Timeline


and

04/19/09:

23:25 Changeset [42668] by adele@apple.com

2009-04-19 Adele Peterson <adele@apple.com>

Reviewed by Darin Adler.

Fix for <rdar://problem/6804809> REGRESSION: In Mail, Home and End do not scroll message

If no scrolling occurs, call tryToPerform on the next responder. Then our WebResponderChainSink
will correctly detect if no responders handle the selector.

  • WebView/WebFrameView.mm: (-[WebFrameView _scrollToBeginningOfDocument]): (-[WebFrameView _scrollToEndOfDocument]): (-[WebFrameView scrollToBeginningOfDocument:]): (-[WebFrameView scrollToEndOfDocument:]): (-[WebFrameView scrollLineUp:]): (-[WebFrameView scrollLineDown:]):
22:47 Changeset [42667] by mitz@apple.com
  • maybe fix release builds
  • dom/ScriptElement.cpp: (WebCore::ScriptElementData::notifyFinished):
22:27 Changeset [42666] by mitz@apple.com
  • fix release builds
  • dom/Document.cpp: (WebCore::Document::executeScriptSoonTimerFired):
21:20 Changeset [42665] by mitz@apple.com

WebCore:

Reviewed by Mark Rowe.

  • fix <rdar://problem/6734365> REGRESSION eBay "Save this search" link does nothing the second time search results are opened

Test: fast/dom/HTMLScriptElement/nested-execution.html

Instead of executing scripts as soon as they finish loading, which may
happen synchronously during execution of another script, defer their
execution until the next run loop iteration, using a queue that is
similar to the HTML5 notion of the "list of scripts that will execute as
soon as possible", and in my testing was consistent with how Firefox
behaved.

  • dom/Document.cpp: (WebCore::Document::Document): Initialize m_executeScriptSoonTimer. (WebCore::Document::~Document): deref() script elements in m_scriptsToExecuteSoon. (WebCore::Document::executeScriptSoonTimerFired): Added. Executes the scripts in m_scriptsToExecuteSoon and deref()s them. (WebCore::Document::executeScriptSoon): Added. Appends to m_scriptsToExecuteSoon and ref()s the script element, which keeps the ScriptElementData alive as well.
  • dom/Document.h:
  • dom/ScriptElement.cpp: (WebCore::ScriptElementData::ScriptElementData): Initialize m_requested. (WebCore::ScriptElementData::requestScript): Set m_requested to true, to prevent further load requests. (WebCore::ScriptElementData::execute): Added. Moved the code from notifyFinished() which should not execute synchronously here, to be called by the Document on a 0-interval timer. (WebCore::ScriptElementData::notifyFinished): Moved the code to dispatch events and evaluate the script, which should not execute synchronously, out of here. (WebCore::ScriptElementData::ignoresLoadRequest): Changed to test for m_requested instead of m_cachedScript, because the latter is cleared before the script is evaluated.
  • dom/ScriptElement.h:

LayoutTests:

Reviewed by Mark Rowe.

  • test and updated tests for <rdar://problem/6734365> eBay "Save this search" link does nothing the second time search results are opened
  • fast/dom/HTMLScriptElement/nested-execution-expected.txt: Added.
  • fast/dom/HTMLScriptElement/nested-execution.html: Added.
  • fast/dom/HTMLScriptElement/resources/nested-execution.js: Added.
  • fast/dom/HTMLScriptElement/script-reexecution.html:
  • fast/dom/HTMLScriptElement/script-set-src.html:
  • fast/dom/script-element-gc.html:
  • fast/loader/local-JavaScript-from-local.html:
  • svg/dom/SVGScriptElement/script-reexecution.svg:
  • svg/dom/SVGScriptElement/script-set-href.svg:
18:31 Changeset [42664] by oliver@apple.com

<rdar://problem/6127705> Excessive memory consumption on image load failure

Reviewed by Antti Koivisto and Darin Adler

When we fail to decode an image we clear most of the image data, but not the
input data itself, which can be backed by a CFData object that itself holds onto
a few hundred Kbs of memory. This patch merely ensures that this buffer gets
cleared.

16:38 Changeset [42663] by darin@apple.com

2009-04-19 Darin Adler <darin@apple.com>

  • StringsNotToBeLocalized.txt: Updated for recent changes.
15:54 Changeset [42662] by weinig@apple.com

2009-04-19 Sam Weinig <sam@webkit.org>

Reviewed by Darin Adler.

Better fix for JSStringCreateWithCFString hardening.

  • API/JSStringRefCF.cpp: (JSStringCreateWithCFString):
15:48 Changeset [42661] by darin@apple.com

ooo

15:47 Changeset [42660] by darin@apple.com

2009-04-19 Darin Adler <darin@apple.com>

Reviewed by Sam Weinig.

Add hoooks so sudden termination will not cause loss of icons or local storage.
rdar://problem/5951701

  • loader/icon/IconDatabase.cpp: Marked various constants static to give them internal linkage. (WebCore::IconDatabase::IconDatabase): Added initial value for m_syncTimer. Removed unneeded initial value for m_defaultIconRecord. Changed assertion to work cross-platform. (WebCore::IconDatabase::wakeSyncThread): Added disableSuddenTermination, so we won't terminate until the sync thread has done its thing. (WebCore::IconDatabase::scheduleOrDeferSyncTimer): Added disableSuddenTermination, so we won't terminate if we have an outstanding sync timer. (WebCore::IconDatabase::syncTimerFired): Added enableSuddenTermination, to balance the one in scheduleOrDeferSyncTimer. (WebCore::IconDatabase::syncThreadMainLoop): Added enableSuddenTermination, to balance the one in wakeSyncThread. (WebCore::IconDatabase::deleteAllPreparedStatements): Use clear() instead of set(0).
  • loader/icon/IconDatabase.h: Use a Timer instead of an OwnPtr<Timer>.
  • storage/LocalStorageArea.cpp: (WebCore::LocalStorageArea::scheduleFinalSync): Added disableSuddenTermination. (WebCore::LocalStorageArea::scheduleItemForSync): Ditto. (WebCore::LocalStorageArea::scheduleClear): Ditto. (WebCore::LocalStorageArea::syncTimerFired): Added a disableSuddenTermination if we schedule a performSync callback for later and an unconditional enableSuddenTermination to balance the ones in the schedule calls above. (WebCore::LocalStorageArea::sync): Factored out the work of the sync function so it can continue to use early return idiom. (WebCore::LocalStorageArea::performSync): Added a call to enableSuddenTermination.
  • storage/LocalStorageArea.h: Added declaration of the sync function.
15:26 Changeset [42659] by weinig@apple.com

2009-04-19 Sam Weinig <sam@webkit.org>

Reviewed by Dan Bernstein.

Fix for <rdar://problem/5860954>
Harden JSStringCreateWithCFString against malformed CFStringRefs.

  • API/JSStringRefCF.cpp: (JSStringCreateWithCFString):
10:05 Changeset [42658] by ddkilzer@apple.com

Make FEATURE_DEFINES completely dynamic

Reviewed by Darin Adler.

Make FEATURE_DEFINES depend on individual ENABLE_FEATURE_NAME
variables for each feature, making it possible to remove all
knowledge of FEATURE_DEFINES from build-webkit.

JavaScriptCore:

  • Configurations/FeatureDefines.xcconfig: Extract a variable from FEATURE_DEFINES for each feature setting.

WebCore:

  • Configurations/FeatureDefines.xcconfig: Extract a variable from FEATURE_DEFINES for each feature setting.

WebKit/mac:

  • Configurations/FeatureDefines.xcconfig: Extract a variable from FEATURE_DEFINES for each feature setting.

WebKitTools:

  • Scripts/build-webkit: Define individual ENABLE_FEATURE_NAME variables instead of the whole FEATURE_DEFINES string when building with Xcode. Miscellaneous clean up:
  • Reordered/sorted command-line switch variables.
  • Reordered/sorted command-line switches in $usage.
  • Reordered/sorted command-line switches in GetOptions().
  • Defined $svgSupport to true if any --svg-feature switch is set in order to match "Implies SVG Support" claim.
  • Build JavaScriptGlue after JavaScriptCore again!
  • Fixed --[no-]geolocation for GTK+ port and sorted GTK+ command-line switches.
  • Added "exit 0" statement at the end of the script.
00:38 Changeset [42657] by ggaren@apple.com

2009-04-18 Geoffrey Garen <ggaren@apple.com>

Reviewed by Alexey Proskuryakov.

More fix for https://bugs.webkit.org/show_bug.cgi?id=21260
Unbounded memory growth when churning elements with anonymous event handler functions


Removed a little more complexity from event handler creation and destruction.


Removed the jsProtectedEventListeners, jsProtectedInlineEventListeners,
and jsInlineEventListeners maps, and all the code for managing them.


ProtectedEventListeners don't exist anymore, so they're easy to nix.


Inline EventListeners do still exist, but there's no reason to track
them in a map. The map exists to enable 'removeEventListener' to associate
a unique JSEventListener with a given JavaScript function. But the
'removeEventListener' API only works with non-inline event listeners!


  • bindings/js/JSDOMGlobalObject.cpp: (WebCore::JSDOMGlobalObject::~JSDOMGlobalObject): (WebCore::JSDOMGlobalObject::findJSEventListener): (WebCore::JSDOMGlobalObject::findOrCreateJSEventListener): (WebCore::JSDOMGlobalObject::createJSInlineEventListener):
  • bindings/js/JSDOMGlobalObject.h:
  • bindings/js/JSEventListener.cpp: (WebCore::JSEventListener::JSEventListener): (WebCore::JSEventListener::clearJSFunctionInline):
  • bindings/js/JSLazyEventListener.cpp: (WebCore::JSLazyEventListener::~JSLazyEventListener): (WebCore::JSLazyEventListener::parseCode):
  • bindings/scripts/CodeGeneratorJS.pm:

04/18/09:

23:42 Changeset [42656] by weinig@apple.com

2009-04-18 Sam Weinig <sam@webkit.org>

Reviewed by Dan Bernstein.

Fix typo. s/VM_MEMORY_JAVASCRIPT_JIT_REGISTER_FILE/VM_MEMORY_JAVASCRIPT_CORE/

  • runtime/Collector.cpp: (JSC::allocateBlock): Fix bozo typo.
23:39 Changeset [42655] by mitz@apple.com
  • try to fix the Windows build
  • editing/VisiblePosition.cpp: (WebCore::VisiblePosition::characterAfter):
23:29 Changeset [42654] by mitz@apple.com
  • revert the previous change, which was to a file that the Windows port does not even use
  • icu/unicode/utf16.h:
23:17 Changeset [42653] by simon.fraser@apple.com

Fix the dates in my Changelog entries

22:28 Changeset [42652] by mitz@apple.com

Reviewed by Geoffrey Garen.

  • try to fix the Windows build
  • icu/unicode/utf16.h:
22:00 Changeset [42651] by simon.fraser@apple.com

2009-04-17 Simon Fraser <simon.fraser@apple.com>

Reviewed by Antti Koivisto.

https://bugs.webkit.org/show_bug.cgi?id=25066

When a <video> element gained a RenderLayer via opacity, reflection etc., the
layer hierarchy was not correctly updated because RenderMedia skipped a level
when asked for its children; it skipped m_controlsShadowRoot's renderer, which
actually has a layer.

Test: fast/layers/video-layer.html

  • rendering/MediaControlElements.cpp: (WebCore::MediaControlShadowRootElement::MediaControlShadowRootElement): Don't manually call setParent() on the renderer. It will happen later as a result of addChild().
  • rendering/RenderMedia.cpp: (WebCore::RenderMedia::createControlsShadowRoot): Add m_controlsShadowRoot's renderer as a child.
  • rendering/RenderMedia.h: (WebCore::RenderMedia::children): Now maintain a RenderObjectChildList, m_children, and remove the unneeded removeChild() method. Make the two children() methods inline.
  • rendering/RenderObject.cpp: (WebCore::RenderObject::moveLayers): Assert if moveLayers() is called with an oldParent that is not the layer's actual parent (which would have revealed this bug).
21:58 Changeset [42650] by ggaren@apple.com

2009-04-18 Geoffrey Garen <ggaren@apple.com>

Reviewed by Cameron Zwarich.

Fixed a layout test failure, caused by my last check-in
(fast/dom/script-element-gc.html).

  • bindings/js/JSDOMBinding.cpp: (WebCore::isObservableThroughDOM):
  • html/HTMLScriptElement.h: (WebCore::HTMLScriptElement::haveFiredLoadEvent): Treat script elements just like image elements: if a script element is loading, mark its JS wrapper, even if the element is not in the document.
19:14 Changeset [42649] by weinig@apple.com

2009-04-18 Sam Weinig <sam@webkit.org>

Reviewed by Anders Carlsson.

Fix for <rdar://problem/6801555> Tag JavaScript memory on SnowLeopard

  • interpreter/RegisterFile.h: (JSC::RegisterFile::RegisterFile):
  • jit/ExecutableAllocatorFixedVMPool.cpp: (JSC::FixedVMPoolAllocator::FixedVMPoolAllocator):
  • jit/ExecutableAllocatorPosix.cpp: (JSC::ExecutablePool::systemAlloc):
  • runtime/Collector.cpp: (JSC::allocateBlock):
18:40 Changeset [42648] by dimich@chromium.org

2009-04-18 Jian Li <jianli@chromium.org>

Reviewed by Dimitri Glazkov.

https://bugs.webkit.org/show_bug.cgi?id=25170
Upstream V8WorkerCustom.cpp and V8WorkerContextCustom.cpp for V8 bindings.

  • bindings/v8/WorkerContextExecutionProxy.cpp: (WebCore::WorkerContextExecutionProxy::findOrCreateEventListener):
  • bindings/v8/WorkerContextExecutionProxy.h:
  • bindings/v8/custom/V8WorkerContextCustom.cpp: Added.
  • bindings/v8/custom/V8WorkerCustom.cpp: Added.
18:40 Changeset [42647] by dimich@chromium.org

2009-04-18 Jian Li <jianli@chromium.org>

Reviewed by Dimitri Glazkov.

https://bugs.webkit.org/post_bug.cg://bugs.webkit.org/show_bug.cgi?id=25167
Change createHiddenXHRDependency and removeHiddenXHRDependency so that they can be used by other custom code in V8 bindings.

