Fix for 3833123, setting a cell's colspan does not update rendering like it should.

[WebKit-https.git] / WebCore / ChangeLog-2005-08-23

2004-12-14  David Hyatt  <hyatt@apple.com>

        Fix for 3833123, setting a cell's colspan does not update rendering like it should.
        
        Reviewed by rjw

        * khtml/html/html_tableimpl.cpp:
        (HTMLTableCellElementImpl::parseHTMLAttribute):
        * khtml/rendering/render_table.cpp:
        (RenderTableCell::RenderTableCell):
        (RenderTableCell::updateFromElement):

2004-12-14  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3864536> crash copying text from other apps and pasting in a sticky in dashboard

        Reviewed by hyatt.

        * khtml/editing/markup.cpp:
        (khtml::createFragmentFromText): ref and deref the fragment since calling appendChild can completely deref it

2004-12-14  David Hyatt  <hyatt@apple.com>

        Fix for 3785211. Make sure to do a layout with the old position before doing a layout after changing to the new position.
        This is actually a regression from my positioned object DHTML optimization.
        
        Reviewed by kocienda

        * khtml/rendering/render_object.cpp:
        (RenderObject::setStyle):

2004-12-14  David Hyatt  <hyatt@apple.com>

        Make sure the class attribute works when newlines are present in the attribute.

        Reviewed by kocienda

        * khtml/html/html_elementimpl.cpp:
        (HTMLNamedAttrMapImpl::parseClassAttribute):

2004-12-14  David Hyatt  <hyatt@apple.com>

        Fix for 3724938, float element is duplicated and paints twice.  The logic for when to paint floats was
        actually fairly screwed up.  This patch simplifies the logic and makes addOverhangingFloats easier to
        read by splitting it into two separate functions.
        
        Reviewed by kocienda

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::layoutBlockChildren):
        (khtml::RenderBlock::clearFloats):
        (khtml::RenderBlock::addOverhangingFloats):
        (khtml::RenderBlock::addIntrudingFloats):
        * khtml/rendering/render_block.h:

2004-12-14  John Sullivan  <sullivan@apple.com>

        Reviewed by Ken.
        
        - added Undo action names for Cut, Paste, and Drag

        * khtml/editing/htmlediting.h:
        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::editingAction):
        overridden to return HTMLEditActionCut
        (khtml::MoveSelectionCommand::editingAction):
        overridden to return HTMLEditActionDrag
        (khtml::ReplaceSelectionCommand::editingAction):
        overridden to return HTMLEditActionPaste

2004-12-14  John Sullivan  <sullivan@apple.com>

        Reviewed by Ken.
        
        - architecture for WebCore part of fix for <rdar://problem/3790011> undoable operations all say "Undo" in the menu, 
        no specific action names
          
        The remaining step is to make each EditCommand subclass override editingAction() to return an
        appropriate value. (Unfortunately the mapping between subclass and user-distinguishable action
        is not completely straightforward, so this next step isn't trivial.)

        * khtml/editing/htmlediting.h:
        new enum for HTMLEditAction

        * khtml/editing/htmlediting.cpp:
        (khtml::EditCommandPtr::editingAction):
        new method, calls through to EditCommand
        (khtml::EditCommand::editingAction):
        new method for subclasses to override, returns HTMLEditActionUnspecified at this level
        (khtml::TypingCommand::editingAction):
        proof of concept override, returns HTMLEditActionTyping
                
        * kwq/KWQKHTMLPart.h:
        declare new private bottleneck method registerCommandForUndoOrRedo
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::registerCommandForUndoOrRedo):
        new bottleneck method to reduce code duplication; now calls over the bridge
        to get the localized string to use for the Undo action name
        (KWQKHTMLPart::registerCommandForUndo):
        now calls new bottleneck method
        (KWQKHTMLPart::registerCommandForRedo):
        now calls new bottleneck method
        
        * kwq/WebCoreBridge.h:
        new enum for WebUndoAction, maps directly to HTMLEditAction.
        Declaration of nameForUndoAction:

2004-12-14  Darin Adler  <darin@apple.com>

        Reviewed by John.

        - added a bunch of missing nil checks; our old version of inherits used to work for nil (by accident)

        * khtml/rendering/render_frames.cpp:
        (RenderPart::~RenderPart): Check widget for nil.
        (RenderPart::setWidget): Ditto.
        (RenderFrame::slotViewCleared): Ditto.
        (RenderPartObject::slotViewCleared): Ditto.

2004-12-13  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        - moved markup-related functions into new sources files in the editing directory
        - removed all of the uses of dynamic_cast, preparing to turn off RTTI to make our code smaller and slightly faster

        * ForwardingHeaders/editing/markup.h: Added.
        * khtml/editing/markup.h: Added.
        * khtml/editing/markup.cpp: Added.

        * WebCore.pbproj/project.pbxproj: Added markup.h/cpp.

