⚠ Archived content — this site is no longer maintained.   Current WebKit documentation is at docs.webkit.org.

Timeline



May 3, 2007:

11:59 PM Changeset in webkit [23342] by sfalken
  • 8 edits in branches/WindowsMerge

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Ada, Oliver.

Checkpoint:
<rdar://problem/4602372> Spelling
<rdar://problem/4769763> Spell checking engine support

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Ada, Oliver.

Checkpoint:
<rdar://problem/4602372> Spelling
<rdar://problem/4769763> Spell checking engine support


Spelling panel is now partially functional with this check-in.
Learn/Ignore not yet implemented, Find Next, Correct, language
selection, grammar toggle works.


Grammar checking is functional with this check-in.

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Ada, Oliver.


Checkpoint:
<rdar://problem/4602372> Spelling
<rdar://problem/4769763> Spell checking engine support

  • Interfaces/IWebEditingDelegate.idl: Add userDescription parameter.
  • WebEditorClient.cpp: (WebEditorClient::updateSpellingUIWithGrammarString): Set user description for spelling errors.
  • WebEditorClient.h:
  • WebView.cpp: (WebView::replaceSelectionWithText): Implemented.

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Ada, Oliver.

Checkpoint:
<rdar://problem/4602372> Spelling
<rdar://problem/4769763> Spell checking engine support

  • platform/win/EditorWin.cpp: (WebCore::findFirstBadGrammarInRange): Account for start offset when reporting bad grammar range (fixed my porting mistake). (WebCore::Editor::advanceToNextMisspelling): Include user description with spelling error. (WebCore::isRangeUngrammatical): Include user description with spelling error.
11:44 PM Changeset in webkit [21246] by sfalken
  • 10 edits in trunk

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Oliver.

Add missing user description parameter to spelling-related editor client method.

  • WebCoreSupport/WebEditorClient.h: (WebEditorClient::updateSpellingUIWithGrammarString):

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Oliver.

Add missing user description parameter to spelling-related editor client method.

  • bridge/EditorClient.h:
  • platform/gdk/EditorClientGdk.cpp: (WebCore::EditorClientGdk::updateSpellingUIWithGrammarString):
  • platform/gdk/EditorClientGdk.h:
  • platform/graphics/svg/SVGImageEmptyClients.h: (WebCore::SVGEmptyEditorClient::updateSpellingUIWithGrammarString):

2007-05-03 Steve Falkenburg <​sfalken@apple.com>

Reviewed by Oliver.

Add missing user description parameter to spelling-related editor client method.

  • WebCoreSupport/EditorClientQt.cpp: (WebCore::EditorClientQt::updateSpellingUIWithGrammarString):
  • WebCoreSupport/EditorClientQt.h:
11:14 PM Changeset in webkit [23341] by aliceli1
  • 3 edits in branches/WindowsMerge/WebCore

Reviewed by Oliver.

Fixed <rdar://5181466> HTML formatted clipboard data is not converted from UTF8 when pasted

  • platform/win/DragDataWin.cpp: (WebCore::fragmentFromHTML):
  • platform/win/PasteboardWin.cpp: (WebCore::fragmentFromCF_HTML): (WebCore::Pasteboard::documentFragment):
11:05 PM Changeset in webkit [23340] by oliver
  • 2 edits in branches/WindowsMerge/WebCore

2007-05-03 Oliver Hunt <​oliver@apple.com>

Reviewed by Steve

Fix misuse of GlobalLock.

  • platform/win/PasteboardWin.cpp: (WebCore::createGlobalData):
9:35 PM Changeset in webkit [21245] by beidson
  • 6 edits in trunk/WebCore

Reviewed by Oliver

Resolve an outstanding FIXME in Loader::numRequests()

Before, numRequests() would iterate through the list of requests pending load and the list of currently
loading requests and tally up a count matching the current DocLoader.

I noticed while studying and cleaning up the loader code that numRequests() is potentially very hot!
Indeed load a complex site with many resources and multiple frames, and this method gets called very often,
tallying up this number every time.