  • bindings/v8/V8Utilities.cpp: Renamed from WebCore/bindings/v8/V8XMLHttpRequestUtilities.cpp. (WebCore::createHiddenDependency): (WebCore::removeHiddenDependency):
  • bindings/v8/V8Utilities.h: Renamed from WebCore/bindings/v8/V8XMLHttpRequestUtilities.h.
  • bindings/v8/custom/V8XMLHttpRequestCustom.cpp: (WebCore::ACCESSOR_SETTER): (WebCore::CALLBACK_FUNC_DECL):
  • bindings/v8/custom/V8XMLHttpRequestUploadCustom.cpp: (WebCore::ACCESSOR_SETTER): (WebCore::CALLBACK_FUNC_DECL):
18:19 Changeset [42646] by ddkilzer@apple.com

2009-04-18 Drew Wilson <amw@apple.com>

<rdar://problem/6781407> VisiblePosition.characterAfter should return UChar32

Reviewed by Dan Bernstein.

JavaScriptCore:

  • wtf/unicode/icu/UnicodeIcu.h: (WTF::Unicode::hasLineBreakingPropertyComplexContextOrIdeographic): Added.

WebCore:

No test case since there is no change in behavior.

  • editing/VisiblePosition.cpp: (WebCore::VisiblePosition::characterAfter): Now handles surrogate pairs correctly and returns a UChar32.
  • editing/VisiblePosition.h: (WebCore::VisiblePosition::characterBefore): Now returns a UChar32.
  • editing/visible_units.cpp: (WebCore::endOfFirstWordBoundaryContext): Renamed firstNonComplexContextLineBreak() to this. Changed it to use the generic requiresContextForWordBoundary() instead of hasLineBreakingPropertyComplexContext(). (WebCore::startOfLastWordBoundaryContext): Replaces lastNonComplexContextLineBreak(), but returns the offset of the character after the last character that does not require context for word breaking. Also changed to use requiresContextForWordBoundary(). (WebCore::previousBoundary): Updated for the above changes in names and semantics, and changed to use requiresContextForWordBoundary(). (WebCore::nextBoundary): Ditto. (WebCore::startWordBoundary): Ditto. (WebCore::endWordBoundary): Ditto. (WebCore::previousWordPositionBoundary): Ditto. (WebCore::nextWordPositionBoundary): Ditto.
  • platform/text/TextBoundaries.h: (WebCore::requiresContextForWordBoundary): Added.
  • platform/text/mac/TextBoundaries.mm:
18:00 Changeset [42645] by ggaren@apple.com

WebCore:

2009-04-18 Geoffrey Garen <ggaren@apple.com>

Reviewed by Cameron Zwarich.

Fixed <rdar://problem/6797442> | https://bugs.webkit.org/show_bug.cgi?id=25285
REGRESSION (r42569): Profiles cannot be selected in inspector profiler


EventListeners are shared; RegisteredEventListeners are not. Therefore,
when a node wrapper needs to invalidate the node's JS event listeners,
it should invalidate its RegisteredEventListeners, not its EventListeners.
Otherwise, it might invalidate an EventListener shared by another node.

  • dom/RegisteredEventListener.h: (WebCore::invalidateEventListeners): ditto

LayoutTests:

2009-04-18 Geoffrey Garen <ggaren@apple.com>

Reviewed by Cameron Zwarich.


Test for <rdar://problem/6797442> | https://bugs.webkit.org/show_bug.cgi?id=25285
REGRESSION (r42569): Profiles cannot be selected in inspector profiler

  • fast/events/event-listener-sharing-expected.txt: Added.
  • fast/events/event-listener-sharing.html: Added.
  • fast/events/resources/event-listener-sharing.js: Added. (gc): (clickHandler):
16:50 Changeset [42644] by weinig@apple.com

2009-04-18 Sam Weinig <sam@webkit.org>

Reviewed by Mark Rowe.

Fix for <rdar://problem/5861045>
A little bit of hardening for UString.

  • runtime/UString.cpp: (JSC::concatenate): (JSC::UString::append):
15:50 Changeset [42643] by weinig@apple.com

2009-04-18 Sam Weinig <sam@webkit.org>

Reviewed by Mark Rowe and Dan Bernstein.

Fix for <rdar://problem/5861188>
A little bit of hardening for Vector.

  • wtf/Vector.h: (WTF::Vector<T, inlineCapacity>::append): (WTF::Vector<T, inlineCapacity>::insert):
13:29 Changeset [42642] by eric.carlson@apple.com

2009-04-18 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Mark Rowe.

<rdar://problem/6781295> video.buffered and video.seekable are not
the same.

  • WebCore.base.exp: Added wkQTMovieMaxTimeSeekable.
  • platform/graphics/mac/MediaPlayerPrivateQTKit.mm: (WebCore::MediaPlayerPrivate::maxTimeSeekable): Return wkQTMovieMaxTimeSeekable.
  • platform/mac/WebCoreSystemInterface.h: Add wkQTMovieMaxTimeSeekable.
  • platform/mac/WebCoreSystemInterface.mm: Ditto.

2009-04-18 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Mark Rowe.

<rdar://problem/6781295> video.buffered and video.seekable are not
the same.

  • WebCoreSupport/WebSystemInterface.m: (InitWebCoreSystemInterface): Add wkQTMovieMaxTimeSeekable.

2009-04-18 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Mark Rowe.

<rdar://problem/6781295> video.buffered and video.seekable are not
the same.

  • WebKitSystemInterface.h:
  • libWebKitSystemInterfaceLeopard.a:
  • libWebKitSystemInterfaceTiger.a:
12:11 Changeset [42641] by eric.carlson@apple.com

2009-04-18 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Mark Rowe.

<rdar://problem/6747241> work around QTKit no longer reaching
QTMovieLoadStateComplete

  • WebCore.base.exp: Export wkQTMovieMaxTimeLoadedChangeNotification.
  • platform/graphics/mac/MediaPlayerPrivateQTKit.mm: (WebCore::MediaPlayerPrivate::createQTMovie): observe QTMovieMaxTimeLoadedChangeNotification. (WebCore::MediaPlayerPrivate::updateStates): compare duuration() with maxTimeLoaded() instead of using QTMovieLoadStateComplete to determine if a movie are fully loaded.
  • platform/mac/WebCoreSystemInterface.h: Add wkQTMovieMaxTimeLoadedChangeNotification.
  • platform/mac/WebCoreSystemInterface.mm: Ditto.

2009-04-18 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Mark Rowe.

<rdar://problem/6747241> work around QTKit no longer reaching
QTMovieLoadStateComplete

  • WebCoreSupport/WebSystemInterface.m: (InitWebCoreSystemInterface): Init the new WKSI exported symbol.
11:49 Changeset [42640] by justin.garcia@apple.com

2009-04-18 Justin Garcia <justin.garcia@apple.com>

Updated test results for https://bugs.webkit.org/show_bug.cgi?id=25277

  • platform/mac/fast/forms/input-text-scroll-left-on-blur-expected.checksum:
  • platform/mac/fast/forms/input-text-scroll-left-on-blur-expected.png:
  • platform/mac/fast/forms/input-text-scroll-left-on-blur-expected.txt:
04:49 Changeset [42639] by staikos@webkit.org

2009-04-18 Nikolas Zimmermann <nikolas.zimmermann@torchmobile.com>

Reviewed by George Staikos.

Make WML work again, most regressions test were failing.

04/17/09:

21:44 Changeset [42638] by barraclough@apple.com

2009-04-17 Gavin Barraclough <barraclough@apple.com>

Reviewed by Geoff Garen.

On x86_64, make all JIT-code allocations from a new heap, managed
by FixedVMPoolAllocator. This class allocates a single large (2Gb)
pool of virtual memory from which all further allocations take place.
Since all JIT code is allocated from this pool, we can continue to
safely assume (as is already asserted) that it will always be possible
to link any JIT-code to JIT-code jumps and calls.

Add new file.

  • jit/ExecutableAllocatorFixedVMPool.cpp: Added. (JSC::FreeListEntry::FreeListEntry): (JSC::AVLTreeAbstractorForFreeList::get_less): (JSC::AVLTreeAbstractorForFreeList::set_less): (JSC::AVLTreeAbstractorForFreeList::get_greater): (JSC::AVLTreeAbstractorForFreeList::set_greater): (JSC::AVLTreeAbstractorForFreeList::get_balance_factor): (JSC::AVLTreeAbstractorForFreeList::set_balance_factor): (JSC::AVLTreeAbstractorForFreeList::null): (JSC::AVLTreeAbstractorForFreeList::compare_key_key): (JSC::AVLTreeAbstractorForFreeList::compare_key_node): (JSC::AVLTreeAbstractorForFreeList::compare_node_node): (JSC::sortFreeListEntriesByPointer): (JSC::sortCommonSizedAllocations): (JSC::FixedVMPoolAllocator::release): (JSC::FixedVMPoolAllocator::reuse): (JSC::FixedVMPoolAllocator::addToFreeList): (JSC::FixedVMPoolAllocator::coalesceFreeSpace): (JSC::FixedVMPoolAllocator::FixedVMPoolAllocator): (JSC::FixedVMPoolAllocator::alloc): (JSC::FixedVMPoolAllocator::free): (JSC::ExecutableAllocator::intializePageSize): (JSC::ExecutablePool::systemAlloc): (JSC::ExecutablePool::systemRelease):

The new 2Gb heap class!

  • jit/ExecutableAllocatorPosix.cpp:

Disable use of this implementation on x86_64.

  • wtf/AVLTree.h:

Add missing variable initialization.

(WTF::::remove):

19:23 Changeset [42637] by justin.garcia@apple.com

Accidently checked in outdated pixel results.

19:04 Changeset [42636] by oliver@apple.com

Fix bug where the VM reentry cache would not correctly unroll the cached callframe

Reviewed by Darin Adler

Fix a check that was intended to mark a cached call as invalid when the callframe could
not be constructed. Instead it was just checking that there was a place to put the
exception. This eventually results in a non-recoverable RegisterFile starvation.

18:59 Changeset [42635] by justin.garcia@apple.com

Forgot to uncomment some code before checking in.

18:53 Changeset [42634] by justin.garcia@apple.com

WebCore:

2009-04-17 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.

https://bugs.webkit.org/show_bug.cgi?id=25277
If there is no line wrapping, a caret can leave its containing block but it shouldn't leave its root line box


Also added code to handle a caret wider than one pixel.


Added fast/inline/25277.html and fast/inline/25277-2.html

  • rendering/RenderText.cpp: (WebCore::RenderText::localCaretRect): Distribute a caret's width to either side of the offset, so that a caret between two characters doesn't hang over one character more than the other. If there is no wrapping, the caret can leave its containing block, but not its root line box.

LayoutTests:

2009-04-17 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.


https://bugs.webkit.org/show_bug.cgi?id=25277
If there is no line wrapping, a caret can leave its containing block but it shouldn't leave its root line box

  • fast/inline/25277-2.html: Added.
  • fast/inline/25277.html: Added.
  • platform/mac/fast/inline/25277-2-expected.checksum: Added.
  • platform/mac/fast/inline/25277-2-expected.png: Added.
  • platform/mac/fast/inline/25277-2-expected.txt: Added.
  • platform/mac/fast/inline/25277-expected.checksum: Added.
  • platform/mac/fast/inline/25277-expected.png: Added.
  • platform/mac/fast/inline/25277-expected.txt: Added.
18:31 Changeset [42633] by ggaren@apple.com

2009-04-17 Geoffrey Garen <ggaren@apple.com>

Reviewed by Oliver Hunt.

More fix for https://bugs.webkit.org/show_bug.cgi?id=21260
Unbounded memory growth when churning elements with anonymous event handler functions


Some refactoring of "inline" event listeners.


Renames:

dispatchEventForType => dispatchEvent
setWindowInlineEventListenerForTypeAndAttribute => setWindowInlineEventListener
removeInlineEventListenerForType => clearInlineEventListener
setInlineEventListenerForType => setInlineEventListener
inlineEventListenerForType => getInlineEventListener

  • bindings/js/JSLazyEventListener.cpp: (WebCore::eventParameterName): (WebCore::JSLazyEventListener::JSLazyEventListener): (WebCore::createInlineEventListener):
  • bindings/js/JSLazyEventListener.h: Added two helper functions for creating "inline" event listeners. These replace Document::createEventListener, and abstract the creation of JS event listeners for DOM attributes out of the DOM. Removed legacy JSProtectedEventListener constructor code for adding the event listener's function to a map, since lazy event listeners don't have functions at construction time.
  • dom/Document.cpp: (WebCore::Document::setFocusedNode):
  • dom/Document.h: (WebCore::Document::isSVGDocument):
  • dom/Element.cpp: (WebCore::Element::setWindowInlineEventListener):
  • dom/Element.h: Updated for renames. Nixed Document::createEventListener, mentioned above. Moved setWindowInlineEventListenerForTypeAndAttribute to Element, for simplicity.
  • dom/InputElement.cpp: (WebCore::InputElement::setValueFromRenderer):
  • dom/Node.cpp: (WebCore::Node::dispatchFocusEvent): (WebCore::Node::dispatchBlurEvent): (WebCore::Node::dispatchEvent): (WebCore::Node::clearInlineEventListener): (WebCore::Node::setInlineEventListener): (WebCore::Node::getInlineEventListener): (WebCore::Node::onabort): (WebCore::Node::setOnabort): (WebCore::Node::etc.):
  • dom/Node.h: Updated for renames.


  • editing/ReplaceSelectionCommand.cpp: (WebCore::ReplacementFragment::ReplacementFragment): Updated for renames.