        * khtml/dom/dom_node.cpp: (Node::toHTML): Call createMarkup since there's no toHTML in NodeImpl any more.
        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::innerHTML): Changed to call createMarkup.
        (HTMLElementImpl::outerHTML): Ditto.

        * khtml/ecma/kjs_window.cpp:
        (Window::retrieveWindow): Comment out assert that uses dynamic_cast.
        (Window::retrieveActive): Ditto.

        * khtml/editing/htmlediting.h: Added forward class declaration needed now that I removed one elsewhere.
        * khtml/xml/dom_docimpl.h: Ditto.

        * khtml/khtml_part.cpp:
        (KHTMLPart::slotDebugDOMTree): Use createMarkup instead of toHTML.
        (KHTMLPart::processObjectRequest): Use inherits instead of dynamic_cast.

        * khtml/rendering/render_image.cpp: (RenderImage::paint): Add an explicit QChar conversion so this code
        still works even with the additional replace overloads added to QString.
        * kwq/KWQTextCodec.mm: (QTextCodec::fromUnicode): Ditto.

        * khtml/rendering/render_object.h: Removed the version of arenaDelete that does not take an object
        base pointer, because it used dynamic_cast in its implementation. Made the other version public.
        * khtml/rendering/render_object.cpp: Ditto.
        * khtml/rendering/render_replaced.cpp: (RenderWidget::deref): Pass object base pointer to arenaDelete.

        * khtml/xml/dom2_rangeimpl.h: Removed extra parameters from toHTML, and unneeded includes and declarations.
        * khtml/xml/dom2_rangeimpl.cpp: (DOM::RangeImpl::toHTML): Changed to call createMarkup, and moved all
        the support code into markup.cpp.

        * khtml/xml/dom_nodeimpl.h: Moved toHTML and related functions into markup.cpp.
        * khtml/xml/dom_nodeimpl.cpp: Ditto.

        * khtml/xml/dom_position.cpp:
        (DOM::startPosition): Implemented the version of this that takes a RangeImpl. Also added null checks
        so these return null positions rather than raising exceptions.
        (DOM::endPosition): Ditto.

        * khtml/khtmlview.h: Added an APPLE_CHANGES function so inherits can detect this class without dynamic_cast.
        * kwq/KWQFrame.h: Ditto.
        * kwq/KWQFrame.mm: (QFrame::isQFrame): Ditto.
        * kwq/KWQKPartsPart.h: Ditto.
        * kwq/KWQKPartsPart.mm: (KParts::ReadOnlyPart::isKPartsReadOnlyPart): Ditto.
        * kwq/KWQScrollView.h: Ditto.
        * kwq/KWQScrollView.mm: (QScrollView::isQScrollView): Ditto.
        * kwq/KWQKHTMLPart.h: Ditto.
        * kwq/KWQKHTMLPart.mm:
        (KHTMLView::isKHTMLView): Ditto.
        (KWQKHTMLPart::setTitle): Added an explicit QChar conversion so this code still works even with the additional
        replace overloads added to QString.
        (KWQKHTMLPart::setStatusBarText): Ditto.
        (KWQKHTMLPart::runJavaScriptAlert): Ditto.
        (KWQKHTMLPart::runJavaScriptConfirm): Ditto.
        (KWQKHTMLPart::runJavaScriptPrompt): Ditto.
        (KWQKHTMLPart::attributedString): Ditto.
        (KWQKHTMLPart::isCharacterSmartReplaceExempt): Ditto.
        (KWQKHTMLPart::isKHTMLPart): That dynamic_cast thing (see above).

        * kwq/KWQObject.h: Added virtual methods for the few cases where we need dynamic_cast-like behavior.
        * kwq/KWQObject.mm:
        (QObject::inherits): Changed to not use dynamic cast.
        (QObject::isKHTMLPart): Added. Returns false.
        (QObject::isKHTMLView): Ditto.
        (QObject::isKPartsReadOnlyPart): Ditto.
        (QObject::isQFrame): Ditto.
        (QObject::isQScrollView): Ditto.

        * kwq/KWQRenderTreeDebug.cpp:
        (write): Changed to use inherits rather than dynamic_cast.
        (writeSelection): Ditto.

        * kwq/KWQSlot.mm: (KWQSlot::call): Call through to the version with just a job pointer parameter rather
        than going straight on to the "no parameters at all" version.

        * kwq/KWQString.h:
        * kwq/KWQString.mm: (QString::replace): Added overloads.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge markupStringFromNode:nodes:]): Changed to call functions in markup.h.
        (-[WebCoreBridge markupStringFromRange:nodes:]): Ditto.
        (-[WebCoreBridge selectedString]): Added an explicit QChar conversion so this code still works even with
        the additional replace overloads added to QString.
        (-[WebCoreBridge stringForRange:]): Ditto.
        (-[WebCoreBridge copyDOMNode:copier:]): Changed to call functions in markup.h.
        (-[WebCoreBridge elementAtPoint:]): QChar conversion.
        (-[WebCoreBridge documentFragmentWithMarkupString:baseURLString:]): Changed to call functions in markup.h.
        (-[WebCoreBridge documentFragmentWithText:]): Changed to call functions in markup.h.

