2010-08-05  Simon Hausmann  <simon.hausmann@nokia.com>

        Reviewed by Laszlo Gombos.

        [Qt] Temporary files should not be created in the current directory

        https://bugs.webkit.org/show_bug.cgi?id=43562

        Don't create temporary files in the current directory, use QDir::tempPath()

        instead, as suggested by the QTemporaryFile documentation.

        * platform/qt/FileSystemQt.cpp:

        (WebCore::openTemporaryFile):

2010-08-05  Simon Hausmann  <simon.hausmann@nokia.com>

        Reviewed by Tor Arne Vestbø.

        [Qt] Clean up the input method handling

        https://bugs.webkit.org/show_bug.cgi?id=43545

        Changed input method hint interface to be more efficient by setting

        all hints in one shot, like in QWidget.

        * platform/qt/QWebPageClient.h:

2010-08-04  Abhishek Arya  <inferno@chromium.org>

        Unreviewed. Put missing equivalent js bindings check.

        Forgot putting js bindings check in http://trac.webkit.org/changeset/64647. This fixes the qt crash.

        * bindings/js/JSDesktopNotificationsCustom.cpp:

        (WebCore::JSNotificationCenter::requestPermission):

2010-08-03  Abhishek Arya  <inferno@chromium.org>

        Reviewed by Alexey Proskuryakov.

        Null the script execution context when disconnecting frame in notifications.

        Make sure that script execution context is valid in notification requestPermission.

        https://bugs.webkit.org/show_bug.cgi?id=43295

        Tests: fast/notifications/notifications-document-close-crash.html

        * bindings/v8/custom/V8NotificationCenterCustom.cpp:

        (WebCore::V8NotificationCenter::requestPermissionCallback):

        * notifications/NotificationCenter.cpp:

        (WebCore::NotificationCenter::disconnectFrame):

2010-08-04  Andreas Kling  <andreas.kling@nokia.com>

        Reviewed by Simon Hausmann.

        [Qt] Use a QImage for the stroke applier scratch context

        This avoids leaking a server-side resource on some graphics systems.

        * platform/graphics/qt/PathQt.cpp:

        (WebCore::scratchContext):

2010-07-21 Grace Kloba  <klobag@gmail.com> , Antonio Gomes  <tonikitoo@webkit.org>

        Reviewed by David Hyatt.

        Enhance the hit testing to take a rectangle instead of a point

        https://bugs.webkit.org/show_bug.cgi?id=40197

        The primary goal of this change is to provide mechanisms for more precise tap

        actions by the users on mobile devices.

        Patch extends the hit testing system to work considering a rectangular area

        as input instead of a point, when applicable. For that, the HitTestResult class

        was modified to take a padding (IntSize). The padding specifies a fuzzy range for

        accepting input events in pixels coordinates for both vertical and horizontal

        orientations. In other words, it tells how much to expand the search rect

        around a supposed touch point.

        If it is non-positive padding (e.g. (-1, -1), (5, -1), (0, 0)), hit testing will behavior

        as the current point based hit testing: methods are no-op'ed to not regress it performance-wise

        since it is the common behavior. When positive padding is provided, the HitTestResult class will

        keep record of all nodes that intersect the built up test area. The logic will continue searching

        when it finds a candidate until the hit test area is fully enclosed by the boundaries of a candidate.

        The result will be a list of nodes in the z-order they are hit-tested. Caller will decide how

        to process them.

        In order to expose the functionality, the patch:

        - Adds a nodesFromRect method to the Document class, exposing the funcionality

        to the DOM. Method returns a NodeList with all nodes that intersect the given

        hit-tested area.

        - Extends hitTestResultAtPoint method of the EventHandler with an extra 'padding'

        parameter, defaulting to IntSize(-1, -1). The rect-based hit test is performed when a

        non-negative padding is passed in.