Updated these files for renames, and to use the new createInlineEventListener
helper function:

  • html/HTMLBodyElement.cpp: (WebCore::HTMLBodyElement::parseMappedAttribute):
  • html/HTMLButtonElement.cpp: (WebCore::HTMLButtonElement::parseMappedAttribute):
  • html/HTMLElement.cpp: (WebCore::HTMLElement::parseMappedAttribute):
  • html/HTMLFormControlElement.cpp: (WebCore::HTMLFormControlElement::onChange):
  • html/HTMLFormElement.cpp: (WebCore::HTMLFormElement::prepareSubmit): (WebCore::HTMLFormElement::reset): (WebCore::HTMLFormElement::parseMappedAttribute):
  • html/HTMLFrameElementBase.cpp: (WebCore::HTMLFrameElementBase::parseMappedAttribute):
  • html/HTMLFrameSetElement.cpp: (WebCore::HTMLFrameSetElement::parseMappedAttribute):
  • html/HTMLImageElement.cpp: (WebCore::HTMLImageElement::parseMappedAttribute):
  • html/HTMLImageLoader.cpp: (WebCore::HTMLImageLoader::dispatchLoadEvent):
  • html/HTMLInputElement.cpp: (WebCore::HTMLInputElement::parseMappedAttribute): (WebCore::HTMLInputElement::onSearch):
  • html/HTMLMediaElement.cpp: (WebCore::HTMLMediaElement::loadInternal):
  • html/HTMLObjectElement.cpp: (WebCore::HTMLObjectElement::parseMappedAttribute):
  • html/HTMLScriptElement.cpp: (WebCore::HTMLScriptElement::parseMappedAttribute): (WebCore::HTMLScriptElement::dispatchLoadEvent): (WebCore::HTMLScriptElement::dispatchErrorEvent):
  • html/HTMLSelectElement.cpp: (WebCore::HTMLSelectElement::parseMappedAttribute):
  • html/HTMLTextAreaElement.cpp: (WebCore::HTMLTextAreaElement::parseMappedAttribute):
  • html/HTMLTokenizer.cpp: (WebCore::HTMLTokenizer::notifyFinished):
  • page/AccessibilityRenderObject.cpp: (WebCore::AccessibilityRenderObject::mouseButtonListener):
  • page/DOMWindow.cpp:
  • page/DOMWindow.h: (WebCore::DOMWindow::eventListeners):
  • page/EventHandler.cpp: (WebCore::EventHandler::canMouseDownStartSelect): (WebCore::EventHandler::canMouseDragExtendSelect): (WebCore::EventHandler::sendScrollEvent):
  • page/Page.cpp: (WebCore::networkStateChanged):
  • rendering/RenderListBox.cpp: (WebCore::RenderListBox::valueChanged):
  • rendering/RenderTextControl.cpp: (WebCore::RenderTextControl::selectionChanged):
  • svg/SVGElement.cpp: (WebCore::SVGElement::parseMappedAttribute):
  • svg/SVGElementInstance.cpp:
  • svg/SVGImageLoader.cpp: (WebCore::SVGImageLoader::dispatchLoadEvent):
  • svg/SVGSVGElement.cpp: (WebCore::SVGSVGElement::parseMappedAttribute):
  • svg/SVGScriptElement.cpp: (WebCore::SVGScriptElement::dispatchErrorEvent):
  • wml/WMLInputElement.cpp: (WebCore::WMLInputElement::defaultEventHandler):
16:52 Changeset [42632] by andersca@apple.com

2009-04-17 Anders Carlsson <andersca@apple.com>

Reviewed by Dan Bernstein.

<rdar://problem/6714488>
REGRESSION (Safari 3-4): Edit menu commands (cut/copy/paste/select all) do not work on Flash content


  • Plugins/Hosted/NetscapePluginInstanceProxy.h:
  • Plugins/Hosted/NetscapePluginInstanceProxy.mm: (WebKit::NetscapePluginInstanceProxy::syntheticKeyDownWithCommandModifier): Send a keyDown event to the plug-in host.


  • Plugins/Hosted/WebHostedNetscapePluginView.mm: (-[WebHostedNetscapePluginView sendModifierEventWithKeyCode:character:]): Call the plug-in instance proxy.


  • Plugins/WebBaseNetscapePluginView.mm: (-[WebBaseNetscapePluginView sendModifierEventWithKeyCode:character:]): Add this. Subclasses are required to override it.


(-[WebBaseNetscapePluginView cut:]):
(-[WebBaseNetscapePluginView copy:]):
(-[WebBaseNetscapePluginView paste:]):
(-[WebBaseNetscapePluginView selectAll:]):
Call sendModifierEventWithKeyCode.


  • Plugins/WebNetscapePluginEventHandler.h: Add syntheticKeyDownWithCommandModifier.


  • Plugins/WebNetscapePluginEventHandlerCarbon.h:
  • Plugins/WebNetscapePluginEventHandlerCarbon.mm: (WebNetscapePluginEventHandlerCarbon::syntheticKeyDownWithCommandModifier): Send the synthetic event.


  • Plugins/WebNetscapePluginEventHandlerCocoa.h:
  • Plugins/WebNetscapePluginEventHandlerCocoa.mm: (WebNetscapePluginEventHandlerCocoa::syntheticKeyDownWithCommandModifier): Send the synthetic event.


  • Plugins/WebNetscapePluginView.mm: (-[WebNetscapePluginView sendModifierEventWithKeyCode:character:]): Call the event handler.
16:11 Changeset [42631] by ddkilzer@apple.com

Simplify FEATURE_DEFINES definition

Reviewed by Darin Adler.

This moves FEATURE_DEFINES and its related ENABLE_FEATURE_NAME
variables to their own FeatureDefines.xcconfig file. It also
extracts a new ENABLE_GEOLOCATION variable so that
FEATURE_DEFINES only needs to be defined once.

JavaScriptCore:

  • Configurations/FeatureDefines.xcconfig: Added.
  • Configurations/JavaScriptCore.xcconfig: Removed definition of ENABLE_SVG_DOM_OBJC_BINDINGS and FEATURE_DEFINES. Added include of FeatureDefines.xcconfig.
  • JavaScriptCore.xcodeproj/project.pbxproj: Added FeatureDefines.xcconfig file.

WebCore:

  • Configurations/FeatureDefines.xcconfig: Added.
  • Configurations/WebCore.xcconfig: Removed definition of ENABLE_SVG_DOM_OBJC_BINDINGS and FEATURE_DEFINES. Added include of FeatureDefines.xcconfig.
  • WebCore.xcodeproj/project.pbxproj: Added FeatureDefines.xcconfig file.
  • bindings/scripts/CodeGeneratorObjC.pm: When creating a list of command-line macro definitions, split on one-or-more spaces instead of a single space since the FEATURE_DEFINES macro may now contain more than one space between macros if some macros aren't defined.

WebKit:

  • WebKit.xcodeproj/project.pbxproj: Added FeatureDefines.xcconfig file.

WebKit/mac:

  • Configurations/FeatureDefines.xcconfig: Added.
  • Configurations/WebKit.xcconfig: Removed definition of ENABLE_SVG_DOM_OBJC_BINDINGS and FEATURE_DEFINES. Added include of FeatureDefines.xcconfig.

WebKitTools:

  • Scripts/build-webkit: Realphabetized code related to enabling Geolation features for WebKit. Also set ENABLE_GEOLOCATION variable for xcodebuild.
15:54 Changeset [42630] by adachan@apple.com

<rdar://problem/6803905> Lots of ASSERTs firing when trying to clear history

Reviewed by Steve Falkenburg.

  • WebHistory.cpp: (WebHistory::data): Return null if there are no history items.
15:29 Changeset [42629] by simon.fraser@apple.com

2009-04-17 Simon Fraser <simon.fraser@apple.com>

Reviewed by Steve Falkenburg

Fix the link for setting the Platform SDK paths in the Windows instructions.

  • building/build.html:
15:28 QtWebKitTodo edited by hausmann@webkit.org
(diff)
15:27 QtWebKitJournal edited by hausmann@webkit.org
(diff)
14:49 Changeset [42628] by beidson@apple.com

WebCore:

2009-04-17 Brady Eidson <beidson@apple.com>

Reviewed by Mark Rowe

<rdar://problem/6786999> Tiger WebKit shouldn't try to POST on back/forward without nagging.

  • platform/network/mac/ResourceHandleMac.mm: (WebCore::ResourceHandle::willLoadFromCache):

LayoutTests:

2009-04-17 Brady Eidson <beidson@apple.com>

Reviewed by Mark Rowe

<rdar://problem/6778137> - Disable post-goback-repost-policy.html until the networking layer allows it to be useful.

  • http/tests/navigation/post-goback-repost-policy.html: Removed.
  • http/tests/navigation/post-goback-repost-policy.html-disabled: Copied from http/tests/navigation/post-goback-repost-policy.html.
13:50 Changeset [42627] by darin@apple.com

2009-04-17 Darin Adler <darin@apple.com>

Reviewed by Timothy Hatcher.

Added sudden termination functions that can be called without involving
the page or client.

  • GNUmakefile.am: Added SuddenTermination.h.
  • WebCore.vcproj/WebCore.vcproj: Ditto.
  • WebCore.xcodeproj/project.pbxproj: Ditto. Also added SuddenTermination.mm.
  • platform/SuddenTermination.h: Added.
  • platform/mac/SuddenTermination.mm: Added.
13:45 Changeset [42626] by eric@webkit.org

Reviewed by Simon Fraser.

Move more code into RenderSVGModelObject

needsLayer() already returns false from RenderSVGModelObject
lineHeight and baselinePosition are from days of RenderPath having RenderBox parents.

I also added comments to explain how focus rings work (now that I understand)
I got rid of two more uses of m_absoluteBounds (nearly dead!)

  • rendering/RenderPath.cpp:
  • rendering/RenderPath.h:
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::paint): (WebCore::RenderSVGContainer::addFocusRingRects):
  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGModelObject.cpp: (WebCore::RenderSVGModelObject::absoluteRects): (WebCore::RenderSVGModelObject::absoluteQuads):
  • rendering/RenderSVGModelObject.h:
13:44 Changeset [42625] by adachan@apple.com
  • Fix a leak in WebHistoryItem::children() by adopting the created instance of WebHistoryItem.
  • Remove the unnecessary call to QueryInterface() which is not related to the leak.

Reviewed by Adam Roben.

  • WebHistoryItem.cpp: (WebHistoryItem::children):
13:42 Changeset [42624] by pkasting@chromium.org

2009-04-16 Peter Kasting <pkasting@google.com>

Reviewed by Adele Peterson.

https://bugs.webkit.org/show_bug.cgi?id=25238
Invalidate the hovered scrollbar part when resetting the pressed part,
since once there is no pressed part we can draw a hovered state.

  • platform/Scrollbar.cpp: (WebCore::Scrollbar::setHoveredPart): (WebCore::Scrollbar::setPressedPart):
13:31 Changeset [42623] by darin@chromium.org

2009-04-16 Darin Fisher <darin@chromium.org>

Reviewed by Darin Adler.

https://bugs.webkit.org/show_bug.cgi?id=25223
REGRESSION: Back button after form submission to the same URL fails to navigate.

Test: http/tests/navigation/post-goback-same-url.html

  • loader/FrameLoader.cpp: (WebCore::FrameLoader::loadItem): Check if the page we are leaving has FormData.
13:09 Changeset [42622] by zimmermann@webkit.org

Not reviewed. Fix WML enabled builds.

12:53 Changeset [42621] by kov@webkit.org

2009-04-17 Gustavo Noronha Silva <gustavo.noronha@collabora.co.uk>

Unreviewed build fix to generate all .mo files at the correct
location.

  • GNUmakefile.am:
12:53 Changeset [42620] by darin@apple.com

2009-04-17 Anders Carlsson <andersca@apple.com>

Reviewed by Darin Adler.

Fix crashes seen in regression tests with hosted plug-ins.

  • Plugins/Hosted/NetscapePluginInstanceProxy.mm: (WebKit::NetscapePluginInstanceProxy::cancelStreamLoad): Check the stream for 0, not the stream ID.
12:35 Changeset [42619] by eric.carlson@apple.com

2009-04-17 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Simon Fraser.

Add support for the media element 'played' attribute.
This involves support of "normalized TimeRanges" as described by
the spec.

Test: media/video-played.html

  • html/HTMLMediaElement.cpp: (WebCore::HTMLMediaElement::HTMLMediaElement): Renamed m_currentTimeDuringSeek->m_lastSeekTime. Added m_playing and m_playedTimeRanges. (WebCore::HTMLMediaElement::loadInternal): Reset m_playing, m_lastSeekTime, and m_playedTimeRanges (WebCore::HTMLMediaElement::seek): Set current time to m_playedTimeRanges (WebCore::HTMLMediaElement::currentTime): m_currentTimeDuringSeek -> m_lastSeekTime (WebCore::HTMLMediaElement::played): Returns m_playedTimeRanges. (WebCore::HTMLMediaElement::updatePlayState): Add the new played time ranges when pausing. Set the m_playing flag.
  • html/HTMLMediaElement.h:
  • html/TimeRanges.cpp: (TimeRanges::add): Merge range(s) if overlap.
  • html/TimeRanges.h: Add various helpers (below). (WebCore::TimeRanges::Range::isPointInRange): Added. (WebCore::TimeRanges::Range::isOverlappingRange): Ditto. (WebCore::TimeRanges::Range::unionWithOverlappingRange): Ditto. (WebCore::TimeRanges::Range::isBeforeRange): Ditto.

2009-04-17 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Simon Fraser.

Add test for media element 'played' attribute.

  • media/video-played-expected.txt: Added.
  • media/video-played.html: Added.
  • media/video-test.js: (disableFullTestDetailsPrinting): New, include less information in 'reportExpected' output to accomodate tests with values that are expected to change from run to run. (reportExpected): Don't log expected and observed values when printFullTestDetails is false. (runSilently): New, same as run() but without logging. (run): Call runSilently to run command.
12:16 Changeset [42618] by darin@apple.com

WebCore:

2009-04-17 Darin Adler <darin@apple.com>

Reviewed by Antti Koivisto.