2004-12-13  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3917956> REGRESSION (Mail): pasting can leave insertion point inside pasted text

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplaceSelectionCommand::doApply): Fix coding mistake. Calculations of bool flag based on 
        leading and trailing whitespace positions was reversed! I must have introduced this error recently
        when changing around this code.

2004-12-13  David Hyatt  <hyatt@apple.com>

        Fix for 3915787, macobserver doesn't paint.  floatRect() needed to be const in the base class.  Also hit-testing
        and painting was using the wrong rect when setting up the x/y of the rect.
        
        Reviewed by mjs

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::paint):
        (khtml::RenderBlock::nodeAtPoint):
        * khtml/rendering/render_object.h:
        (khtml::RenderObject::floatRect):

2004-12-13  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3917863> REGRESSION (Mail): pasting two lines of plain text copied from an RTF document results in two styles

        Code to figuire out the end node to merge was missing the font tag in the second paragraph
        written out by AppKit convert-to-HTML function. I refined the algorithm to be smarter.

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplacementFragment::mergeEndNode): Refine algorithm used to walk through the fragment being pasted
        looking for the node that is the last inline in the last block of the fragment. The old algorithm was 
        insufficiently powerful.
        (khtml::ReplacementFragment::enclosingBlock): New helper function.
        * khtml/editing/htmlediting.h: Add declaration for new helper function.
        * layout-tests/editing/pasteboard/paste-text-011-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-011.html: Added.

2004-12-13  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        WebCore side of fix for this bug:
        
        <rdar://problem/3768372> REGRESSION (Mail): paste of text ending in whitespace loses whitespace

        Note that we are coordinating with Doug Davidson on the AppKit team to make a complete fix for this
        bug. This change involves our half of the needed changes.
        
        Note that a lot of this change has to do with changing code to use a <br> element instead of
        a comment node as the mechanism to annotate HTML with information used to fix the bug. In some
        other places, code to handle comments in markup can be removed since we do not use comments for
        such annotations after this change.

        * khtml/editing/htmlediting.cpp: Remove isComment() helper; no longer needed.
        (khtml::ReplacementFragment::ReplacementFragment): Change m_hasInterchangeNewlineComment name to m_hasInterchangeNewline.
        (khtml::ReplacementFragment::isInterchangeNewlineNode): Name changed from isInterchangeNewlineComment.
        (khtml::ReplacementFragment::isInterchangeConvertedSpaceSpan): Local variable name convertedSpaceSpanClass changed to
        convertedSpaceSpanClassString to match other uses of the idiom used here.
        (khtml::ReplaceSelectionCommand::doApply): Change hasInterchangeNewlineComment() name to hasInterchangeNewline().
        * khtml/editing/htmlediting.h: Change names as noted in .cpp file. Remove isComment() helper; no longer needed.
        (khtml::ReplacementFragment::hasInterchangeNewline):  Change hasInterchangeNewlineComment() name to hasInterchangeNewline().
        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::createContextualFragment): No longer has includeCommentsInDOM flag; no longer needed as we do not
        annotate fragments with comments any longer.
        * khtml/html/html_elementimpl.h: Ditto.
        * khtml/xml/dom2_rangeimpl.cpp: Remove addCommentToHTMLMarkup() helper. No longer needed.
        (DOM::interchangeNewlineMarkupString): New helper to return <br> element markup we use to annotate content for interchange.
        (DOM::RangeImpl::toHTML): No longer uses addCommentToHTMLMarkup; now calls interchangeNewlineMarkupString(). Remove
        spurious semi-colon.
        * khtml/xml/dom2_rangeimpl.h: Remove obsolete addCommentToHTMLMarkup() function and EAddToMarkup enum.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge documentFragmentWithMarkupString:baseURLString:]): No longer pass bool to ask for including comments
        in DOM when calling createContextualFragment().

2004-12-10  John Sullivan  <sullivan@apple.com>

        fixed deployment build bustage that John Louch ran into

        * kwq/KWQTextEdit.mm:
        (QTextEdit::setScrollBarModes):
        move bool declaration inside exception-handling block to avoid obscure
        compiler error

2004-12-10  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Richard.

        <rdar://problem/3907484> REGRESSION (125-173): crash when KWQTextField is dealloc'ed while setting focus (profoundlearning.com)
        
        * kwq/KWQWidget.mm:
        (QWidget::setFocus): Handle the case where setting focus removed
        us from the superview - this can happen due to style changes on
        focus change.

2004-12-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Hyatt

        Fix for this bug:
        
        <rdar://problem/3915008> REGRESSION (Mail): Too much white space between lines separated by carriage returns

        There are a number of interesting things we could do to fix this bug, including SPI and involving
        the WebKit delegate, etc., however it seems reasonable to start with a hard-coded default that
        will fix the bug in the general case until such time as we can come up with more specific
        solutions. 
        