        Test: fast/dom/nodesFromRect-basic.html

        * WebCore.base.exp:

        * dom/Document.cpp:

        (WebCore::Document::nodesFromRect): This method exposes the rect based funcionality to

        the DOM. It works similarly to elementFromPoint, however receiving a rectangular area

        as input instead of a point, and returning a z-index ordered list of nodes (not elements)

        whose area intersect the hit test rect.

        * dom/Document.h: Ditto.

        * dom/Document.idl: Ditto.

        * page/EventHandler.cpp:

        (WebCore::EventHandler::hitTestResultAtPoint): The funcionality is also exposed through this

        method. Patch adds a additional IntSize parameter to work as the padding area, building up

        the hit test rect.

        * page/EventHandler.h: Ditto.

        * rendering/HitTestResult.cpp:

        (WebCore::HitTestResult::HitTestResult): Rect based hit test constructor. Receives a

        padding IntSize as parameter. It can be (0,0).

        (WebCore::HitTestResult::operator=): Modified to assign the m_rectBasedTestResult as well.

        (WebCore::HitTestResult::append): Merge to HitTestResult objects in a way that the

        list node's of both objects get amended.

        (WebCore::HitTestResult::addNodeToRectBasedTestResult): Adds a given Node to the list of

        hit nodes.

        * rendering/HitTestResult.h:

        (WebCore::HitTestResult::padding): Returns the padding as an IntSize.

        (WebCore::HitTestResult::isRectBasedTest): Returns if the HitTestResult is rect based or not.

        (WebCore::HitTestResult::.rectBasedTestResult): Returns the list nodes hit.

        (WebCore::HitTestResult::rectFromPoint): Returns the hit test rect given the hit test point

        and padding.

        * rendering/RenderLayer.cpp:

        (WebCore::RenderLayer::hitTestLayer):

        (WebCore::RenderLayer::hitTestList):

        (WebCore::RenderLayer::hitTestChildLayerColumns):

        * rendering/EllipsisBox.cpp:

        (WebCore::EllipsisBox::nodeAtPoint): Method is modified to support rect based hit test extension.

        Now it not just checks if the boundary of the node being hit-tested contains a hit test point, but

        instead it checks if the boundary of the node intersects a hit test rect. It is implemented so

        that the common case (point based hit test) works as previously.

        * rendering/InlineFlowBox.cpp:

        (WebCore::InlineFlowBox::nodeAtPoint): Ditto.

        * rendering/InlineTextBox.cpp:

        (WebCore::InlineTextBox::nodeAtPoint): Ditto.

        * rendering/RenderBlock.cpp:

        (WebCore::RenderBlock::nodeAtPoint): Ditto.

        (WebCore::RenderBlock::hitTestColumns): Ditto.

        * rendering/RenderBox.cpp:

        (WebCore::RenderBox::nodeAtPoint): Ditto.

        * rendering/RenderImage.cpp:

        (WebCore::RenderImage::nodeAtPoint): Ditto.

        * rendering/RenderLineBoxList.cpp:

        (WebCore::RenderLineBoxList::hitTest):

        * rendering/RenderSVGRoot.cpp:

        (WebCore::RenderSVGRoot::nodeAtPoint): Ditto.

        * rendering/RenderTable.cpp:

        (WebCore::RenderTable::nodeAtPoint): Ditto.

        * rendering/RenderTableSection.cpp:

        (WebCore::RenderTableSection::nodeAtPoint): Ditto.

        * rendering/RenderWidget.cpp:

        (WebCore::RenderWidget::nodeAtPoint): Ditto.

2010-07-22  Ben Murdoch  <benm@google.com>

        Reviewed by Simon Fraser.

        Touch events do not affect the :active CSS state

        https://bugs.webkit.org/show_bug.cgi?id=39493

        Clean the code up by using a typedef based on comments

        post landing of the original patch.

        No change in functionality so no new tests.

        * rendering/HitTestRequest.h: Typedef the HitTestRequestType to unsigned.

        * WebCore.exp.in: Update export signature.