Bug 25210: don't use ObjC methods to wrap/unwrap DOM objects with ObjC
https://bugs.webkit.org/show_bug.cgi?id=25210

  • WebCore.base.exp: Export the core and kit function overloads that are used in WebKit. Also resorted this file.
  • WebCore.xcodeproj/project.pbxproj: Re-added DOMHTMLCanvasElement.h to fix a strange problem with the file type.
  • bindings/objc/DOM.mm: Removed the many extraneous includes. Added new includes so we can use and implement the new core and kit functions. Import DOMInternal.h first so it can do its public/private magic. (-[DOMNode _rootObject]): Changed to use early-return style. (kitClass): Added. Tells the wrapper generator what kind of node class to create to wrap DOM nodes. This has the code that used to be in the _wrapNode method. (kit): Added. Takes an EventTarget and makes the appropriate type of wrapper, dpending on whether it's a Node or an SVGElementInstance. This replace methods that used to be on both of those classes. (-[DOMNode boundingBox]): Changed to use early return, and the core function. (-[DOMNode lineBoxRects]): Ditto. (-[DOMNode renderedImage]): Ditto. (-[DOMRange boundingBox]): Ditto. (-[DOMRange lineBoxRects]): Ditto. (-[DOMElement image]): Ditto. (-[DOMElement _font]): Ditto. (-[DOMElement _imageTIFFRepresentation]): Ditto. (-[DOMElement _getURLAttribute:]): Ditto. (-[DOMElement isFocused]): Ditto. (kit): Hand-wrote the version of this for NodeFilter, since it's a protocol, not a class in the binding. (core): Ditto. (-[DOMNodeFilter acceptNode:]): Use core.
  • bindings/objc/DOMAbstractView.mm: (-[DOMAbstractView document]): Use kit. (-[DOMAbstractView _disconnectFrame]): Call removeDOMWrapper, which is no longer in the WebCore namespace, since it's headed for WebKit in the future. (core): Hand-wrote the version of this for DOMAbstractView.mm, since we store a pointer to the Frame, not the DMWindow. (kit): Ditto.
  • bindings/objc/DOMCSS.mm: (kitClass): Added. Tells the wrapper generator what kind of node class to create to wrap CSS objects. This has the code that used to be in the _wrapStyleSheet, _wrapCSSRule, and _wrapCSSValue methods.
  • bindings/objc/DOMEvents.mm: (kitClass): Added. Tells the wrapper generator what kind of node class to create to wrap event objects. This has the code that used to be in the _wrapEvent methods.
  • bindings/objc/DOMHTML.mm: (-[DOMHTMLDocument createDocumentFragmentWithMarkupString:baseURL:]): Use kit and core. (-[DOMHTMLDocument createDocumentFragmentWithText:]): Ditto. (-[DOMHTMLDocument _createDocumentFragmentWithMarkupString:baseURLString:]): Ditto. (-[DOMHTMLInputElement _rectOnScreen]): Ditto. (-[DOMHTMLInputElement _replaceCharactersInRange:withString:selectingFromIndex:]): Ditto. (-[DOMHTMLInputElement _selectedRange]): Ditto. (-[DOMHTMLInputElement _setAutofilled:]): Ditto. (-[DOMHTMLSelectElement _activateItemAtIndex:]): Ditto. (-[DOMHTMLInputElement _isEdited]): Ditto. (-[DOMHTMLTextAreaElement _isEdited]): Ditto. (kitClass): Added. Gives HTMLOptionCollection objects an appropriate wrapper.
  • bindings/objc/DOMInternal.h: Removed most of the imports. There's no reason to have this header include all the other internal DOM headers. Removed unneeded interfaces, including the redeclaration of _init in DOMObject, since it's already inherited from WebScriptObject. Moved all the functions out of the WebCore namespace since this code is destined for WebKit, which doesn't use a namespace. Added kit and core functions for the two classes that live in this header.
  • bindings/objc/DOMInternal.mm: (-[WebScriptObject _initializeScriptDOMNodeImp]): Use core.
  • bindings/objc/DOMObject.h: Made the _internal field be private instead of protected. This also allows the "#define private public" trick to be used to make the field accessible inside the bindings code.
  • bindings/objc/DOMObject.mm: (-[DOMObject dealloc]): Updated for namespace change. (-[DOMObject sheet]): Use core and kit functions.
  • bindings/objc/DOMRGBColor.mm: (-[DOMRGBColor dealloc]): Updated for namespace change. (-[DOMRGBColor red]): Use kit function. (-[DOMRGBColor green]): Ditto. (-[DOMRGBColor blue]): Ditto. (-[DOMRGBColor alpha]): Ditto. (core): Added. Replaces the _RGBColor method. (kit): Added. Replaces the _wrapRGBColor method.
  • bindings/objc/DOMSVGPathSegInternal.mm: (kitClass): Added. Replaces the _wrapSVGPathSeg method.
  • bindings/objc/DOMUtility.mm: (JSC::createDOMWrapper): Use kit function. (createDOMWrapper): Ditto.
  • bindings/objc/DOMXPath.mm: (core): Added. Replaces the _xpathNSResolver method. (kit): Added. Replaces the _wrapXPathNSResolver method.
  • bindings/objc/ObjCEventListener.mm: (WebCore::ObjCEventListener::find): Changed to use early return. (WebCore::ObjCEventListener::handleEvent): Use kit function.
  • bindings/objc/ObjCNodeFilterCondition.mm: (WebCore::ObjCNodeFilterCondition::acceptNode): Use kit function.
  • bindings/objc/WebScriptObject.mm: (+[WebScriptObject scriptObjectForJSObject:originRootObject:rootObject:]): Updated for namespace change.
  • bindings/scripts/CodeGeneratorObjC.pm: Changed spelling of "license" to U.S. spelling. Removed GetObjCTypeMaker, GetObjCTypeGetterName, GetInternalTypeGetterSignature, and GetInternalTypeMakerSignature. Changed includes to conform to new scheme. Generate core and kit functions instead of methods and calls to those functions. Added a new attribute, Polymorphic, to indicate classes that have derived subclasses. Removed the old ObjCCustomInternalImpl attribute.
  • css/CSSRule.idl: Use Polymorphic attribute, so core and kit functions will be generated, with kit calling kitClass. This requires less hand- written code than the old ObjCCustomInternalImpl.
  • css/CSSValue.idl: Ditto.
  • css/StyleSheet.idl: Ditto.
  • dom/Event.idl: Ditto.
  • dom/Node.idl: Ditto.
  • html/HTMLCollection.idl: Ditto.
  • svg/SVGPathSeg.idl: Ditto.
  • editing/TextAffinity.h: Moved kit and core functions out of the WebCore namespace since this code is destined for WebKit, which doesn't use a namespace
  • platform/mac/ClipboardMac.mm: (WebCore::ClipboardMac::declareAndWriteDragImage): Use kit function.
  • platform/mac/DragDataMac.mm: (WebCore::DragData::asFragment): Use core function.
  • platform/mac/PasteboardMac.mm: (WebCore::Pasteboard::writeSelection): Use kit function.
  • svg/SVGElementInstance.idl: Removed now-unneeded ObjCCustomInternalImpl function. This can be generated in a standard way now.
  • svg/SVGViewSpec.idl: Ditto.

WebKit/mac:

2009-04-17 Darin Adler <darin@apple.com>

Reviewed by Antti Koivisto.

Bug 25210: don't use ObjC methods to wrap/unwrap DOM objects with ObjC
https://bugs.webkit.org/show_bug.cgi?id=25210

  • DOM/WebDOMOperations.mm: (-[DOMNode markupString]): Use the core function instead of an Objective-C method. (-[DOMNode _subresourceURLs]): Ditto. (-[DOMDocument _focusableNodes]): Ditto. (-[DOMRange webArchive]): Ditto. (-[DOMRange markupString]): Ditto.
  • Misc/WebElementDictionary.mm: Added now-needed include since the core and kit functions now come from the internal headers from DOM bindings.
  • Misc/WebNSPasteboardExtras.mm: Ditto.
  • Plugins/WebNullPluginView.mm: Ditto.
  • Plugins/WebPluginController.mm: Ditto.
  • WebCoreSupport/WebChromeClient.mm: Ditto.
  • WebCoreSupport/WebInspectorClient.mm: Ditto.
  • WebCoreSupport/WebPasteboardHelper.mm: Ditto.
  • WebView/WebHTMLView.mm: Ditto.
  • WebCoreSupport/WebEditorClient.mm: Made kit function have internal linkage since it's only used in this file. Someone had instead added a declaration to suppress the warning you would otherwise get. Removed the core function. (WebEditorClient::textFieldDidBeginEditing): Added correct type checking. Previously the function would check only that something was an HTMLElement, but then cast it to HTMLInputElement. Also call kit instead of old wrap method. (WebEditorClient::textFieldDidEndEditing): Ditto. (WebEditorClient::textDidChangeInTextField): Ditto. (WebEditorClient::doTextFieldCommandFromEvent): Ditto. (WebEditorClient::textWillBeDeletedInTextField): Ditto. (WebEditorClient::textDidChangeInTextArea): Ditto, but for HTMLTextAreaElement.
  • WebView/WebFrame.mm: Removed the core and kit functions here which are no longer needed since they're automatically generated now. (-[WebFrame _nodesFromList:]): Use kit. (-[WebFrame _markupStringFromRange:nodes:]): Use core. (-[WebFrame _stringForRange:]): More of the same. (-[WebFrame _caretRectAtNode:offset:affinity:]): Ditto. (-[WebFrame _firstRectForDOMRange:]): Ditto. (-[WebFrame _scrollDOMRangeToVisible:]): Ditto. (-[WebFrame _rangeByAlteringCurrentSelection:SelectionController::direction:SelectionController::granularity:]): Ditto. (-[WebFrame _convertNSRangeToDOMRange:]): Ditto. (-[WebFrame _convertDOMRangeToNSRange:]): Ditto. (-[WebFrame _markDOMRange]): Ditto. (-[WebFrame _smartDeleteRangeForProposedRange:]): Ditto. (-[WebFrame _smartInsertForString:replacingRange:beforeString:afterString:]): Ditto. (-[WebFrame _documentFragmentWithMarkupString:baseURLString:]): Ditto. (-[WebFrame _documentFragmentWithNodesAsParagraphs:]): Ditto. (-[WebFrame _replaceSelectionWithNode:selectReplacement:smartReplace:matchStyle:]): Ditto. (-[WebFrame _characterRangeAtPoint:]): Ditto. (-[WebFrame _typingStyle]): Ditto. (-[WebFrame _setTypingStyle:withUndoAction:]): Ditto. (-[WebFrame _pauseAnimation:onNode:atTime:]): Ditto. (-[WebFrame _pauseTransitionOfProperty:onNode:atTime:]): Ditto. (-[WebFrame _replaceSelectionWithFragment:selectReplacement:smartReplace:matchStyle:]): Ditto.
  • WebView/WebFrameInternal.h: Removed the core and kit functions here which are no longer needed since they're automatically generated now.
  • WebView/WebHTMLRepresentation.mm: (-[WebHTMLRepresentation attributedStringFrom:startOffset:to:endOffset:]): Use core. (formElementFromDOMElement): Ditto. (inputElementFromDOMElement): Ditto.
  • WebView/WebTextIterator.mm: (-[WebTextIterator initWithRange:]): Ditto. (-[WebTextIterator currentRange]): Ditto. (-[WebTextIterator currentNode]): Ditto.
  • WebView/WebView.mm: (-[WebView textIteratorForRect:]): Ditto. (-[WebView setSelectedDOMRange:affinity:]): Ditto.
11:46 Changeset [42617] by kov@webkit.org

2009-04-17 Gustavo Noronha Silva <gustavo.noronha@collabora.co.uk>

Reviewed by Alexey Proskuryakov.

https://bugs.webkit.org/show_bug.cgi?id=25251
Crash when frame creation removes it from the parent

This brings us up-to-date with the expectations of sub-frame
creation, following what Mac has done. It fixes crashes when
calling init on the child frame removes it from the page.

This makes the following test pass instead of crash:

fast/loader/frame-creation-removal.html

  • WebCoreSupport/FrameLoaderClientGtk.cpp: (WebKit::FrameLoaderClient::createFrame):
11:03 Changeset [42616] by beidson@apple.com

WebCore:

2009-04-17 Brady Eidson <beidson@apple.com>

Reviewed by Maciej Stachowiak

<rdar://problem/5753299> "Private Browsing" doesn't affect HTML 5 Database storage

If the Page is in private browsing mode when a statement is queued up, flag that
statement as read-only.

It would be better to set the read-only flag on the database thread when it actually
executes, but that would either require making Settings access a lock-protected operation
or involve a synchronous callback to the WebCore thread, neither of which seem worth it.
If it becomes an issue in practice, we can revisit this later.

Test: storage/private-browsing-readonly.html

  • storage/Database.cpp: (WebCore::Database::setAuthorizerReadOnly):
  • storage/Database.h:
  • storage/DatabaseAuthorizer.cpp: For all action types that would alter the database, deny when in read-only mode. (WebCore::DatabaseAuthorizer::reset): (WebCore::DatabaseAuthorizer::createTable): (WebCore::DatabaseAuthorizer::dropTable): (WebCore::DatabaseAuthorizer::allowAlterTable): (WebCore::DatabaseAuthorizer::createIndex): (WebCore::DatabaseAuthorizer::dropIndex): (WebCore::DatabaseAuthorizer::createTrigger): (WebCore::DatabaseAuthorizer::dropTrigger): (WebCore::DatabaseAuthorizer::createVTable): (WebCore::DatabaseAuthorizer::dropVTable): (WebCore::DatabaseAuthorizer::allowDelete): (WebCore::DatabaseAuthorizer::allowInsert): (WebCore::DatabaseAuthorizer::allowUpdate): (WebCore::DatabaseAuthorizer::setReadOnly):
  • storage/DatabaseAuthorizer.h:
  • storage/SQLStatement.cpp: (WebCore::SQLStatement::create): Add a m_readOnly flag. (WebCore::SQLStatement::SQLStatement): Ditto. (WebCore::SQLStatement::execute): If m_readOnly is set, tell the authorizer to change to read-only mode.
  • storage/SQLStatement.h:
  • storage/SQLTransaction.cpp: (WebCore::SQLTransaction::executeSQL): Flag the statement as read-only if the Page is currently in private browsing mode.