The FIXME was to keep a collection of Requests mapped to each DocLoader. In reality, since this map would
simply be used for retrieving a count, that was overkill. Keeping a request count in the DocLoader itself
along with maintaining that count in Loader as requests come and go is a much better way to do this.

  • loader/DocLoader.cpp: (WebCore::DocLoader::DocLoader): (WebCore::DocLoader::incrementRequestCount): (WebCore::DocLoader::decrementRequestCount): (WebCore::DocLoader::requestCount): Emulate the defunct Loader::numRequests()
  • loader/DocLoader.h:
  • loader/FrameLoader.cpp: (WebCore::numRequests): Call DocLoader::requestCount() directly (WebCore::FrameLoader::checkCompleted): Use numRequests()
  • loader/loader.cpp: (WebCore::Loader::load): Increment the DocLoader's request count (WebCore::Loader::servePendingRequests): If the SubresourceLoader failed to create, decrement the count (WebCore::Loader::didFinishLoading): If the Request is not Multipart, decrement the count (WebCore::Loader::didFail): If the Request is not Multipart, decrement the count (WebCore::Loader::didReceiveResponse): If the Request becomes Multipart, decrement the count (WebCore::Loader::cancelRequests): Decrement the count for the pending requests being tossed, and ASSERT the count is zero after all requests have been cancelled
  • loader/loader.h:
9:34 PM Changeset in webkit [23339] by oliver
  • 3 edits in branches/WindowsMerge/WebCore

2007-05-03 Oliver Hunt <​oliver@apple.com>

Reviewed by Steve.

Add logic to draw underlines for grammar and spelling errors.
(rdar://problem/4769832)

  • platform/win/GraphicsContextWin.cpp: (WebCore::spellingPatternColor): (WebCore::grammarPatternColor): Can't define global const with constructor so use function scoped globals instead. (WebCore::setCGStrokeColor): Utility function pulled over from GraphicsContextCG (WebCore::GraphicsContext::drawLineForMisspellingOrBadGrammar):


  • platform/win/TemporaryLinkStubs.cpp: Remove old drawLineForMisspellingOrBadGrammar stub
8:55 PM Changeset in webkit [21244] by ggaren
  • 3 edits in trunk/WebCore

Reviewed by Brady Eidson.


Some cleanup in preparation for fixing <rdar://problem/5055182> The
back/forward cache has no global cap

  • loader/FrameLoader.cpp: (WebCore::FrameLoader::provisionalLoadStarted): Moved "Can I cache this page?" logic to canCachePage(). Moved "I am caching this page" logic to cachePageToHistoryItem(). (WebCore::FrameLoader::canCachePage): Moved above logic here. (WebCore::FrameLoader::cachePageToHistoryItem): Removed bogus NULL check and failure return value. Moved above logic here. Don't call setCachedPage() until the CachedPage is fully constructed. (WebCore::FrameLoader::purgePageCache): Removed comment that duplicated the function name.
  • loader/FrameLoader.h:
6:54 PM Changeset in webkit [21243] by harrison
  • 2 edits in trunk/WebCore

Reviewed by harrison.

<rdar://problem/5120281> CrashTracer: [USER] 1 crash in Mail at WebCore::CSSComputedStyleDeclaration::copyInheritableProperties() const

Problem is that checkAncestor and lastClosed can be nil when createMarkup is called unrendered nodes, and that was not checked for.


  • editing/markup.cpp: (WebCore::createMarkup): Nil-check checkAncestor and lastClosed.
5:51 PM Changeset in webkit [21242] by bdash
  • 1 copy in tags/Safari-4522.7

New tag.

5:50 PM Changeset in webkit [21241] by bdash
  • 4 edits in branches/Safari-522

Versioning.

4:08 PM Changeset in webkit [21240] by thatcher
  • 2 edits in trunk/WebKit

Reviewed by Kevin.

<rdar://problem/4975212> REGRESSION: With NetNewsWire 2.1.1, the contextual menu shows extra menu items when focus
is placed in input or textarea field

The NetNewsWire UI delegate isn't expecting calls for form controls, so we need to do a linked-on-or-after check.
If the application was linked against Tiger or earlier and the element is a text form control, just return the
default menu items and bypass the delegate call completely.