        * WebCore.order: Ditto.

        * page/EventHandler.cpp: use the new typedef.

        * page/EventHandler.h: ditto.

2010-07-24  Andreas Kling  <andreas.kling@nokia.com>

        Reviewed by Simon Hausmann.

        [Qt] tst_QWebFrame::callQtInvokable() fails

        https://bugs.webkit.org/show_bug.cgi?id=41065

        Converting JS objects to QVariantMaps was broken.

        This is a partial revert of <http://trac.webkit.org/changeset/61478>

        * bridge/qt/qt_runtime.cpp:

        (JSC::Bindings::convertValueToQVariant): Remove erroneous crash guard

        around JSObject::getPropertyNames()

2010-07-22  Kim Grönholm  <kim.1.gronholm@nokia.com>

        Reviewed by NOBODY (OOPS!).

        Generate TransformAction events from touch gestures

        https://bugs.webkit.org/show_bug.cgi?id=39979

        Enabling transformaction events to be generated from multi-touch

        gestures by adding gesture handling machinery and eventhandler hooks.

        Test: fast/events/transformaction/basic-transformaction-events.html

        Tests basic single- and multi-touch sequences and the corresponding

        TransformAction events that are generated.

        A manual test can be found from:

        https://bug-39757-attachments.webkit.org/attachment.cgi?id=57226

        * WebCore.pro:

        * page/EventHandler.cpp:

        (WebCore::hasTransformActionEventListener):

        (WebCore::EventHandler::handleTouchEvent):

        * page/EventHandler.h:

        * page/TouchGestureHandler.cpp: Added.

        (WebCore::TouchGestureHandler::TouchGestureHandler):

        (WebCore::TouchGestureHandler::updateRefreshNeededState):

        (WebCore::TouchGestureHandler::updateTouchPoints):

        (WebCore::TouchGestureHandler::reset):

        (WebCore::TouchGestureHandler::calcAngle):

        (WebCore::TouchGestureHandler::diffAngle):

        (WebCore::TouchGestureHandler::calcPos):

        (WebCore::TouchGestureHandler::calcDistance):

        * page/TouchGestureHandler.h: Added.

        (WebCore::TouchGestureHandler::translateX):

        (WebCore::TouchGestureHandler::translateY):

        (WebCore::TouchGestureHandler::pos):

        (WebCore::TouchGestureHandler::screenPos):

        (WebCore::TouchGestureHandler::scale):

        (WebCore::TouchGestureHandler::rotate):

2010-07-22  Kim Grönholm  <kim.1.gronholm@nokia.com>

        Reviewed by NOBODY (OOPS!).

        Add TransformActionEvent support

        https://bugs.webkit.org/show_bug.cgi?id=39757

        Added only the necessary TransformAction event interfaces and not e.g.

        any eventhandler hooks that generate and dispatch them.

        Test: fast/events/transformaction/create-transformaction-event.html

        More tests will be added with the machinery that generates and

        dispatches these events.

        * Android.derived.jscbindings.mk:

        * Android.derived.v8bindings.mk:

        * Android.mk:

        * CMakeLists.txt:

        * DerivedSources.cpp:

        * DerivedSources.make:

        * GNUmakefile.am:

        * WebCore.gypi:

        * WebCore.pri:

        * WebCore.pro:

        * bindings/generic/RuntimeEnabledFeatures.cpp:

        * bindings/generic/RuntimeEnabledFeatures.h:

        (WebCore::RuntimeEnabledFeatures::transformactionEnabled):

        (WebCore::RuntimeEnabledFeatures::setTransformActionEnabled):

        (WebCore::RuntimeEnabledFeatures::ontransformactionstartEnabled):

        (WebCore::RuntimeEnabledFeatures::ontransformactionupdateEnabled):

        (WebCore::RuntimeEnabledFeatures::ontransformactionendEnabled):

        * bindings/js/JSEventCustom.cpp:

        (WebCore::toJS):

        * bindings/v8/custom/V8EventCustom.cpp:

        (WebCore::toV8):

        * dom/Document.cpp:

        (WebCore::Document::createEvent):

        (WebCore::Document::addListenerTypeIfNeeded):

        * dom/Document.h:

        (WebCore::Document::):

        * dom/Document.idl:

        * dom/Element.h:

        * dom/Element.idl:

        * dom/Event.cpp:

        (WebCore::Event::isTransformActionEvent):

        (WebCore::Event::fromUserGesture):

        * dom/Event.h:

        * dom/EventNames.h:

        * dom/TransformActionEvent.cpp: Added.

        (WebCore::TransformActionEvent::TransformActionEvent):

        (WebCore::TransformActionEvent::initTransformActionEvent):

        * dom/TransformActionEvent.h: Added.

        (WebCore::TransformActionEvent::create):

        (WebCore::TransformActionEvent::translateX):

        (WebCore::TransformActionEvent::translateY):

        (WebCore::TransformActionEvent::translateSpeedX):

        (WebCore::TransformActionEvent::translateSpeedY):

        (WebCore::TransformActionEvent::scale):

        (WebCore::TransformActionEvent::scaleSpeed):

        (WebCore::TransformActionEvent::rotate):

        (WebCore::TransformActionEvent::rotateSpeed):

        (WebCore::TransformActionEvent::TransformActionEvent):

        (WebCore::TransformActionEvent::isTransformActionEvent):

        * dom/TransformActionEvent.idl: Added.

        * html/HTMLAttributeNames.in:

        * html/HTMLElement.cpp:

        (WebCore::HTMLElement::parseMappedAttribute):

        * page/DOMWindow.h:

        * page/DOMWindow.idl:

2010-01-07  Laszlo Gombos  <laszlo.1.gombos@nokia.com>

        Reviewed by NOBODY (OOPS!).

        [RVCT] ACID3 test crash

        https://bugs.webkit.org/show_bug.cgi?id=33280

        Workaround developed by Yongjun Zhang.

        * dom/Element.cpp:

        (WebCore::Element::setAttribute):

2010-07-22  Kent Tamura  <tkent@chromium.org>

        Reviewed by Darin Adler.

        Small refactoring for input value sanitization

        https://bugs.webkit.org/show_bug.cgi?id=42807

        Rename some functions to clarify their roles.

        * dom/InputElement.cpp:

        (WebCore::replaceEOLAndLimitLength):

          Renamed from sanitizeUserInputValue().

        (WebCore::InputElement::sanitizeValueForTextField):

          Renamed from sanitizeValue(), and call replaceEOLAndLimitLength()

          instead of sanitizeUserInputValue().

        (WebCore::InputElement::sanitizeUserInputValue):

          Just call replaceEOLAndLimitLength().

        (WebCore::InputElement::updateValueIfNeeded):

          Use non-static sanitizeValue() for consistency.

        * dom/InputElement.h:

        * html/HTMLInputElement.cpp:

        (WebCore::HTMLInputElement::setInputType):

          Revert the change of r63876. It is not needed because of the

          updateValueIfNeeded() change.

        (WebCore::HTMLInputElement::sanitizeValue):

          Apply the sanitizeValue() renaming.

2010-07-22  Tony Gentilcore  <tonyg@chromium.org>

        Reviewed by Darin Fisher.

        webkitPerformance.timing.responseEnd should not include document parse time

        https://bugs.webkit.org/show_bug.cgi?id=42797

        No new tests because timing based test would be flaky.

        * loader/FrameLoader.cpp:

        (WebCore::FrameLoader::finishedLoading): finishedLoading() is called by the platform at the right time, but didReceiveData() synchronously invokes parsing without returning to the event loop prior to this. So by the time the didFinishLoading() method is executed, parsing is finished. The solution is to move this time to didReceiveData() prior to parsing.