LayoutTests:

2009-04-17 Brady Eidson <beidson@apple.com>

Reviewed by Maciej Stachowiak

<rdar://problem/5753299> "Private Browsing" doesn't affect HTML 5 Database storage

  • storage/private-browsing-readonly-expected.txt: Added.
  • storage/private-browsing-readonly.html: Added.
10:59 QtWebKitTodo edited by ariya@webkit.org
(diff)
10:46 Changeset [42615] by andersca@apple.com

2009-04-17 Anders Carlsson <andersca@apple.com>

Reviewed by Sam Weinig.

WebKit side of <rdar://problem/6449642>.


  • Plugins/Hosted/HostedNetscapePluginStream.h: (WebKit::HostedNetscapePluginStream::create): New function that creates a stream from a frame loader.


  • Plugins/Hosted/HostedNetscapePluginStream.mm: (WebKit::HostedNetscapePluginStream::HostedNetscapePluginStream): Add the constructor that takes a frame loader.


  • Plugins/Hosted/NetscapePluginHostManager.h:
  • Plugins/Hosted/NetscapePluginHostManager.mm: (WebKit::NetscapePluginHostManager::instantiatePlugin): Pass "fullFrame" to the plug-in host.


  • Plugins/Hosted/NetscapePluginHostProxy.mm: (WKPCCancelLoadURL): Call NetscapePluginInstanceProxy::cancelStreamLoad.


  • Plugins/Hosted/NetscapePluginInstanceProxy.h: (WebKit::NetscapePluginInstanceProxy::create): Pass "fullFrame" to the constructor.


(WebKit::NetscapePluginInstanceProxy::manualStream):
New getter for the manual stream.


  • Plugins/Hosted/NetscapePluginInstanceProxy.mm: (WebKit::NetscapePluginInstanceProxy::NetscapePluginInstanceProxy): Take the implicit request into account if we have a full frame plug-in.


(WebKit::NetscapePluginInstanceProxy::setManualStream):
Setter for the manual stream.


(WebKit::NetscapePluginInstanceProxy::cancelStreamLoad):
Cancel the manual stream if necessary.


  • Plugins/Hosted/WebHostedNetscapePluginView.h: WebHostedNetscapePluginView now conforms to the WebPluginManualLoader protocol.


  • Plugins/Hosted/WebHostedNetscapePluginView.mm: (-[WebHostedNetscapePluginView createPlugin]): Pass "fullFrame" to instantiatePlugin.


(-[WebHostedNetscapePluginView pluginView:receivedResponse:]):
(-[WebHostedNetscapePluginView pluginView:receivedData:]):
(-[WebHostedNetscapePluginView pluginView:receivedError:]):
(-[WebHostedNetscapePluginView pluginViewFinishedLoading:]):
Call the equivalent manual stream functions.


  • WebCoreSupport/WebFrameLoaderClient.mm: (WebFrameLoaderClient::createPlugin): Use a macro for getting the plug-in view type.
10:44 Changeset [42614] by adachan@apple.com

Fix a leak in WebHistory::data().

Reviewed by Adam Roben.

  • WebHistory.cpp: (WebHistory::data):
10:42 Changeset [42613] by eric@webkit.org

Reviewed by Antti Koivisto.

Move RenderPath and RenderSVGContainer onto a unified clippedOverflowRectForRepaint
https://bugs.webkit.org/show_bug.cgi?id=25268

Lots of minus lines. Now we're sharing more sane code
(which will respect -webkit-transforms! and scroll offsets correctly)

Which means this fixes:
https://bugs.webkit.org/show_bug.cgi?id=20769 too!

We're no longer expanding the paint rect "for anti-aliasing", since
I can't find a case where that's required. If it is, repaintRectInLocalCoordinates()
should be fixed to handle those cases instead of here.

This fixes svg/custom/scroll-hit-test (now that we respect scroll offsets when repainting)
as well as improves our focus ring drawing seen in svg/custom/focus-ring
focus rings are now closer to transformed content by a couple pixels (they were needlessly outset by the antialiasing hack)
Also, it fixes the dumped rects for markers, causing a progression in svg/custom/marker-overflow-clip

  • rendering/RenderPath.cpp: (WebCore::RenderPath::repaintRectInLocalCoordinates): (WebCore::RenderPath::setPath):
  • rendering/RenderPath.h:
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::repaintRectInLocalCoordinates):
  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGModelObject.cpp: (WebCore::RenderSVGModelObject::clippedOverflowRectForRepaint): (WebCore::RenderSVGModelObject::computeRectForRepaint):
  • rendering/RenderSVGModelObject.h:
  • rendering/RenderSVGRoot.cpp: (WebCore::RenderSVGRoot::computeRectForRepaint):
  • rendering/RenderSVGRoot.h:
09:53 Changeset [42612] by mitz@apple.com

Rubber-stamped by Anders Carlsson.

  • remove tests affected by <rdar://problem/6661401> from the skip list, now that it is fixed.
  • platform/mac-snowleopard/Skipped:
09:00 Changeset [42611] by cfleizach@apple.com

Bug 25259: AX: VoiceOver cmd-cntl-space does not follow the google.com->more link
https://bugs.webkit.org/show_bug.cgi?id=25259

When a node becomes visible or hidden, accessibility needs to be told to update.

08:52 Changeset [42610] by kevino@webkit.org

wx build fix, added missing header.

02:20 Changeset [42609] by hausmann@webkit.org

2009-04-17 Simon Hausmann <simon.hausmann@nokia.com>

Unreviewed build fix.

Fix the build with enabled SVG filters

01:49 Changeset [42608] by hausmann@webkit.org

2009-04-17 Laszlo Gombos <laszlo.1.gombos@nokia.com>

Reviewed by Simon Hausmann.

https://bugs.webkit.org/show_bug.cgi?id=25242

Remove QtCore 4.5 dependency from QWebElement test

00:34 Changeset [42607] by oliver@apple.com

Bug 25027: JavaScript parseInt wrong on negative numbers
<https://bugs.webkit.org/show_bug.cgi?id=25027>

Contributed by Mihnea Ovidenie
Reviewed by Oliver Hunt

When dealing with negative numbers, parseInt should use ceil instead of floor.

04/16/09:

23:20 Changeset [42606] by slewis@apple.com

2009-04-16 Stephanie Lewis <slewis@apple.com>

Reviewed by Oliver Hunt.

<rdar://problem/6744652> 32-bit to 64-bit: Javascript hash tables double in size

Remove perfect hash optimization which removes 1 MB of overhead on 32-bit and almost 2 MB on 64-bit. Removing the optimization was not a regression on SunSpider and the acid 3 test still passes.

  • create_hash_table:
  • runtime/Lookup.cpp: (JSC::HashTable::createTable): (JSC::HashTable::deleteTable):
  • runtime/Lookup.h: (JSC::HashEntry::initialize): (JSC::HashEntry::next): (JSC::HashTable::entry):
  • runtime/Structure.cpp: (JSC::Structure::getEnumerableNamesFromClassInfoTable):
21:41 Changeset [42605] by oliver@apple.com

Fix subtle error in optimised VM reentry in Array.sort

Reviewed by Gavin Barraclough

Basically to ensure we don't accidentally invalidate the cached callframe
we should be using the cached callframe rather than our own exec state.
While the old behaviour was wrong i have been unable to actually create a
test case where anything actually ends up going wrong.

21:23 Changeset [42604] by dglazkov@chromium.org

2009-04-16 Eric Roman <eroman@chromium.org>

Reviewed by Dimitri Glazkov.

https://bugs.webkit.org/show_bug.cgi?id=25196
Add a missing V8DOMMap include.

  • bindings/v8/WorkerScriptController.cpp: Included V8DOMMap.h.
21:19 Changeset [42603] by dglazkov@chromium.org

2009-04-16 Dimitri Glazkov <dglazkov@chromium.org>

Reviewed by Darin Fisher.

https://bugs.webkit.org/show_bug.cgi?id=25260
Finish V8 custom bindings for InspectorController.

  • bindings/v8/custom/V8InspectorControllerCustom.cpp: (WebCore::CALLBACK_FUNC_DECL): Added.
20:10 Changeset [42602] by alice.liu@apple.com

2009-04-16 Alice Liu <alice.liu@apple.com>

Reviewed by Mark Rowe.

Fix for heavy leakage of WebURLResponses seen during the stress test

  • WebCoreSupport/WebFrameLoaderClient.cpp: (WebFrameLoaderClient::shouldCacheResponse):
18:36 Changeset [42601] by kevino@webkit.org

Reviewed by Kevin Ollivier.

Switch to wxGraphicsBitmap when using wxGraphicsContext so that we're not doing unnecessary conversions internally when drawing bitmaps.

https://bugs.webkit.org/show_bug.cgi?id=25248

18:23 Changeset [42600] by justin.garcia@apple.com

WebCore:

2009-04-16 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.

Performing a block style change in an editable root that contains only a br adds a newline
https://bugs.webkit.org/show_bug.cgi?id=25256

  • editing/CompositeEditCommand.cpp: (WebCore::CompositeEditCommand::moveParagraphContentsToNewBlockIfNecessary): Fix a bad check for an root block that didn't handle a root that contained a placeholder.

LayoutTests:

2009-04-16 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.


Performing a block style change in an editable root that contains only a br adds a newline
https://bugs.webkit.org/show_bug.cgi?id=25256

  • editing/execCommand/25256-expected.txt: Added.
  • editing/execCommand/25256.html: Added.
17:34 Changeset [42599] by eric@webkit.org

Reviewed by Simon Fraser.

Fix computeContainerBoundingBox to respect viewport translations
https://bugs.webkit.org/show_bug.cgi?id=25254

Previously computeContainerBoundingBox was just adding the viewportTranslation.
Which, although it has a confusing name, only contains the scale and translation
for the viewBox/preserveAspectRatio adjustment. It does not contain the
translation for the x/y offset of the viewport.
localToParentTransform() does contain this offset, so we use that instead
of the previous hacky code.

  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGViewportContainer.h:
  • rendering/SVGRenderSupport.cpp: (WebCore::computeContainerBoundingBox):
16:27 Changeset [42598] by eric@webkit.org

Reviewed by Simon Fraser.

Add a localToParentTransform() method which includes the
full transform between local and parent local coordinates.
https://bugs.webkit.org/show_bug.cgi?id=25226

Moved m_localTransform up into RenderSVGTransformableContainer
and added a comment to RenderSVGViewportContainer to make it more
clear that it does not have a localTransform().

This patch reveals two more failed design decisions:

  1. Use of RenderBox::absoluteTransform() to mean "absoluteTransform()

including only my localTransform()" callers are probably using it
incorrectly anyway and are just masking bugs in the confused code.

  1. computeContainerBoundingBox does not include viewport translations in

its computed bounding box, so bounding boxes will be off for parents of
inner <svg> elements. I'll fix this an updated the results in a separate change.

No functional changes, thus no tests.

  • rendering/RenderObject.cpp: (WebCore::RenderObject::localToParentTransform): (WebCore::RenderObject::absoluteTransform):
  • rendering/RenderObject.h:
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::layout):
  • rendering/RenderSVGContainer.h: (WebCore::RenderSVGContainer::viewportTransform):
  • rendering/RenderSVGHiddenContainer.cpp:
  • rendering/RenderSVGHiddenContainer.h: (WebCore::RenderSVGHiddenContainer::absoluteTransform):
  • rendering/RenderSVGRoot.cpp: (WebCore::RenderSVGRoot::localToParentTransform): (WebCore::RenderSVGRoot::absoluteTransform):
  • rendering/RenderSVGRoot.h:
  • rendering/RenderSVGTransformableContainer.cpp: (WebCore::RenderSVGTransformableContainer::localToParentTransform): (WebCore::RenderSVGTransformableContainer::localTransform):
  • rendering/RenderSVGTransformableContainer.h:
  • rendering/RenderSVGViewportContainer.cpp: (WebCore::RenderSVGViewportContainer::localToParentTransform): (WebCore::RenderSVGViewportContainer::absoluteTransform):
  • rendering/RenderSVGViewportContainer.h:
  • rendering/SVGRenderSupport.cpp: (WebCore::computeContainerBoundingBox):
15:29 Changeset [42597] by darin@chromium.org

2009-04-16 Adam Langley <agl@google.com>

Reviewed by Darin Fisher.

Fix Chromium's rendering of <option> elements inside of <optgroups>.

https://bugs.webkit.org/show_bug.cgi?id=25244

Test (will included as a manual test by https://bugs.webkit.org/show_bug.cgi?id=24900):

<select>

<optgroup label="Should be bold">

<option>Should not be bold</option>

</optgroup>

</select>

  • css/themeWin.css: adding comments
  • platform/chromium/PopupMenuChromium.cpp: (WebCore::PopupListBox::getRowFont): use menuStyle() rather than the item's style
15:25 Changeset [42596] by zecke@webkit.org

2009-04-16 Holger Hans Peter Freyther <zecke@selfish.org>

Rubber-stamped by Simon Hausmann.

Make the operator== and operator!= non inline to
allow looking at the d-pointer in the future.

  • Api/qwebelement.cpp: (QWebElement::operator==): (QWebElement::operator!=):
  • Api/qwebelement.h:
15:17 Changeset [42595] by xan@webkit.org

2009-04-16 Xan Lopez <xlopez@igalia.com>

Unreviewed build fix.

  • WebView.cpp:
15:13 Changeset [42594] by ggaren@apple.com

2009-04-16 Geoffrey Garen <ggaren@apple.com>

Build fix: updated check-for-global-initializers to account for code motion.

  • Scripts/check-for-global-initializers:
15:12 Changeset [42593] by darin@chromium.org

2009-04-16 Eric Roman <eroman@chromium.org>

Reviewed by Alexey Proskuryakov.

https://bugs.webkit.org/show_bug.cgi?id=25218

Make DOMWindow::inlineEventListenerForType not have the "inline"
keyword. As this method gets used outside of DOMWindow.

  • page/DOMWindow.cpp: (WebCore::DOMWindow::inlineEventListenerForType):
15:02 Changeset [42592] by bolsinga@apple.com

2009-04-16 Greg Bolsinga <bolsinga@apple.com>

Reviewed by Antti Koivisto.

Fix <rdar://problem/6766969>


When deferred repaints are enabled and being reset and there is a active
timer, stop the timer and do the deferred repaints immediately.