  • WebCoreSupport/WebContextMenuClient.mm: (isPreVersion3Client): Cache the result of the WebKitLinkedOnOrAfter call (fixMenusToSendToOldClients): Call the new isPreVersion3Client() (fixMenusReceivedFromOldClients): Ditto. (WebContextMenuClient::getCustomMenuFromDefaultItems): Return the default menu items if the element is a text form control.
3:50 PM Changeset in webkit [21239] by bdash
  • 2 edits in branches/Safari-522/WebKit

Merge r21238.

3:48 PM Changeset in webkit [21238] by bdash
  • 2 edits in trunk/WebKit

2007-05-03 Mark Rowe <​mrowe@apple.com>

Reviewed by Geoff and Kevin.

<rdar://problem/5141290> WebAssertions.h is still needed by some internal clients

Second shot at fixing this error. Stub out the macros rather than forwarding to JavaScriptCore, which would
leave clients using this header trying to resolve JavaScriptCore symbols against WebKit when linking. This
should only happen in production builds when assertions should be disabled anyway as anyone building a
development configuration should be in a position to move away from using this header.

  • Misc/WebAssertions.h:
3:23 PM Changeset in webkit [21237] by thatcher
  • 2 edits in trunk/WebCore

Reviewed by Kevin.

<rdar://problem/5016318> Contextual menu item for opening links is missing in compose windows but present in viewer

Add the Open Link, Open Link in New Window, Download Linked File and Copy Link menu
items to the content editable context menu.

  • platform/ContextMenu.cpp: (WebCore::ContextMenu::populate):
3:10 PM Changeset in webkit [21236] by brmorris
  • 2 edits in S60/trunk/WebKit

brmorris, reviewed by sachin

DESC: stub this out in SDK development environment:

  • BrowserCore/Formcontrols/src/FControlInputSkin.cpp: (CFormInputSkin::SelectedLanguage):
3:04 PM Changeset in webkit [21235] by hausmann
  • 14 edits
    1 add in trunk

Gdk build fixes using the Qt build's qmake tool from Holger Freyther, reviewed by Zack.

2:50 PM Changeset in webkit [21234] by thatcher
  • 2 edits in trunk/WebKit

Reviewed by Kevin.

<rdar://problem/5067707> REGRESSION: "Open Link" contextual menu item appears twice in Mail

Remove the check for Mail in fixMenusToSendToOldClients and fixMenusReceivedFromOldClients
when linked on or after Leopard. The isAppleMail() function is still used for Tiger Mail fixups.

  • WebCoreSupport/WebContextMenuClient.mm: (fixMenusToSendToOldClients): (fixMenusReceivedFromOldClients):
2:37 PM Changeset in webkit [21233] by adele
  • 2 edits in trunk/WebCore

Use BUILDING_ON_TIGER.

  • page/mac/FrameMac.mm: (WebCore::Frame::setUseSecureKeyboardEntry):
2:35 PM Changeset in webkit [21232] by beidson
  • 3 edits in trunk/WebCore

Reviewed by Darin

Some more minor cleanup that resolves a FIXME and ditches another pointless method

No change in functionality == no layout test

  • loader/FrameLoader.cpp: (WebCore::FrameLoader::commitProvisionalLoad): Updated a comment (WebCore::FrameLoader::continueLoadAfterWillSubmitForm): Folded startLoading() into this method (WebCore::FrameLoader::continueLoadAfterNavigationPolicy): continueAfterWillSubmitForm becomes continueLoadAfterWillSubmitForm
  • loader/FrameLoader.h: Removed startLoading()
2:30 PM Changeset in webkit [21231] by brmorris
  • 3 edits in S60/trunk/S60Internals

brmorris, reviewed by rasvcan

DESC: land some changes to wmlengine that i forgot to commit :(

  • S60DynamicLibs.zip: updated wmlengine.lib
  • S60Headers.zip: updated wmlengineinterface.h
2:29 PM Changeset in webkit [21230] by adele
  • 2 edits in trunk/WebCore

Reviewed by Tim Hatcher.