        So, I added a helper method to create <p> elements with an inline style that sets top and bottom margins
        to 0.1em.

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorCommand::createParagraphElement): New factory method to create
        paragraph elements to insert. Also adds style information to keep the <p> from having "too-big" margins.
        (khtml::InsertParagraphSeparatorCommand::doApply): Call new factory method.
        * khtml/editing/htmlediting.h: Add createParagraphElement() declaration.

2004-12-10  Darin Adler  <darin@apple.com>

        Reviewed by Hyatt.

        - fixed <rdar://problem/3910419> setting style={overflow:hidden} for <textarea> does not prevent appearance of scrollbars

        * khtml/rendering/render_form.h: Remove now-unneeded wrap parameter.
        * khtml/rendering/render_form.cpp:
        (RenderSubmitButton::rawText): Convert to QChar explicitly.
        (RenderLineEdit::updateFromElement): Ditto.
        (RenderLineEdit::slotTextChanged): Ditto.
        (RenderSelect::updateFromElement): Ditto.
        (TextAreaWidget::TextAreaWidget): Moved out most of the initialization since it's not something
        that requires a derived class. Now we don't use this class at all for WebCore, but they still
        have it for KDE.
        (TextAreaWidget::event): Moved out the ifdefs.
        (RenderTextArea::RenderTextArea): Moved setting code from TextAreaWidget here. Put a bunch that
        we don't need at all inside !APPLE_CHANGES, and removed the setting for scroll bars, since that's
        now done in setStyle.
        (RenderTextArea::handleFocusOut): Use type QTextEdit instead of TextAreaWidget since that's all
        that's needed and WebCore no longer has TextAreaWidget.
        (RenderTextArea::calcMinMaxWidth): Ditto.
        (RenderTextArea::setStyle): Add code to set scroll bar modes based on wrap setting combined with
        overflow style.
        (RenderTextArea::updateFromElement): Use type QTextEdit.
        (RenderTextArea::text): Ditto.
        (RenderTextArea::select): Ditto.

        * kwq/KWQTextArea.mm:
        (-[KWQTextArea _configureTextViewForWordWrapMode]): Don't set horizontal scroller visibility here,
        since it's now handled by QTextEdit.
        (-[KWQTextArea initWithFrame:]): Don't set vertical scroller visibility or scroller auto-hiding.

        * kwq/KWQTextEdit.h: Add setScrollBarModes function to be used instead of separate setter for
        the horizontal and vertical mode; needed because AppKit switches "autohide" for both at once.
        * kwq/KWQTextEdit.mm: (QTextEdit::setScrollBarModes): Added.

2004-12-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Darin

        Fix for this bug:
        
        <rdar://problem/3915047> HItting return in empty document inserts <p> but 
        insertion point does not move

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorCommand::doApply): The issue is that the
        code to insert the <p> element for the return is not detecting the fact
        that the document is empty. Inserting a <p> into an empty body will not
        "add a new line" as the user expects. With this change, we'll add a second 
        <p> when the root editable element has no rendered kids.

2004-12-10  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by John.

        <rdar://problem/3912979> REGRESSION (125-173): repro crash in HTMLCollectionImpl code (www.clubtravel.ie)

        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::traverseNextItem): Pass base when traversing
        the initial one step, otherwise we might inadvertantly step
        outside the collection base, thereby causing assertion failures or
        other badness later.

2004-12-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Chris

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorCommand::doApply): There is a starting block which is supposed to 
        act as the root node for this operation. However, a loop was incorrectly coded, and a parent node
        search could escape this node. Also, one other piece to code to move nodes to the new <p> element
        should do nothing if the starting point for the selection is itself the starting block.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::doApply): Similar changes, in spirit, to the above
        function, though the names and concepts are slightly different.

2004-12-10  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Ken.

        <rdar://problem/3907705> REGRESSION (172-173): DHTML menus are broken at hrweb.apple.com
        
        * khtml/dom/html_document.cpp:
        (HTMLDocument::nameableItems): New method, wrapper for HTMLCollection creation.
        * khtml/dom/html_document.h:
        * khtml/ecma/kjs_html.cpp:
        (KJS::HTMLDocument::tryGet): use doc.nameableItems(), not doc.all()!
        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::traverseNextItem): Added new DOC_NAMEABLE_ITEMS type, this represents
        the items that can be accessed directly as a document propery, in particular forms, images,
        objects, applets and embeds.
        (HTMLCollectionImpl::updateNameCache): Fix some nameCache/idCache confusion.
        (HTMLFormCollectionImpl::updateNameCache): Ditto.
        * khtml/html/html_miscimpl.h:
        (DOM::HTMLCollectionImpl::): Added new type.

2004-12-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3914779> REGRESSION (Mail): Cannot arrow navigate to position before last character on text-wrapped line

        * khtml/rendering/render_text.cpp:
        (RenderText::caretRect): Code was not detecting space at the end of a line correctly. Now it does.