  • page/FrameView.cpp: (WebCore::FrameView::resetDeferredRepaintDelay):
15:01 Changeset [42591] by christian@webkit.org

2009-04-16 Christian Dywan <christian@twotoasts.de>

Reviewed by Holger Freyther.

http://bugs.webkit.org/show_bug.cgi?id=25042
destroying webview widget directly causes crash.

  • tests/testwebframe.c: (test_webkit_web_frame_create_destroy): Add a test case for destroying a web view shortly after packing it in a window.
14:53 Changeset [42590] by xan@webkit.org

WebCore:

2009-04-16 Xan Lopez <xlopez@igalia.com>

Reviewed by Sam Weinig.

https://bugs.webkit.org/show_bug.cgi?id=25241
Remove superfluous NotImplemented.h includes.

  • html/CanvasRenderingContext2D.cpp:
  • html/CanvasStyle.cpp:
  • page/AccessibilityRenderObject.cpp:
  • platform/chromium/ScrollbarThemeChromiumLinux.cpp:
  • platform/graphics/GraphicsContext.cpp:
  • platform/graphics/cairo/ImageBufferCairo.cpp:
  • platform/graphics/cairo/PathCairo.cpp:
  • platform/graphics/chromium/SimpleFontDataLinux.cpp:
  • platform/graphics/gtk/FontGtk.cpp:
  • platform/graphics/gtk/IconGtk.cpp:
  • platform/graphics/qt/IconQt.cpp:
  • platform/graphics/qt/ImageQt.cpp:
  • platform/graphics/qt/ImageSourceQt.cpp:
  • platform/graphics/skia/ImageBufferSkia.cpp:
  • platform/graphics/skia/ImageSkia.cpp:
  • platform/graphics/win/ColorSafari.cpp:
  • platform/graphics/win/FontWin.cpp:
  • platform/graphics/win/GraphicsContextCGWin.cpp:
  • platform/graphics/win/GraphicsContextCairoWin.cpp:
  • platform/graphics/win/SimpleFontDataCairoWin.cpp:
  • platform/graphics/wx/ImageWx.cpp:
  • platform/graphics/wx/TransformationMatrixWx.cpp:
  • platform/gtk/PlatformScreenGtk.cpp:
  • platform/gtk/PopupMenuGtk.cpp:
  • platform/gtk/ScrollViewGtk.cpp:
  • platform/gtk/ScrollbarGtk.cpp:
  • platform/network/cf/ResourceHandleCFNet.cpp:
  • platform/posix/FileSystemPOSIX.cpp:
  • platform/qt/DragDataQt.cpp:
  • platform/qt/FileSystemQt.cpp:
  • platform/qt/Localizations.cpp:
  • platform/qt/PopupMenuQt.cpp:
  • platform/win/EditorWin.cpp:
  • platform/win/PasteboardWin.cpp:
  • platform/wx/KeyboardEventWx.cpp:
  • platform/wx/PopupMenuWx.cpp:
  • platform/wx/SharedTimerWx.cpp:
  • plugins/gtk/PluginViewGtk.cpp:
  • plugins/mac/PluginPackageMac.cpp:
  • plugins/qt/PluginPackageQt.cpp:
  • plugins/win/PluginViewWin.cpp:
  • rendering/RenderThemeChromiumLinux.cpp:
  • svg/graphics/SVGImage.cpp:

WebKit/gtk:

2009-04-16 Xan Lopez <xlopez@igalia.com>

Reviewed by Sam Weinig.

https://bugs.webkit.org/show_bug.cgi?id=25241
Remove superfluous NotImplemented.h includes.

  • webkit/webkitprivate.cpp:

WebKit/win:

2009-04-16 Xan Lopez <xlopez@igalia.com>

Reviewed by Sam Weinig.

https://bugs.webkit.org/show_bug.cgi?id=25241
Remove superfluous NotImplemented.h includes

  • WebDownloadCFNet.cpp:
  • WebURLAuthenticationChallengeSender.cpp:
  • WebView.cpp:
14:19 Changeset [42589] by ggaren@apple.com

2009-04-16 Geoffrey Garen <ggaren@apple.com>

Reviewed by Sam Weinig.

More fix for https://bugs.webkit.org/show_bug.cgi?id=21260
Unbounded memory growth when churning elements with anonymous event handler functions


Some cleanup in event handling code.

  • bindings/js/JSDOMGlobalObject.cpp:
  • bindings/js/JSDOMGlobalObject.h: Removed findJSProtectedEventListener and findJSProtectedEventListener because they are now unused.
  • bindings/js/JSEventListener.cpp:
  • bindings/js/JSEventListener.h: (WebCore::JSEventListener::clearGlobalObject):
  • bindings/js/JSLazyEventListener.cpp: (WebCore::JSLazyEventListener::JSLazyEventListener): (WebCore::JSLazyEventListener::~JSLazyEventListener): (WebCore::JSLazyEventListener::globalObject):
  • bindings/js/JSLazyEventListener.h: (WebCore::JSLazyEventListener::clearGlobalObject): Nixed JSProtectedEventListener, and merged it with JSLazyEventListener, the only remaining event listener that still GC-protects its function and global object.
  • bindings/js/ScriptController.cpp: (WebCore::ScriptController::ScriptController):
  • bindings/js/ScriptController.h: (WebCore::ScriptController::setEventHandlerLineNumber): (WebCore::ScriptController::eventHandlerLineNumber): Renamed handlerLineno => handlerLineNumber, because anything less would be uncivilized. Removed createInlineEventListener because it mostly did irrelevent work, so it just got in the way of understanding how event handler creation works.
  • dom/Document.cpp: (WebCore::Document::createEventListener):
  • dom/XMLTokenizerLibxml2.cpp: (WebCore::XMLTokenizer::startElementNs):
  • html/HTMLTokenizer.cpp: (WebCore::HTMLTokenizer::processToken): Ditto.
14:19 Changeset [42588] by justin.garcia@apple.com

Checking in pixel results for a recently added test.

13:58 Changeset [42587] by bdakin@apple.com

2009-04-16 Beth Dakin <bdakin@apple.com>

Reviewed by Dave Hyatt.

Fix for https://bugs.webkit.org/show_bug.cgi?id=20765 Website
crashes on load due to messy HTML in search form
-and corresponding-
<rdar://problem/6210633>

Before this patch, <form>s were allowed to sit inside tables
without being wrapped by anonymous table parts. There was also a
concept that such a form could be "demoted" and would not be
allowed to have any children. This patch has the HTML parser mark
form elements that have been demoted as such, and then the demoted
forms are not given renderers. I also removed the code that allowed
forms to sit in tables without anonymous table sections. So now any
forms that do manage to get a renderer inside a table will also be
wrapped with appropriate table parts.

  • html/HTMLFormElement.cpp: (WebCore::HTMLFormElement::HTMLFormElement): (WebCore::HTMLFormElement::rendererIsNeeded):
  • html/HTMLFormElement.h: (WebCore::HTMLFormElement::setDemoted): (WebCore::HTMLFormElement::isDemoted):
  • html/HTMLParser.cpp: (WebCore::HTMLParser::insertNode):
  • rendering/RenderTable.cpp: (WebCore::RenderTable::addChild):
  • rendering/RenderTableRow.cpp: (WebCore::RenderTableRow::addChild):
  • rendering/RenderTableSection.cpp: (WebCore::RenderTableSection::addChild):
13:48 Changeset [42586] by bdakin@apple.com

WebCore:

2009-04-16 Beth Dakin <bdakin@apple.com>

Reviewed by Dave Hyatt.

Fix for https://bugs.webkit.org/show_bug.cgi?id=20765 Website
crashes on load due to messy HTML in search form
-and corresponding-
<rdar://problem/6210633>

Before this patch, <form>s were allowed to sit inside tables
without being wrapped by anonymous table parts. There was also a
concept that such a from could be "demoted" and would not be
allowed to have any children. This patch has the HTML parser mark
form elements that have been demoted as such, and then the demoted
forms are not given renderers. I also removed the code that allowed
forms to sit in tables without anonymous table sections. So now any
forms that do manage to get a renderer inside a table will also be
wrapped with appropriate table parts.

  • html/HTMLFormElement.cpp: (WebCore::HTMLFormElement::HTMLFormElement): (WebCore::HTMLFormElement::rendererIsNeeded):
  • html/HTMLFormElement.h: (WebCore::HTMLFormElement::setDemoted): (WebCore::HTMLFormElement::isDemoted):
  • html/HTMLParser.cpp: (WebCore::HTMLParser::insertNode):
  • rendering/RenderTable.cpp: (WebCore::RenderTable::addChild):
  • rendering/RenderTableRow.cpp: (WebCore::RenderTableRow::addChild):
  • rendering/RenderTableSection.cpp: (WebCore::RenderTableSection::addChild):

LayoutTests:

2009-04-16 Beth Dakin <bdakin@apple.com>

Reviewed by Dave Hyatt.

Tests for https://bugs.webkit.org/show_bug.cgi?id=20765 Website
crashes on load due to messy HTML in search form
-and corresponding-
<rdar://problem/6210633>

New tests:

  • fast/forms/form-added-to-table.html: Added.
  • fast/forms/form-in-malformed-markup.html: Added.
  • platform/mac/fast/forms/form-added-to-table-expected.checksum: Added.
  • platform/mac/fast/forms/form-added-to-table-expected.png: Added.
  • platform/mac/fast/forms/form-added-to-table-expected.txt: Added.
  • platform/mac/fast/forms/form-in-malformed-markup-expected.checksum: Added.
  • platform/mac/fast/forms/form-in-malformed-markup-expected.png: Added.
  • platform/mac/fast/forms/form-in-malformed-markup-expected.txt: Added.

Tests with new results:

  • platform/mac/fast/forms/formmove3-expected.txt:
  • platform/mac/fast/forms/preserveFormDuringResidualStyle-expected.txt:
  • platform/mac/fast/invalid/table-residual-style-crash-expected.txt:
  • platform/mac/fast/table/fixed-table-non-cell-in-row-expected.txt:
  • platform/mac/fast/table/inline-form-assert-expected.txt:
  • platform/mac/fast/table/insert-cell-before-form-expected.txt:
  • platform/mac/fast/table/insert-row-before-form-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug1318-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug2516-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug30559-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug32447-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug34538-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug4527-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug78162-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug96334-expected.txt:
  • platform/mac/tables/mozilla/bugs/bug96343-expected.txt:
  • platform/mac/tables/mozilla_expected_failures/bugs/bug1725-expected.txt:
13:02 Changeset [42585] by darin@chromium.org

2009-04-16 Xiaomei Ji <xji@chromium.org>

Reviewed by Simon Fraser.

Fix https://bugs.webkit.org/show_bug.cgi?id=24527
caret does not paint the first time you focus on a 0px right padding RTL div

Test: fast/forms/caret-rtl.html

  • rendering/RenderBlock.cpp: (WebCore::RenderBlock::localCaretRect): When calculating x-axis if alignment is alignRight, we need to subtract the caretWidth so that the caret at IntRect(x, y, caretWidth, height) is inside the block.
11:41 Changeset [42584] by justin.garcia@apple.com

Moving this test to platform/mac, it's mac only.

11:11 Changeset [42583] by justin.garcia@apple.com

WebCore:

2009-04-16 Justin Garcia <justin.garcia@apple.com>

Reviewed by Simon Fraser.

https://bugs.webkit.org/show_bug.cgi?id=25228
SelectionController::absoluteCaretBounds returns an inflated caret (the caret repaint rect)


Return the bounds of the transformed caret, not the transformed repaint rect for the caret (which is inflated).

  • editing/SelectionController.cpp: (WebCore::SelectionController::localCaretRect): (WebCore::SelectionController::absoluteBoundsForLocalRect): (WebCore::SelectionController::caretRepaintRect): (WebCore::SelectionController::recomputeCaretRect):
  • editing/SelectionController.h:

LayoutTests:

2009-04-16 Justin Garcia <justin.garcia@apple.com>

Reviewed by Simon Fraser.


https://bugs.webkit.org/show_bug.cgi?id=25228
SelectionController::absoluteCaretBounds returns an inflated caret (the caret repaint rect)

  • editing/selection/25228.html: Added.
  • platform/mac/editing/selection/25228-expected.checksum: Added.
  • platform/mac/editing/selection/25228-expected.png: Added.
  • platform/mac/editing/selection/25228-expected.txt: Added.
11:11 Changeset [42582] by kov@webkit.org

2009-04-16 Zan Dobersek <zandobersek@gmail.com>

Reviewed by Adam Roben.

Compare test's response mime type and dump test as text properly.

  • DumpRenderTree/gtk/DumpRenderTree.cpp: (dump):
11:00 Changeset [42581] by eric.carlson@apple.com

2009-04-16 Pierre d'Herbemont <pdherbemont@apple.com>

Reviewed by Simon Fraser.

https://bugs.webkit.org/show_bug.cgi?id=25187
Bug 25187: <video> may not be rendered upon certain racy conditions

Always create the video renderer when the load state reaches QTMovieLoadStateLoaded
and the element is visible, not only when the movie has just been opened.

  • platform/graphics/mac/MediaPlayerPrivateQTKit.mm: (WebCore::MediaPlayerPrivate::updateStates): Make sure we do not make false assumption on the state changes order. Create the renderer if none is present, when movie is loaded instead.
10:24 Changeset [42580] by eric@webkit.org

Reviewed by Geoffrey Garen.

Add filterBoundingBox() to RenderSVGModelObject to share a tiny bit of code

RenderSVGModelObject has much more interesting code it can share,
but I'm just trying to get the commits out of my local branch and into
the repository in any order I can. :) This one was small.

This will be used by the unified RenderSVGModelObject
clippedOverflowRectForRepaint patch coming soon.

  • rendering/RenderPath.cpp: (WebCore::RenderPath::clippedOverflowRectForRepaint):
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::clippedOverflowRectForRepaint):
  • rendering/RenderSVGModelObject.cpp: (WebCore::RenderSVGModelObject::filterBoundingBox):
  • rendering/RenderSVGModelObject.h:
10:11 Changeset [42579] by ggaren@apple.com

2009-04-16 Geoffrey Garen <ggaren@apple.com>

Reviewed by Eric Seidel and Adam Roben.