Fix for <rdar://problem/4727607> REGRESSION: KeyScript is deprecated, need a new solution for forcing Roman input for password fields on Leopard

  • page/mac/FrameMac.mm: (WebCore::Frame::setUseSecureKeyboardEntry): Use TSMSetDocumentProperty and TSMRemoveDocumentProperty on Leopard disable and enable non-ascii input sources.
11:50 AM Changeset in webkit [21229] by beidson
  • 3 edits in trunk/WebCore

Reviewed by Anders

commitProvisionalLoad() was the name of two methods.
One of them was private and only called from one other private place.
This was driving me crazy. I had to change that.

No change in functionality == no layout test

  • loader/FrameLoader.cpp: (WebCore::FrameLoader::transitionToCommitted): roll commitProvisionalLoad() into this method
  • loader/FrameLoader.h: Removed commitProvisionalLoad()
11:37 AM Changeset in webkit [21228] by ggaren
  • 2 edits in trunk/WebKitTools

Build fix: Add AllInOneFile.o to the ignore list, since it includes files
in the ignore list.


  • Scripts/check-for-global-initializers:
9:23 AM Changeset in webkit [21227] by darin
  • 6 edits in trunk/WebCore

Reviewed by Anders.

  • make ICU initialization lazy to speed up application start-up time (when it includes an initial webpage load)
  • platform/TextCodecICU.h: Replaced registerEncodingNames and registerCodecs with registerBaseEncodingNames, registerBaseCodecs, registerExtendedEncodingNames, and registerExtendedCodecs.
  • platform/TextCodecICU.cpp: (WebCore::newTextCodecICU): Moved up to the top of the file. (WebCore::TextCodecICU::registerBaseEncodingNames): Added. Registers just UTF-8. (WebCore::TextCodecICU::registerBaseCodecs): Ditto. (WebCore::TextCodecICU::registerExtendedEncodingNames): Renamed. (WebCore::TextCodecICU::registerExtendedCodecs): Ditto. (WebCore::TextCodecICU::createICUConverter): Compare with the name "GBK" in a way that does not require extending the maps.
  • platform/TextEncoding.cpp: (WebCore::TextEncoding::usesVisualOrdering): Added code so this will just return false if we haven't used any extended encoding names yet. The result will be correct, and we won't ask for an encoding name that will require registering the extended encoding names. (WebCore::TextEncoding::isJapanese): Ditto. (WebCore::TextEncoding::backslashAsCurrencySymbol): Ditto, but return a backslash instead of false.
  • platform/TextEncodingRegistry.h: Added a new function named noExtendedTextEncodingNameUsed.
  • platform/TextEncodingRegistry.cpp: (WebCore::buildBaseTextCodecMaps): Added. Registers the base encodings, including Latin-1 because it's the default, and the UTF encodings. (WebCore::extendTextCodecMaps): Added. Registers all the other encodings. (WebCore::newTextCodec): Removed the all to buildTextCodecMap because we always build the map when making the canonical encoding name. (WebCore::atomicCanonicalTextEncodingName): Changed code to build only the base maps at first, then extend the maps the first time we run into a name we don't know. (WebCore::noExtendedTextEncodingNameUsed): Added.

May 2, 2007:

9:39 PM Changeset in webkit [21226] by bdash
  • 2 edits in trunk/WebCore

2007-05-03 Mark Rowe <​mrowe@apple.com>

Qt build fix. In r21185 the Qt version of WebCore::scaleDragImage was
updated to have the incorrect signature.

  • platform/qt/DragImageQt.cpp: (WebCore::scaleDragImage):
3:27 PM Changeset in webkit [21225] by bdakin
  • 2 edits in trunk/WebCore
  • dom/Node.h: Just fixing a variable name because it's driving me mad.
3:16 PM Changeset in webkit [21224] by bdakin
  • 15 edits in trunk/WebCore

Reviewed by Hyatt.