2004-12-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:

        <rdar://problem/3914755> REGRESSION (Mail): Insertion point disappears after pasting paragraph

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplaceSelectionCommand::doApply): Selection could end up in a "placeholder" node
        that was removed from the document when pasting. This caused the disappearance. Now this
        is detected, and the selection is shifted to a node that is in the document.

2004-12-09  Richard Williamson   <rjw@apple.com>

        Check to disable threaded decoding during
        layout tests wasn't invoking function, just checking address of 
        function, which would always return true.
        
        * kwq/WebCoreImageRendererFactory.m:
        (+[WebCoreImageRendererFactory shouldUseThreadedDecoding]):

2004-12-09  David Hyatt  <hyatt@apple.com>

        Fix for 3892686, left/top overflow was not being propagated properly up to containing blocks because of a math
        error.

        Also fix a bug I noticed on the same page where relative position offsets were not being added in properly for
        all inlines when repainting.
        
        Reviewed by rjw

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::layoutBlockChildren):
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::getAbsoluteRepaintRect):

2004-12-09  David Hyatt  <hyatt@apple.com>

        Fix for 3867545, finance.yahoo.com lays out incorrectly.  Add a quirk that will prevent tables from moving
        down below floats when there is insufficient space.  Instead we will match Gecko and just spill out of the
        containing block to the right.  This appears to be the more common desired behavior, despite being wrong.
        WinIE sometimes wraps and sometimes doesn't, but it's really hard for us to match its inconsistency.
        
        Reviewed by john

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::getClearDelta):
        * layout-tests/apple-only/base/www.cnn.com/index-expected.txt:
        * layout-tests/apple-only/base/www.ebay.com/index-expected.txt:
        * layout-tests/apple-only/base/www.excite.com/index-expected.txt:
        * layout-tests/fast/block/margin-collapse/102-expected.txt:
        * layout-tests/fast/block/margin-collapse/102.html:

2004-12-09  Richard Williamson   <rjw@apple.com>

        Fixed <rdar://problem/3914078> worldclock crashing gc related

        Use ProtectedValue for Context2D instance members.

        Reviewed by Maciej.

        * khtml/ecma/kjs_html.h:

2004-12-09  John Sullivan  <sullivan@apple.com>

        Reviewed by Dave.
        
        - fixed <rdar://problem/3731099> Move AXTitle string for image elements to AXDescription

        * kwq/KWQAccObject.mm:
        (-[KWQAccObject title]): moved image alt tag code out of here
        (-[KWQAccObject accessibilityDescription]): moved image alt tag code into this new method
        (-[KWQAccObject accessibilityAttributeNames]): include AXDescription in the set of attributes
        that ordinary elements return; this means that ordinary elements that aren't images will return
        a nil description, which isn't ideal, but is in keeping with the way the rest of these attributes
        work here.
        (-[KWQAccObject accessibilityAttributeValue:]):
        call accessibilityDescription when asked for AXDescription

2004-12-09  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Fix for this bug:
        
        <rdar://problem/3910425> REGRESSION (Mail): Crash in ReplaceSelectionCommand; selection is empty, leading to null deref

        * khtml/editing/htmlediting.cpp:
        (khtml::MoveSelectionCommand::doApply): The node representing the destination for the move may have
        been deleted. If this is the case, set the destination to the node the delete command provides in
        its ending selection.

=== Safari-175 ===

2004-12-09  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John
        
        Fix for this bug:
        
        <rdar://problem/3912841> REGRESSION (173-TOT): Some images report 0x0 dimensions on layout tests, causes spurious test failures

        The new threaded image decoding capability can throw off layout tests. The issue is that the decoding
        callback may not be delivered before the program asks for the dimensions of an image in order to
        wrote the layout dimensions. More generally, I think we need to ensure that there are no races in
        layout tests, so I have added a flag to the render tree debug code that we can set when debugging.

        * kwq/KWQRenderTreeDebug.cpp:
        (debuggingRenderTree): New function. Returns flag which tells whether the program is debugging the render tree.
        (externalRepresentation): Sets debuggingRenderTree flag to true;
        * kwq/KWQRenderTreeDebug.h:
        * kwq/WebCoreImageRendererFactory.m:
        (+[WebCoreImageRendererFactory shouldUseThreadedDecoding]): Checks debuggingRenderTree flag and will not
        do threaded decoding in any case if the flag is set.

2004-12-09  Chris Blumenberg  <cblu@apple.com>

        Fix for busting XMLHTTPRequest.

        Reviewed by kocienda.

        * khtml/misc/loader.cpp:
        (Loader::servePendingRequests): pass true for deliverAllData
        * kwq/KWQKJob.h:
        (KIO::get): take deliverAllData param
        (KIO::http_post): ditto
        * kwq/KWQKJobClasses.h:
        * kwq/KWQKJobClasses.mm:
        (KIO::TransferJob::TransferJob): if deliverAllData, create signal with data param
        (KIO::TransferJob::emitResult): if deliverAllData, call signal with data param

2004-12-09  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        <rdar://problem/3911011> REGRESSION (Mail): Spaces at end of line causing word wrap lost when copied/pasted

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::renderedText): Fixed the code so that spaces at the end of lines are not skipped.

