, the special case is not needed as the default: case handles
main resources well: it returns early with false as resourceBuffer() is null.

R=pfeldman@chromium.org

Review URL: https://codereview.chromium.org/350263002

git-svn-id: svn://svn.chromium.org/blink/trunk@177032 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Share more code between boundingBoxForCompositing and physicalBoundingBoxIncludingReflectionAndStackingChildren

This also makes it easier to see what the code is computing.

BUG=

Review URL: https://codereview.chromium.org/356853003

git-svn-id: svn://svn.chromium.org/blink/trunk@177031 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Service Worker API is promise-based, so this helper should be too.

Also take care of a FIXME now that testharness has been updated.

BUG=
R=horo@chromium.org

Review URL: https://codereview.chromium.org/354953002

git-svn-id: svn://svn.chromium.org/blink/trunk@177030 bbb929c8-8fbe-4397-9dbb-9b2b20218538

R=caseq, pfeldman, sergeyv

Review URL: https://codereview.chromium.org/351903003

git-svn-id: svn://svn.chromium.org/blink/trunk@177029 bbb929c8-8fbe-4397-9dbb-9b2b20218538

BUG=381728

Review URL: https://codereview.chromium.org/354613002

git-svn-id: svn://svn.chromium.org/blink/trunk@177028 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=hajimehoshi
NOTRY=true
BUG=376188

Review URL: https://codereview.chromium.org/351353003

git-svn-id: svn://svn.chromium.org/blink/trunk@177027 bbb929c8-8fbe-4397-9dbb-9b2b20218538

Also separated incremental invalidation code into
virtual RenderObject::updatePaintInvalidationReason() and
invalidatePaintIncrementally().

Override the functions in RenderBox to contain box-related code
originally in RenderObject::invalidatePaintAfterLayoutIfNeeded().

This doesn't change any current behavior, but makes it possible
to change the box decoration invalidation code to depend on rects
that is only available in RenderBox.

This is preparation for final fix of bug 380472.

BUG=380472
R=leviw@chromium.org

Review URL: https://codereview.chromium.org/350733003

git-svn-id: svn://svn.chromium.org/blink/trunk@177026 bbb929c8-8fbe-4397-9dbb-9b2b20218538

TBR=hajimehoshi
NOTRY=true
BUG=364377

Review URL: https://codereview.chromium.org/354073002

git-svn-id: svn://svn.chromium.org/blink/trunk@177025 bbb929c8-8fbe-4397-9dbb-9b2b20218538