Fix for <rdar://problem/4513966> Never-ending identical repaint @
orbitz.com (Safari eating 85% CPU)

We need to be a little bit more savvy about when we do a FULL style
recalc since it can be unnecessary and expensive.

  • css/CSSMutableStyleDeclaration.cpp: (WebCore::CSSMutableStyleDeclaration::setChanged): Now takes a StyleChangeType as a parameter. (WebCore::CSSMutableStyleDeclaration::setProperty): Send InlineStyleChange as the paramter. This is the optimization!
  • css/CSSMutableStyleDeclaration.h: setChanged takes a StyleChangeType as a parameter.
  • dom/Document.cpp: setChanged now takes a StyleChangeType instead of a bool. (WebCore::Document::recalcStyle): Same.
  • dom/Element.cpp: (WebCore::Element::recalcStyle): This is the fix!! Only set change to Force is we use descendant rules AND it's a FullStyleChange.
  • dom/Node.cpp: m_hasStyle was an unused bit on Node, so this patch gets rid of it and uses that extra bit for m_changed to be a StyleChangeType instead of a bool. (WebCore::Node::Node): (WebCore::Node::setChanged): (WebCore::Node::dump):
  • dom/Node.h: Same. (WebCore::): Define the StyleChangeType enum. (WebCore::Node::hasClass): (WebCore::Node::changed): (WebCore::Node::styleChangeType): (WebCore::Node::setHasClass):
  • dom/StyledElement.cpp: (WebCore::StyledElement::parseMappedAttribute): Don't bother calling setHasStyle() since m_hasStyle is gone now. It was never used, and this was the only place it was set.
  • dom/Text.cpp: (WebCore::Text::recalcStyle): setChanged now takes a StyleChangeType instead of a bool.
  • html/HTMLFrameSetElement.cpp: (WebCore::HTMLFrameSetElement::recalcStyle): Same.
  • html/HTMLSelectElement.cpp: (WebCore::HTMLSelectElement::restoreState): Same. (WebCore::HTMLSelectElement::reset): Same.
  • html/HTMLTextAreaElement.cpp: (WebCore::HTMLTextAreaElement::setValue): Same.
  • ksvg2/misc/SVGTimer.cpp: (WebCore::SVGTimer::applyAnimations): Same.
  • ksvg2/svg/SVGUseElement.cpp: (WebCore::SVGUseElement::recalcStyle): Same.
  • platform/graphics/svg/SVGResource.cpp: (WebCore::SVGResource::repaintClients): Same.
12:36 PM Changeset in webkit [21223] by andersca
  • 2 edits in trunk/WebKit

Reviewed by Brady.

<rdar://problem/5151113> Assertion firing in [FrameProgressEntry addChild:forDataSource:] when navigating cnn.com


The assertion fired because a plug-in was trying to load a subresource when a new load had started but not yet
committed. The check that would have prevented this was removed in order to fix <rdar://problem/5085897>.


This puts back the check but changes it to allow loads where the target is the same frame as the plugin's parent frame.


  • Plugins/WebBaseNetscapePluginView.mm: (-[WebBaseNetscapePluginView loadRequest:inTarget:withNotifyData:sendNotification:]):
7:54 AM Changeset in webkit [21222] by harrison
  • 3 edits in trunk/WebCore

Reviewed by Antti.

<rdar://problem/4859132> Grammar must always be checked in entire-sentence chunks, and shouldn't show markers for current sentence

Restore the remainder of the original fix now that blocking bug rdar://5174862 is addressed.

John made the original patch on 2007-04-10, but this fell prey to rdar://5157329.
Justin backed out the whole patch on 2007-04-30 to avoid the crash.
Justin then restored most of the original patch, but used word granularity instead, to keep the build working.
I just addressed the root cause of rdar://5157329 in rdar://5174862.
This patch switches the temp uses of word granularity back to sentence granularity.


  • editing/mac/EditorMac.mm: (WebCore::Editor::markMisspellingsAfterTypingToPosition):
  • page/Frame.cpp: (WebCore::Frame::respondToChangedSelection):
6:11 AM Changeset in webkit [21221] by harrison
  • 6 edits
    4 adds in trunk

JavaScriptCore:

Reviewed by Antti.