2004-12-07  Richard Williamson   <rjw@apple.com>

        Support threaded image decoding on machines w/ > 2 CPUs.

        Reviewed by Maciej and Chris.

        * khtml/misc/loader.cpp:
        (CachedImageCallback::notifyUpdate):
        (CachedImageCallback::notifyFinished):
        (CachedImageCallback::notifyDecodingError):
        (CachedImageCallback::handleError):
        (CachedImageCallback::clear):
        (CachedImage::CachedImage):
        (CachedImage::clear):
        (CachedImage::data):
        (CachedImage::checkNotify):
        (Loader::servePendingRequests):
        (Loader::slotFinished):
        (Loader::numRequests):
        (Loader::cancelRequests):
        (Loader::removeBackgroundDecodingRequest):
        * khtml/misc/loader.h:
        (khtml::CachedImageCallback::CachedImageCallback):
        (khtml::CachedImageCallback::ref):
        (khtml::CachedImageCallback::deref):
        (khtml::CachedImage::decoderCallback):
        * khtml/rendering/render_object.cpp:
        (RenderObject::setPixmap):
        * kwq/KWQPixmap.h:
        * kwq/KWQPixmap.mm:
        (-[WebImageCallback initWithCallback:khtml::]):
        (-[WebImageCallback _commonTermination]):
        (-[WebImageCallback dealloc]):
        (-[WebImageCallback finalize]):
        (-[WebImageCallback notify]):
        (-[WebImageCallback setImageSourceStatus:]):
        (-[WebImageCallback status]):
        (QPixmap::shouldUseThreadedDecoding):
        (QPixmap::receivedData):
        * kwq/WebCoreImageRenderer.h:
        * kwq/WebCoreImageRendererFactory.h:
        * kwq/WebCoreImageRendererFactory.m:
        (+[WebCoreImageRendererFactory shouldUseThreadedDecoding]):
        (+[WebCoreImageRendererFactory setShouldUseThreadedDecoding:]):

2004-12-07  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::doApply): Fix a problem with my
        change in thei code yesterday. Call to insertBlockPlaceholderIfNeeded() must be done
        after new block has been inserted, otherwise a crash can result. Shuffle down call 
        to insertBlockPlaceholderIfNeeded() a couple of lines (where the node is inserted), 
        and all is good.

2004-12-07  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3907422> REGRESSION (Mail): Pasting quoted content can place content after body element

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplaceSelectionCommand::doApply): Detect when the body element is the "reference block" used
        for determining the location for inserting content. Do not allow an insert before or after if the
        reference block is the body. Perform insertNodeAt(0) and appendNode, respectively, in the block-is-body case.
        * layout-tests/editing/inserting/insert-3907422-fix-expected.txt: Added.
        * layout-tests/editing/inserting/insert-3907422-fix.html: Added.

2004-12-07  Darin Adler  <darin@apple.com>

        Reviewed by Don.

        - fixed <rdar://problem/3908701> REGRESSION: Cursor does not change to "hand" over active links

        * kwq/KWQEvent.mm: (positionForEvent): Add NSMouseMoved to list of events that have mouse location.

2004-12-07  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Changed name of constant from KHTMLInterchangeNewline to AppleInterchangeNewline.
        I discussed this with Hyatt and he agreed that going with "Apple" names was OK.

        * khtml/editing/html_interchange.h: Name change, as described above.
        * khtml/editing/htmlediting.cpp:
        (khtml::ReplacementFragment::isInterchangeNewlineComment): Ditto.
        * khtml/xml/dom2_rangeimpl.cpp:
        (DOM::RangeImpl::toHTML): Ditto.

2004-12-07  Ken Kocienda  <kocienda@apple.com>

        Reviewed by me
        
        Added a couple more layout tests.

        * layout-tests/editing/deleting/delete-line-013-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-013.html: Added.
        * layout-tests/editing/deleting/delete-line-014-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-014.html: Added.

2004-12-06  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Hyatt.

        - fixed <rdar://problem/3906974> assertion failure in QWidget::beforeMouseDown clicking on <select multiple>
        
        * kwq/KWQListBox.mm:
        (QListBox::QListBox): Initialize KWQListBoxScrollView with this.
        (-[KWQListBoxScrollView initWithListBox:]): Make this class a KWQWidgetHolder.
        (-[KWQListBoxScrollView widget]): See above.
        (-[KWQTableView mouseDown:]): Pass outerView rather than self to beforeMouseDown and
        afterMouseDown, to avoid triggering an assertion failure.