More fix for https://bugs.webkit.org/show_bug.cgi?id=21260
Unbounded memory growth when churning elements with anonymous event handler functions


Simplified some EventHandler creation code.


Removed a pile of code whose sole purpose was to allow SVG event handlers
to supply a parameter named "evt" instead of the normal "event", and
replaced it with a single parameter to JSLazyEventListener's constructor
specifying the parameter name to use.

  • bindings/js/JSDOMWindowBase.h:
  • bindings/js/JSLazyEventListener.cpp: (WebCore::JSLazyEventListener::JSLazyEventListener): (WebCore::JSLazyEventListener::parseCode):
  • bindings/js/JSLazyEventListener.h: (WebCore::JSLazyEventListener::create):
  • bindings/js/ScriptController.cpp: (WebCore::ScriptController::createInlineEventListener):
  • bindings/js/ScriptController.h:
  • dom/Document.cpp: (WebCore::Document::createEventListener):
10:09 Changeset [42578] by eric@webkit.org

Reviewed by David Hyatt.

Split out objectBoundingBox and repaintRectInLocalCoordinates from relativeBBox
in preparation for simplifying (and fixing) repaint logic in the SVG rendering tree.
https://bugs.webkit.org/show_bug.cgi?id=25224

objectBoundingBox() matches the SVG 1.1 concept of a "bounding box"

repaintRectInLocalCoordinates() should return the a rect covering all painted content.
However, repaintRectInLocalCoordinates() still only returns the "stroke bounding box"
in this patch. In a future patch, repaintRectInLocalCoordinates will be fixed to return
a rect covering all painted content. In order to avoid changing several hundred layout test results, I've left
the behavior as-is for now. The returned rect is used by various repaintRectInLocalCoordinates
implementations and sometimes adjusted to include all painted content, but not always, and
the places where the adjustments are made are sometimes wrong. Again, will be fixed in
an upcoming patch.

This patch discovered a bug in Font::drawTextUsingSVGFont, which is probably causing
bounding-box relative gradients on SVGFont glyphs to not paint correctly.
I chose not to try and fix the bug in this patch and instead left a FIXME.

This patch also discovered that at least tspan.getBBox() is broken. This
along with the foreignObject.getBBox() change will be tested (and fixed) in a
later patch. https://bugs.webkit.org/show_bug.cgi?id=25225

No change in behavior (besides the above mentioned foreignObject.getBBox()), thus no tests.

  • rendering/RenderForeignObject.cpp: (WebCore::RenderForeignObject::objectBoundingBox): this is a behavior improvement for getBBox() test case coming in a later patch (WebCore::RenderForeignObject::repaintRectInLocalCoordinates): only really used for layout test results, might some day be used for repaint.
  • rendering/RenderForeignObject.h:
  • rendering/RenderObject.cpp: (WebCore::RenderObject::objectBoundingBox): new methods, ASSERT if used but not implemented. (WebCore::RenderObject::repaintRectInLocalCoordinates):
  • rendering/RenderObject.h:
  • rendering/RenderPath.cpp: (WebCore::RenderPath::objectBoundingBox): (WebCore::RenderPath::repaintRectInLocalCoordinates): (WebCore::RenderPath::clippedOverflowRectForRepaint): (WebCore::RenderPath::lineHeight): (WebCore::RenderPath::baselinePosition): (WebCore::RenderPath::paint): (WebCore::RenderPath::addFocusRingRects):
  • rendering/RenderPath.h:
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::paint): (WebCore::RenderSVGContainer::objectBoundingBox): (WebCore::RenderSVGContainer::repaintRectInLocalCoordinates):
  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGGradientStop.h: (WebCore::RenderSVGGradientStop::objectBoundingBox): (WebCore::RenderSVGGradientStop::repaintRectInLocalCoordinates):
  • rendering/RenderSVGHiddenContainer.cpp: (WebCore::RenderSVGHiddenContainer::objectBoundingBox): (WebCore::RenderSVGHiddenContainer::repaintRectInLocalCoordinates):
  • rendering/RenderSVGHiddenContainer.h:
  • rendering/RenderSVGImage.cpp: (WebCore::RenderSVGImage::objectBoundingBox): (WebCore::RenderSVGImage::repaintRectInLocalCoordinates): (WebCore::RenderSVGImage::calculateAbsoluteBounds): (WebCore::RenderSVGImage::addFocusRingRects):
  • rendering/RenderSVGImage.h:
  • rendering/RenderSVGRoot.cpp: (WebCore::RenderSVGRoot::paint): (WebCore::RenderSVGRoot::objectBoundingBox): (WebCore::RenderSVGRoot::repaintRectInLocalCoordinates):
  • rendering/RenderSVGRoot.h:
  • rendering/RenderSVGTSpan.h: (WebCore::RenderSVGTSpan::objectBoundingBox): (WebCore::RenderSVGTSpan::repaintRectInLocalCoordinates):
  • rendering/RenderSVGText.cpp: (WebCore::RenderSVGText::clippedOverflowRectForRepaint): (WebCore::RenderSVGText::absoluteRects): (WebCore::RenderSVGText::absoluteQuads): (WebCore::RenderSVGText::objectBoundingBox): (WebCore::RenderSVGText::repaintRectInLocalCoordinates):
  • rendering/RenderSVGText.h:
  • rendering/RenderSVGViewportContainer.cpp: (WebCore::RenderSVGViewportContainer::viewportTransform):
  • rendering/SVGRenderSupport.cpp: (WebCore::computeContainerBoundingBox):
  • rendering/SVGRenderSupport.h:
  • rendering/SVGRenderTreeAsText.cpp: (WebCore::operator<<):
  • svg/SVGFont.cpp: (WebCore::Font::drawTextUsingSVGFont):
  • svg/SVGLocatable.cpp: (WebCore::SVGLocatable::getBBox):
  • svg/SVGPatternElement.cpp: (WebCore::SVGPatternElement::buildPattern):
  • svg/graphics/SVGPaintServerGradient.cpp: (WebCore::createMaskAndSwapContextForTextGradient): (WebCore::clipToTextMask): (WebCore::SVGPaintServerGradient::setup):
  • svg/graphics/SVGPaintServerPattern.cpp: (WebCore::SVGPaintServerPattern::setup):
08:42 Changeset [42577] by mitz@apple.com

WebCore:

Reviewed by Darin Adler.

  • fix <rdar://problem/6032139> Table cell widths calculated incorrectly on table that uses table-layout:fixed, colspans, and a mix of percentage and pixel widths

Test: fast/table/fixed-granular-cols.html

The incorrect widths resulted from incorrect handling of the case where
the <col> elements are more granular than the table cells.

  • rendering/FixedTableLayout.cpp: (WebCore::FixedTableLayout::calcWidthArray): When processing <col> elements, append effective columns or split existing effective columns as needed.

LayoutTests:

Reviewed by Darin Adler.

  • test for <rdar://problem/6032139> Table cell widths calculated incorrectly on table that uses table-layout:fixed, colspans, and a mix of percentage and pixel widths
  • fast/table/fixed-granular-cols.html: Added.
  • platform/mac/fast/table/fixed-granular-cols-expected.checksum: Added.
  • platform/mac/fast/table/fixed-granular-cols-expected.png: Added.
  • platform/mac/fast/table/fixed-granular-cols-expected.txt: Added.
08:40 Changeset [42576] by aroben@apple.com

Skip yet another JavaScriptCore test that sometimes fails on Windows

See Bug 25160: Various ecma/Date tests sometimes fail on Windows (but
not Mac)
<https://bugs.webkit.org/show_bug.cgi?id=25160>

  • Scripts/run-javascriptcore-tests: Skip ecma/Date/15.9.2.2-6.js.
05:01 Changeset [42575] by oliver@apple.com

Optimise op_resolve_base

Reviewed by Gavin Barraclough.

If we can statically find a property we are trying to resolve
the base of, the base is guaranteed to be the global object.

04:36 Changeset [42574] by oliver@apple.com

Improve performance of read-write-modify operators

Reviewed by Gavin Barraclough

Implement cross scope optimisation for read-write-modify
operators, to avoid unnecessary calls to property resolve
helper functions.

03:10 Changeset [42573] by oliver@apple.com

Improve performance of remaining array enumeration functions

Reviewed by Gavin Barraclough

Make use of function entry cache for remaining Array enumeration functions.

00:54 Changeset [42572] by ap@webkit.org

<rdar://problem/6795285> Infinite recursion in ResourceHandle::receivedRequestToContinueWithoutCredential

Rolled out <http://trac.webkit.org/projects/webkit/changeset/42536> - the two instances of
authentication challenge are different, after all. Added a FIXME comment about a possible
future improvement.

  • platform/network/ResourceHandle.cpp: (WebCore::ResourceHandle::clearAuthentication):
  • platform/network/ResourceHandleInternal.h: (WebCore::ResourceHandleInternal::ResourceHandleInternal):
  • platform/network/cf/ResourceHandleCFNet.cpp: (WebCore::ResourceHandle::didReceiveAuthenticationChallenge): (WebCore::ResourceHandle::receivedCredential): (WebCore::ResourceHandle::receivedRequestToContinueWithoutCredential): (WebCore::ResourceHandle::receivedCancellation):
  • platform/network/mac/ResourceHandleMac.mm: (WebCore::ResourceHandle::didReceiveAuthenticationChallenge): (WebCore::ResourceHandle::didCancelAuthenticationChallenge): (WebCore::ResourceHandle::receivedCredential): (WebCore::ResourceHandle::receivedRequestToContinueWithoutCredential): (WebCore::ResourceHandle::receivedCancellation):
00:24 Changeset [42571] by oliver@apple.com

Improve performance of Array.sort

Reviewed by Gavin Barraclough.

Cache the VM entry for Array.sort when using a JS comparison function.

04/15/09:

21:07 Changeset [42570] by oliver@apple.com

Bug 25229: Need support for Array.prototype.reduceRight
<https://bugs.webkit.org/show_bug.cgi?id=25229>

Reviewed by Gavin Barraclough.

Implement Array.reduceRight

20:51 Changeset [42569] by ggaren@apple.com

2009-04-15 Geoffrey Garen <ggaren@apple.com>

Reviewed by Sam Weinig.


More fix for https://bugs.webkit.org/show_bug.cgi?id=21260
Unbounded memory growth when churning elements with anonymous event handler functions

Also fixed <rdar://problem/6792909> WebInspector crashes when reloading
a page with breakpoints set


  • bindings/js/JSEventListener.cpp: (WebCore::JSEventListener::clearJSFunctionInline): (WebCore::JSEventListener::markJSFunction):
  • bindings/js/JSEventListener.h: Actually clear our function and global object pointers when our client instructs us to. (Oops!) Also, mark our global object while we still intend to use it.
  • bindings/js/JSSVGElementInstanceCustom.cpp: (WebCore::JSSVGElementInstance::mark): (WebCore::toJS):
  • bindings/scripts/CodeGeneratorJS.pm:
  • svg/SVGElementInstance.idl: For every SVGElementInstance wrapper, ensure that the corresponding element also has a wrapper, to keep its event listeners alive.
20:33 Changeset [42568] by dglazkov@chromium.org

2009-04-15 Dimitri Glazkov <dglazkov@chromium.org>

Reviewed by Darin Fisher.

https://bugs.webkit.org/show_bug.cgi?id=25201
Add ScriptFunctionCall/ScriptObject for V8.

  • bindings/v8/ScriptFunctionCall.cpp: Added.
  • bindings/v8/ScriptFunctionCall.h: Added.
  • bindings/v8/ScriptObject.cpp: Added.
  • bindings/v8/ScriptObject.h: Added.
  • bindings/v8/ScriptObjectQuarantine.cpp: Added.
  • bindings/v8/ScriptObjectQuarantine.h: Added.
  • bindings/v8/ScriptScope.cpp: Added.
  • bindings/v8/ScriptScope.h: Added.
  • bindings/v8/ScriptState.h: (WebCore::ScriptState::ScriptState): Added new constructors. (WebCore::ScriptState::frame): Added Frame ptr accessor.
  • bindings/v8/ScriptString.h: (WebCore::ScriptString::ScriptString): Added default constructor.
  • bindings/v8/ScriptValue.h: (WebCore::ScriptValue::isEqual): Added.
19:12 Changeset [42567] by oliver@apple.com

Bug 25227: Array.filter triggers an assertion when the target array shrinks while being filtered
<https://bugs.webkit.org/show_bug.cgi?id=25227>

Reviewed by Gavin Barraclough.

We correct this simply by making the fast array path fall back on the slow path if
we ever discover the fast access is unsafe.

17:49 Changeset [42566] by kov@webkit.org

2009-04-15 Gustavo Noronha Silva <gustavo.noronha@collabora.co.uk>

Unreviewed build fix. Add RenderSVGModelObject.* to GTK+'s build.

  • GNUmakefile.am:
17:02 Changeset [42565] by sfalken@apple.com

Redo last build fix in a more straightforward way.

  • DumpRenderTree/win/DumpRenderTree.cpp: (sharedCFURLCache):
16:46 Changeset [42564] by sfalken@apple.com

Fix Windows build of DumpRenderTree.

  • DumpRenderTree/win/DumpRenderTree.cpp: (sharedCFURLCache): (main):
16:35 Changeset [42563] by oliver@apple.com

Bug 25159: Support Array.prototype.reduce
<https://bugs.webkit.org/show_bug.cgi?id=25159>

Reviewed by Gavin Barraclough

Implement Array.prototype.reduce

15:46 Changeset [42562] by kevino@webkit.org

Reviewed by Kevin Ollivier.

Implement support for wxWebViewNewWindowEvent.

https://bugs.webkit.org/show_bug.cgi?id=19043

15:32 Changeset [42561] by eric@webkit.org

Reviewed by Simon Fraser.

Add minimal RenderSVGModelObject and make RenderPath and RenderSVGContainer use it.
https://bugs.webkit.org/show_bug.cgi?id=25221

RenderSVGModelObject is the base rendering class for things which live
in the SVG rendering tree. Not all SVG renders are RenderSVGModelObjects yet.