        * loader/MainResourceLoader.cpp:

        (WebCore::MainResourceLoader::didReceiveData): Update the time each time didReceiveData() is called.

        (WebCore::MainResourceLoader::didFinishLoading): When finished, and after parsing, set responseEnd appropriately.

        * loader/MainResourceLoader.h:

        * page/Timing.cpp:

        (WebCore::Timing::resourceLoadTimeRelativeToAbsolute): Now that responseEnd is set correctly, it can be much shorter for very fast loads (like local files). This exposed the fact that this skew check was not tight enough. We need to make sure that the whole range of values from the ResourceLoadTiming API fit within fetchStart-responseEnd, not just that requestTime fits in this range.

2010-07-21  Kent Tamura  <tkent@chromium.org>

        Reviewed by Darin Adler.

        Assertion failure by changing the type of an input element with a

        non-number value to 'range'.

        https://bugs.webkit.org/show_bug.cgi?id=42643

        Test: fast/forms/input-value-sanitization.html

        * html/HTMLInputElement.cpp:

        (WebCore::HTMLInputElement::setInputType):

        Update the value by HTMLInputElement::sanitizeValue() in a case that

        storesValueSeparateFromAttribute() state is not changed.

2010-07-21  Adam Barth  <abarth@webkit.org>

        Reviewed by Eric Seidel.

        Implement spec changes for basefont and bgsound

        https://bugs.webkit.org/show_bug.cgi?id=42792

        * html/HTMLTreeBuilder.cpp:

        (WebCore::HTMLTreeBuilder::processStartTagForInBody):

        (WebCore::HTMLTreeBuilder::processStartTag):

        (WebCore::HTMLTreeBuilder::processStartTagForInHead):

2010-07-21  Adam Barth  <abarth@webkit.org>

        Reviewed by Eric Seidel.

        Update a list of tags to match changes in the HTML5 spec

        https://bugs.webkit.org/show_bug.cgi?id=42791

        We need to atomize these strings at some point.

        * html/HTMLTreeBuilder.cpp:

        (WebCore::HTMLTreeBuilder::processStartTagForInBody):

2010-07-21  Adam Barth  <abarth@webkit.org>

        Reviewed by Eric Seidel.

        HTMLTreeBuilder needs to update to match new spec behavior for <button>

        https://bugs.webkit.org/show_bug.cgi?id=42233

        The spec has some typos in this area, but this is my best guess for

        what Ian means.  I've filed bugs against the spec for each typo.

        * html/HTMLElementStack.cpp:

        (WebCore::HTMLNames::isScopeMarker):

        (WebCore::HTMLNames::isButtonScopeMarker):

        (WebCore::HTMLElementStack::inButtonScope):

        * html/HTMLElementStack.h:

        * html/HTMLTreeBuilder.cpp:

        (WebCore::HTMLTreeBuilder::processFakePEndTagIfPInButtonScope):

        (WebCore::HTMLTreeBuilder::processCloseWhenNestedTag):

        (WebCore::HTMLTreeBuilder::processStartTagForInBody):

        (WebCore::HTMLTreeBuilder::processEndTagForInBody):

        * html/HTMLTreeBuilder.h:

2010-07-21  Justin Schuh  <jschuh@chromium.org>

        Reviewed by Oliver Hunt.

        Prevent DeleteButtonController enable state from changing when not editing

        https://bugs.webkit.org/show_bug.cgi?id=42659

        Test: svg/custom/use-invalid-html.xhtml

        * dom/ContainerNode.cpp:

        (WebCore::ContainerNode::cloneChildNodes):

2010-07-21  Beth Dakin  <bdakin@apple.com>

        Reviewed by Dan Bernstein.

        Fix for https://bugs.webkit.org/show_bug.cgi?id=42605 New border-

        radius path-based drawing code has some issues with corner-joins

        GraphicsContext::clipConvexPolygon() now takes a boolean parameter 

        indicating whether or not the clip should be antialiased.