2004-12-06  David Hyatt  <hyatt@apple.com>

        Fix for 3615411, the linesAppended optimization was old and broken, and it's easier with the new code fixes
        made by me, kocienda and harrison to just remove it.
        
        Reviewed by mjs

        * khtml/rendering/bidi.cpp:
        (khtml::RenderBlock::layoutInlineChildren):
        * khtml/rendering/render_block.cpp:
        (khtml:::RenderFlow):
        * khtml/rendering/render_block.h:
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::dirtyLinesFromChangedChild):

2004-12-06  David Hyatt  <hyatt@apple.com>

        Fix for 3787133, some web pages print with many blank pages.  Make sure to use the real page print rect and
        not a damage rect that can be changed when intersected with the clip regions of the web page.
        
        Reviewed by john

        * khtml/rendering/render_canvas.h:
        (khtml::RenderCanvas::printRect):
        (khtml::RenderCanvas::setPrintRect):
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::paintLines):
        * khtml/rendering/render_list.cpp:
        (RenderListMarker::paint):
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::adjustPageHeight):

2004-12-06  David Harrison  <harrison@apple.com>

        Reviewed by Ken Kocienda and Dave Hyatt (OOPS!).

                <rdar://problem/3849947> Typing after pasting line does not appear until after window resize.
                
                
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::dirtyLinesFromChangedChild):
                Dirty the line above because new child can inval the cached line break position of previous line.

2004-12-06  David Hyatt  <hyatt@apple.com>

        Fix for 3254464, radio buttons do not work for quiz on netscape.com.  Left/top overflow needed to be implemented.
        This also fixes 3106907, link hover color only partially set on rollover and the more general architecture bug
        3126929, handle top/left overflow.

        This patch also fixes 3902891, scroll bar of position:fixed content moves when a page is scrolled.

        Finally, the Emerson regression 3869718 (error involving computing the rightmost/lowest position of overflow:auto
        regions and web pages) has been fixed.
        
        Reviewed by mjs

        * khtml/rendering/bidi.cpp:
        (khtml::RenderBlock::computeHorizontalPositionsForLine):
        (khtml::RenderBlock::checkLinesForOverflow):
        * khtml/rendering/render_block.cpp:
        (khtml:::RenderFlow):
        (khtml::RenderBlock::overflowHeight):
        (khtml::RenderBlock::overflowWidth):
        (khtml::RenderBlock::overflowLeft):
        (khtml::RenderBlock::overflowTop):
        (khtml::RenderBlock::overflowRect):
        (khtml::RenderBlock::layoutBlock):
        (khtml::RenderBlock::layoutBlockChildren):
        (khtml::RenderBlock::paint):
        (khtml::RenderBlock::floatRect):
        (khtml::RenderBlock::lowestPosition):
        (khtml::RenderBlock::rightmostPosition):
        (khtml::RenderBlock::leftmostPosition):
        (khtml::RenderBlock::nodeAtPoint):
        * khtml/rendering/render_block.h:
        * khtml/rendering/render_box.h:
        (khtml::RenderBox::borderBox):
        (khtml::RenderBox::borderTopExtra):
        (khtml::RenderBox::borderBottomExtra):
        * khtml/rendering/render_layer.cpp:
        (RenderLayer::paintScrollbars):
        (mustExamineRenderer):
        (RenderLayer::intersectsDamageRect):
        (RenderLayer::containsPoint):
        * khtml/rendering/render_line.cpp:
        (khtml::InlineFlowBox::placeBoxesHorizontally):
        (khtml::InlineFlowBox::verticallyAlignBoxes):
        * khtml/rendering/render_line.h:
        (khtml::InlineBox::leftOverflow):
        (khtml::InlineBox::rightOverflow):
        (khtml::InlineFlowBox::setVerticalOverflowPositions):
        (khtml::RootInlineBox::RootInlineBox):
        (khtml::RootInlineBox::leftOverflow):
        (khtml::RootInlineBox::rightOverflow):
        (khtml::RootInlineBox::setVerticalOverflowPositions):
        (khtml::RootInlineBox::setHorizontalOverflowPositions):
        * khtml/rendering/render_object.h:
        (khtml::RenderObject::borderBox):
        (khtml::RenderObject::overflowLeft):
        (khtml::RenderObject::overflowTop):
        (khtml::RenderObject::overflowRect):
        (khtml::RenderObject::floatRect):
        * khtml/rendering/render_table.cpp:
        (RenderTable::layout):
        (RenderTable::paint):
        (RenderTable::paintBoxDecorations):
        (RenderTable::calcMinMaxWidth):
        * khtml/rendering/render_table.h:
        (khtml::RenderTableCell::borderTopExtra):
        (khtml::RenderTableCell::borderBottomExtra):
        * kwq/KWQRect.h:
        * kwq/KWQRect.mm:
        (QRect::inflate):

2004-12-06  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by John.

        - fixed <rdar://problem/3903797> scripts can cause other frames/windows to execute arbitrary script using javascript: URLs
        
        I changed all unprotected places that can navigate a different
        window or frame from script to check for a javascript: URL, and if
        found, to check for safety using cross-site-script rules. 