More patches coming. This is just adding the class, future patches
will pull more logic out of RenderPath and RenderSVGContainer into RenderSVGModelObject.

  • WebCore.pro:
  • WebCore.scons:
  • WebCore.vcproj/WebCore.vcproj:
  • WebCore.xcodeproj/project.pbxproj:
  • rendering/RenderPath.cpp: (WebCore::RenderPath::RenderPath):
  • rendering/RenderPath.h:
  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::RenderSVGContainer):
  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGModelObject.cpp: Added. (WebCore::RenderSVGModelObject::RenderSVGModelObject):
  • rendering/RenderSVGModelObject.h: Added. (WebCore::RenderSVGModelObject::requiresLayer):
  • rendering/RenderTreeAsText.cpp:
  • rendering/SVGRenderTreeAsText.cpp:
15:22 Changeset [42560] by justin.garcia@apple.com

Fixed an expected result and added a test to the Skipped list while I investigate why its expected results didn't change.

14:32 Changeset [42559] by sfalken@apple.com

<rdar://problem/6785760> WebCore should use a maximum of 6 connections per host, not 4

Reviewed by Adam Roben.

  • platform/network/ResourceRequestBase.cpp:
  • platform/network/cf/ResourceRequestCFNet.cpp: (WebCore::initializeMaximumHTTPConnectionCountPerHost):
14:30 Changeset [42558] by sfalken@apple.com

Updated WebKitSystemInterface for Windows.
Changes needed for <rdar://problem/6785760>

  • win/include/WebKitSystemInterface/WebKitSystemInterface.h:
  • win/lib/WebKitSystemInterface.lib:
  • win/lib/WebKitSystemInterface_debug.lib:
13:28 Changeset [42557] by weinig@apple.com

2009-04-15 Sam Weinig <sam@webkit.org>

Reviewed by Mark Rowe.

Add special casing to bindings generator so that custom functions
starting with xml or xslt conform to WebKit style guidelines.

  • bindings/js/JSDOMWindowCustom.cpp: (WebCore::JSDOMWindow::xmlHttpRequest): (WebCore::JSDOMWindow::xsltProcessor):
  • bindings/scripts/CodeGenerator.pm:
13:25 Changeset [42556] by aroben@apple.com

Skip yet another JavaScriptCore test that sometimes fails on Windows

See Bug 25160: Various ecma/Date tests sometimes fail on Windows (but
not Mac)
<https://bugs.webkit.org/show_bug.cgi?id=25160>

  • Scripts/run-javascriptcore-tests: Skip ecma/Date/15.9.2.2-2.js (who didn't see it coming?).
13:24 Changeset [42555] by eric@webkit.org

Reviewed by Simon Fraser.

Remove more dead code from RenderSVGHiddenContainer.

lineHeight and baselinePosition used to be on RenderSVGContainer
(of which this is a subclass) because RenderSVGContainer used
to be a RenderBox and always the renderer for <svg>
Now <svg> uses RenderSVGRoot when needing a RenderBox (inside HTML)
and RenderSVGViewportContainer (when inside SVG content)
so there is no need for RenderSVGHiddenContainer to have these HTML-specific methods.

  • rendering/RenderSVGHiddenContainer.cpp:
  • rendering/RenderSVGHiddenContainer.h:
13:01 Changeset [42554] by eric@webkit.org

Reviewed by Simon Fraser.

Remove unneeded (broken) code from SVG renderers
https://bugs.webkit.org/show_bug.cgi?id=25214

Mostly due to my historical confusions about the render tree
and some methods not being removed after classes were split.

RenderSVGRoot is an RenderBox and should just use all the
standard RenderBox methods for inspector and repaint rects.

RenderSVGContainer is *not* a RenderBox (any more) and thus
doesn't need lineHeight or width/height or calcBounds.

RenderSVGViewportContainer had some broken code which tried
to see if the click was inside the container at all, but it
was using width/height metrics based off of the containing
block (from calcWidth) which is wrong (since its real
width/height are from its containing viewport not containing block).

  • rendering/RenderSVGContainer.cpp: (WebCore::RenderSVGContainer::RenderSVGContainer): (WebCore::RenderSVGContainer::layout):
  • rendering/RenderSVGContainer.h:
  • rendering/RenderSVGRoot.cpp:
  • rendering/RenderSVGRoot.h:
  • rendering/RenderSVGViewportContainer.cpp: (WebCore::RenderSVGViewportContainer::layout): (WebCore::RenderSVGViewportContainer::nodeAtPoint):
12:42 Changeset [42553] by justin.garcia@apple.com

Updated expected result for a text dragging test that pastes text on Windows but not on Mac (which is expected).

12:33 Changeset [42552] by eric@webkit.org

Reviewed by Simon Fraser and Sam Weinig.

https://bugs.webkit.org/show_bug.cgi?id=25217
Simplify nearestViewportElement and farthestViewportElement using isViewportElement and a for loop
Add ASSERT(element) to all these functions since they should never be passed a null element.

  • svg/SVGLocatable.cpp: (WebCore::isViewportElement): (WebCore::SVGLocatable::nearestViewportElement): (WebCore::SVGLocatable::farthestViewportElement): (WebCore::SVGLocatable::getBBox): (WebCore::SVGLocatable::getCTM): (WebCore::SVGLocatable::getScreenCTM):
11:41 Changeset [42551] by weinig@apple.com

2009-04-15 Sam Weinig <sam@webkit.org>

Reviewed by Anders Carlsson.

Remove support for the ObjCIvar extended attribute from the Objective-C
bindings generator. It is not used anymore.

  • bindings/scripts/CodeGeneratorObjC.pm:
11:30 Changeset [42550] by eric.carlson@apple.com

2009-04-15 Eric Carlson <eric.carlson@apple.com>

Fix incorrect expected result.

  • http/tests/security/local-video-source-from-remote-expected.txt:
11:22 Changeset [42549] by justin.garcia@apple.com

WebCore:

2009-04-15 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.

https://bugs.webkit.org/show_bug.cgi?id=25204
Create a fast path for ReplaceSelectionCommand that merges text nodes


During simple pastes, where we're just pasting a text node into a run of text, we would split the current
text and insert the new node in between. This is slow and we hit this bug:

https://bugs.webkit.org/show_bug.cgi?id=6148

in the layout and rendering code where adjacent text nodes don't shape correctly in Arabic.


This change creates a fast path for ReplaceSelectionCommand that inserts text directly into the
text node that holds the selection (very similar to the fast path we wrote for InsertTextCommand).

  • editing/ReplaceSelectionCommand.cpp: (WebCore::ReplaceSelectionCommand::doApply): (WebCore::ReplaceSelectionCommand::performTrivialReplace):
  • editing/ReplaceSelectionCommand.h:
  • editing/TextIterator.cpp:

LayoutTests:

2009-04-15 Justin Garcia <justin.garcia@apple.com>

Reviewed by Dan Bernstein.


https://bugs.webkit.org/show_bug.cgi?id=25204
Create a fast path for ReplaceSelectionCommand that merges text nodes


Simplified DOM/Render trees:

  • editing/pasteboard/4840662-expected.txt:
  • platform/mac/editing/pasteboard/bad-placeholder-expected.txt:
  • platform/mac/editing/pasteboard/paste-match-style-001-expected.txt:
  • platform/mac/editing/pasteboard/paste-text-019-expected.txt:
  • platform/mac/editing/pasteboard/paste-xml-expected.txt:

Removed redundant styles that were generated during nesting prevention that is no longer needed:

  • platform/mac/editing/pasteboard/4076267-2-expected.txt:
  • platform/mac/editing/pasteboard/5156401-1-expected.txt:


No longer incorrectly uncollapsing unrendered whitespace:

  • editing/pasteboard/paste-into-anchor-text-expected.txt:
  • platform/mac/editing/pasteboard/5387578-expected.txt:


No longer invalidating the old selected DOM range during a paste:

  • editing/pasteboard/copy-in-password-field-expected.txt:
11:18 Changeset [42548] by darin@chromium.org

2009-04-15 Adam Langley <agl@google.com>

Reviewed by Darin Fisher.

Move VDMX parsing into the Chromium Linux port.

https://bugs.webkit.org/show_bug.cgi?id=25116

VDMX tables are optional tables in TrueType fonts which contain the
exact pixel height of a given font at a given pel size. In order to
match Windows font metrics we have to use these numbers.

Previously, the parsing was performed in Skia. As part of the merge
with upstream Skia, an interface for getting table data from a font
has been added to Skia and we're moving the parsing into WebKit.

This does not change any layout tests.

  • platform/graphics/chromium/FontPlatformDataLinux.cpp: (WebCore::FontPlatformData::uniqueID):
  • platform/graphics/chromium/FontPlatformDataLinux.h:
  • platform/graphics/chromium/FontTrueTypeLinux.cpp: Added.
  • platform/graphics/chromium/FontTrueTypeLinux.h: Added.
  • platform/graphics/chromium/SimpleFontDataLinux.cpp: (WebCore::SimpleFontData::platformInit): (WebCore::SimpleFontData::platformWidthForGlyph):
11:15 Changeset [42547] by darin@chromium.org

2009-04-15 Stephen White <senorblanco@chromium.org>

Reviewed by Dimitri Glazkov.

https://bugs.webkit.org/show_bug.cgi?id=25180

Restore stroked outlines to PlatformContextSkia::drawRect().
These were removed inadvertently in r41805, aka
https://bugs.webkit.org/show_bug.cgi?id=24662.
SkRect is { left, top, right, bottom }, not { left, top, width, height }.

  • platform/graphics/skia/PlatformContextSkia.cpp: (PlatformContextSkia::drawRect):
11:14 Changeset [42546] by mitz@apple.com

WebCore:

Reviewed by Dave Hyatt.

  • fix <rdar://problem/6777374> Generated content with display: run-in causes a crash

Test: fast/runin/generated.html

  • rendering/RenderBlock.cpp: (WebCore::RenderBlock::handleRunInChild): Check if the run-in block is generated, and if so, make the RenderInline anonymous instead of passing a 0 node to the RenderInline constructor. If the run-in itself is generated, do move :before and :after children from the block into the inline, as they will not be regenerated. Changed nested ifs into early returns.

LayoutTests:

Reviewed by Dave Hyatt.

  • test for <rdar://problem/6777374> Generated content with display: run-in causes a crash
  • fast/runin/generated.html: Added.
  • platform/mac/fast/runin/generated-expected.checksum: Added.
  • platform/mac/fast/runin/generated-expected.png: Added.
  • platform/mac/fast/runin/generated-expected.txt: Added.
11:13 Changeset [42545] by darin@chromium.org

2009-04-15 Eric Roman <eroman@chromium.org>

Reviewed by Geoffrey Garen.

https://bugs.webkit.org/show_bug.cgi?id=25192

Expose DOMWindow::removeInlineEventListenerForType and
DOMWindow::inlineEventListenerForType as public.

  • page/DOMWindow.h:
11:06 Changeset [42544] by darin@chromium.org

2009-04-14 Eric Roman <eroman@chromium.org>

Reviewed by Darin Adler.

https://bugs.webkit.org/show_bug.cgi?id=25198

Wrap RegisteredEventListener's markEventListeners and invalidateEventListeners
in a USE(JSC), since it doesn't compile with V8 bindings.

  • dom/RegisteredEventListener.h:
10:26 Changeset [42543] by eric.carlson@apple.com

2009-04-15 Eric Carlson <eric.carlson@apple.com>

Reviewed by Alexey Proskuryakov.

Use a media file inside of LayoutTests/http/ so a cgi isn't necessary to load
it. Fixes a test failure introduced in r42533.

  • http/tests/resources/silence.mpg: Added.
  • http/tests/security/local-video-source-from-remote-expected.txt:
  • http/tests/security/local-video-source-from-remote.html: change the remote url, update a comment.
  • http/tests/security/resources/load-media.cgi: Removed.
10:20 Changeset [42542] by darin@chromium.org

2009-04-15 Eric Roman <eroman@chromium.org>

Reviewed by Darin Adler.

https://bugs.webkit.org/show_bug.cgi?id=25194

Fix a caller of Settings::javaScriptCanOpenWindowsAutomatically() to
use the new name.

  • bindings/v8/custom/V8DOMWindowCustom.cpp:
08:25 Changeset [42541] by ap@webkit.org

Reviewed by Darin Adler.

https://bugs.webkit.org/show_bug.cgi?id=25186
There is no XSLT component in bugzilla

  • projects/xslt/index.html: Changed to search by bug title.
08:09 Changeset [42540] by ariya@webkit.org

2009-04-15 Antonio Gomes <antonio.gomes@openbossa.org>

Reviewed by Ariya Hidayat.

Rename extend() method in QWebElementSelection to append().

  • Api/qwebelement.cpp: (QWebElementSelection::append):
  • Api/qwebelement.h:
  • tests/qwebelement/tst_qwebelement.cpp: (tst_QWebElement::appendSelection):
06:42 Changeset [42539] by hausmann@webkit.org

2009-04-15 Antonio Gomes <antonio.gomes@openbossa.org>

Reviewed by Simon Hausmann.

Fixed nit/typo in QWebElement documentation.

06:24 WebKitFeatures edited by sascha@brawer.ch
Fixed broken links to www.css3.info (diff)
04:56 QtWebKitTodo edited by ariya@webkit.org
(diff)
04:50 QtWebKitTodo edited by elbunce@xendom.com
(diff)
04:48 QtWebKitTodo edited by elbunce@xendom.com
(diff)
00:31 Changeset [42538] by oliver@apple.com

Move CallFrameClosure from inside the Interpreter class to its own file.

Reviewed by NOBODY(Build fix)

00:13 Changeset [42537] by oliver@apple.com

Bug 25202: Improve performance of repeated callbacks into the VM

Reviewed by Cameron Zwarich

Add the concept of a CachedCall to native code for use in Array
prototype and similar functions where a single callback function
is called repeatedly with the same number of arguments.

Used Array.prototype.filter as the test function and got a 50% win
over a naive non-caching specialised version. This makes the native
implementation of Array.prototype.filter faster than the JS one once
more.

Note: See TracTimeline for information about the timeline view.