<rdar://problem/5174862> Crash resulting from DeprecatedString::insert()

Added insertion support for more than one value.


  • wtf/Vector.h: (WTF::::insert): Added support for inserting multiple values.


(WTF::::prepend):
New. Insert at the start of vectors. Convenient for vectors used as strings.

LayoutTests:

Reviewed by Antti.

  • test for <rdar://problem/5174862> Crash resulting from DeprecatedString::insert()


  • editing/selection/move-by-sentence-001-expected.checksum: Added.
  • editing/selection/move-by-sentence-001-expected.png: Added.
  • editing/selection/move-by-sentence-001-expected.txt: Added.
  • editing/selection/move-by-sentence-001.html: Added.

WebCore:

Reviewed by Antti.

<rdar://problem/5174862> Crash resulting from DeprecatedString::insert()


Test added: editing/selection/move-by-sentence-001.html

  • editing/visible_units.cpp: (WebCore::previousBoundary): (WebCore::nextBoundary): Use a UChar Vector instead of DeprecatedString. Avoid creating an extra string for secure bullet replacement unless it is actually needed.


  • platform/DeprecatedString.cpp: (WebCore::DeprecatedString::insert): Call forceUnicode() before setLength(), so that only the unicode buffer is resized.

(WebCore::DeprecatedString::setLength):

Adjust the unicode buffer even if the ascii buffer is valid.

2:21 AM Changeset in webkit [21220] by bdash
  • 3 edits in trunk/WebKitTools

2007-05-02 Holger Freyther <​freyther@kde.org>

Reviewed by Mark Rowe.

  • Scripts/build-webkit: Remove the CMake call.
  • Scripts/webkitdirs.pm: Remove isQtWithQMake as QMake is now the only buildsystem for the Qt port.
2:20 AM Changeset in webkit [21219] by bdash
  • 6 edits in trunk/WebCore

2007-05-02 Mitz Pettel <​mitz@webkit.org>

Reviewed by Brady.

No test added since there is no change in functionality.

  • loader/FrameLoader.cpp: (WebCore::FrameLoader::didOpenURL): (WebCore::FrameLoader::begin): (WebCore::FrameLoader::open):
  • loader/FrameLoader.h: Made clear() private.
  • page/Frame.cpp: (WebCore::Frame::setDocument): Changed the argument into a PassRefPtr.
  • page/Frame.h:
  • xml/XSLTProcessor.cpp: (WebCore::XSLTProcessor::createDocumentFromSource):
2:17 AM Changeset in webkit [21218] by bdash
  • 3 edits
    5 adds in trunk

2007-05-02 Mitz Pettel <​mitz@webkit.org>

Reviewed by Dave Hyatt.

Test: fast/replaced/image-resize-width.html

  • rendering/RenderImage.cpp: (WebCore::RenderImage::imageChanged): Moved the call to dirtyPrefWidths() before the call to calcWidth().

2007-05-02 Mitz Pettel <​mitz@webkit.org>

Reviewed by Dave Hyatt.

  • fast/replaced/image-resize-width-expected.checksum: Added.
  • fast/replaced/image-resize-width-expected.png: Added.
  • fast/replaced/image-resize-width-expected.txt: Added.
  • fast/replaced/image-resize-width.html: Added.
  • fast/replaced/resources/rectangle-red-200x100.png: Added.
1:44 AM Changeset in webkit [21217] by hyatt
  • 4 adds in trunk/LayoutTests/fast/block/positioning

Add layout test for dashboard bug.

1:39 AM Changeset in webkit [21216] by hyatt
  • 2 edits in trunk/WebCore

Fix for bug 13566, broken Dashboard widgets. setPrefWidthsDirty did
the wrong thing for text children of a positioned element.

Reviewed by mitz

fast/block/positioning/pref-width-change.html

  • rendering/RenderObject.cpp: (WebCore::RenderObject::setPrefWidthsDirty):
Note: See TracTimeline for information about the timeline view.