        * platform/graphics/GraphicsContext.h:

        * platform/graphics/cairo/GraphicsContextCairo.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/cg/GraphicsContextCG.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/haiku/GraphicsContextHaiku.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/qt/GraphicsContextQt.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/skia/GraphicsContextSkia.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/wince/GraphicsContextWince.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        * platform/graphics/wx/GraphicsContextWx.cpp:

        (WebCore::GraphicsContext::clipConvexPolygon):

        Compute 4 booleans determining if each edge is a "matching" edge of 

        not. Send the two appropriate values to clipBorderSidePolygon() so 

        we can decide there whether or not to antialias the clip based on 

        matchy-ness.

        * rendering/RenderBoxModelObject.cpp:

        (WebCore::RenderBoxModelObject::paintBorder):

        If both edges are matching, don't antialias. If neither edge 

        matches, never antialias. If one edge matches and one does not, 

        apply two clips to get the appropriate antialiasing in each corner.

        (WebCore::RenderBoxModelObject::clipBorderSidePolygon):

        * rendering/RenderBoxModelObject.h:

2010-07-21  Brady Eidson  <beidson@apple.com>

        Reviewed by Geoffrey Garen.

        Break out "scheme registration" functionality from SecurityOrigin to a SchemeRegistry

        https://bugs.webkit.org/show_bug.cgi?id=42783

        No new tests. (No change in behavior)

        * GNUmakefile.am:

        * WebCore.exp.in:

        * WebCore.gypi:

        * WebCore.pro:

        * WebCore.vcproj/WebCore.vcproj:

        * WebCore.xcodeproj/project.pbxproj:

        * loader/FrameLoader.cpp:

        (WebCore::FrameLoader::isMixedContent):

        (WebCore::FrameLoader::loadFrameRequest):

        * page/SecurityOrigin.cpp:

        (WebCore::SecurityOrigin::SecurityOrigin):

        (WebCore::SecurityOrigin::canLoad):

        (WebCore::SecurityOrigin::isLocal):

        * page/SecurityOrigin.h:

        * platform/SchemeRegistry.cpp: Added.

        (WebCore::localSchemes):

        (WebCore::secureSchemes):

        (WebCore::schemesWithUniqueOrigins):

        (WebCore::SchemeRegistry::registerURLSchemeAsLocal):

        (WebCore::SchemeRegistry::removeURLSchemeRegisteredAsLocal):

        (WebCore::SchemeRegistry::localURLSchemes):

        (WebCore::SchemeRegistry::shouldTreatURLAsLocal):

        (WebCore::SchemeRegistry::shouldTreatURLSchemeAsLocal):

        (WebCore::SchemeRegistry::registerURLSchemeAsNoAccess):

        (WebCore::SchemeRegistry::shouldTreatURLSchemeAsNoAccess):

        (WebCore::SchemeRegistry::registerURLSchemeAsSecure):

        (WebCore::SchemeRegistry::shouldTreatURLSchemeAsSecure):

        * platform/SchemeRegistry.h: Added.

2010-07-21  Chris Marrin  <cmarrin@apple.com>

        Reviewed by Simon Fraser.

        Assertion failure in AnimationBase::updateStateMachine() coming out of paused state

        https://bugs.webkit.org/show_bug.cgi?id=37993

        Added logic to properly handle pausing and resuming when in the 

        AnimationStateStartWaitStyleAvailable state. This was causing an

        assert when going out of the pause state because the paused flag

        was not set.

        The fix is a straightforward implementation, going into a new

        AnimationStatePausedWaitStyleAvailable state and setting the paused

        flag (actually setting the m_pauseTime variable to something other

        than -1). Also added handling of the new state, both when the 

        "style available" callback comes in while in this state and when

        unpausing while in this state. 

        For now a LayoutTest is not possible since there's no way to go in

        and out of the pause state. I've opened https://bugs.webkit.org/show_bug.cgi?id=42790

        to track this.