        I considered a few other possible exploits and made no change:

        - document.location is already protected because the document
        object itself is protected

        - frame.src, frame.location, iframe.src and targetted links are
        all safe because setting the URL of a frame to a javascript: URL
        executes the script in the context of the parent

        * khtml/ecma/kjs_window.cpp:
        (WindowFunc::tryCall):
        (Location::put):
        (LocationFunc::tryCall):

2004-12-06  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Maciej

        Fix for this bug:
        
        <rdar://problem/3890955> 8A314: Forward delete sometimes fails to delete the selected quoted text

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::handleSpecialCaseBRDelete): Fixed bonehead coding mistake in the
        check for one of the special cases being checked for in this function. The specific case 
        intends to check for a selection that is only a <br> after a block ends (as in </div><br>). If it
        sees such markup, it deletes only the <br> and bails. However, this code would run in *any*
        case where a selection ended in a <br> after a block and would not delete any part of the
        selection preceding the <br>. Bad. I have tightened the check to see that only a <br> is
        selected.

        Fixing the bug above was accomplished with an additional call to DOM::Position::downstream. This
        new use of the function exposed this bug:
        
        <rdar://problem/3907666> Incorrectly coded loop in Position::downstream can lead to infinite loop

        * khtml/xml/dom_position.cpp:
        (DOM::Position::downstream): I am ashamed of my first cut at this. Rewrote the loop so it does 
        not have this fatal flaw. It is a much better design as well.
        
        * layout-tests/editing/deleting/delete-3800834-fix-expected.txt: Changes made this test
        have what I consider to be a better result. Going with it.

2004-12-06  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3871718> REGRESSION (125-168): text marked bold with font that does not have bold variant copies as non-bold

        Reviewed by hyatt.

        * kwq/DOM.mm:
        (-[DOMElement _font]): new SPI for AppKit
        * kwq/DOMPrivate.h:

2004-12-06  Darin Adler  <darin@apple.com>

        Reviewed by Maciej.

        - fixed <rdar://problem/3906327> Select All of a large document is slow (>15 secs on my machine for attached specimen)

        * kwq/KWQScrollView.mm: (QScrollView::updateContents): Intersect with visibleRect before calling through
        to NSView to dirty; NSView could also be more efficient in this case (I filed 3906343).

2004-12-06  John Sullivan  <sullivan@apple.com>

        Darin found what appears to be the real leak that we were falsely blaming
        on the 'leaks' tool (3880245). I made the change, and ran layout tests and PLT to make
        sure nothing barfed.

        * khtml/css/cssparser.cpp:
        (CSSParser::parseValue):
        call clearProperties() instead of just setting numParsedProperties to 0
        (CSSParser::parseDeclaration):
        ditto
        (CSSParser::createStyleDeclaration):
        ditto

2004-12-06  Ken Kocienda  <kocienda@apple.com>

        Reviewed by me
        
        New layout tests.

        * layout-tests/editing/inserting/insert-div-001-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-001.html: Added.
        * layout-tests/editing/inserting/insert-div-002-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-002.html: Added.
        * layout-tests/editing/inserting/insert-div-003-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-003.html: Added.
        * layout-tests/editing/inserting/insert-div-004-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-004.html: Added.
        * layout-tests/editing/inserting/insert-div-005-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-005.html: Added.
        * layout-tests/editing/inserting/insert-div-006-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-006.html: Added.
        * layout-tests/editing/inserting/insert-div-007-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-007.html: Added.
        * layout-tests/editing/inserting/insert-div-008-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-008.html: Added.
        * layout-tests/editing/inserting/insert-div-009-expected.txt: Added.
        * layout-tests/editing/inserting/insert-div-009.html: Added.

2004-12-06  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Fix for this bug:
        
        <rdar://problem/3906948> REGRESSION (Mail): Insert paragraph code can make the insertion point "stick" in place.

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorCommand::doApply): Call insertBlockPlaceholderIfNeeded(), passing block
        being added to this function. This ensures that the added block has a height.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::doApply): Ditto.

2004-12-06  Ken Kocienda  <kocienda@apple.com>
 
        Reviewed by John
        
        * khtml/dom/dom_string.cpp:
        (DOM::DOMString::substring): Expose method already on DOMStrimgImpl.
        * khtml/dom/dom_string.h: Ditto.
        * khtml/editing/htmlediting.cpp:
        (khtml::CompositeEditCommand::rebalanceWhitespace): New helper to create and execute a
        RebalanceWhitespaceCommand instance.
        (khtml::DeleteSelectionCommand::doApply): Call rebalanceWhitespace() after running command.
        (khtml::InsertLineBreakCommand::doApply): Ditto.
        (khtml::InsertParagraphSeparatorCommand::doApply): Ditto.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::doApply): Ditto.
        (khtml::InsertTextCommand::input): Ditto.
        (khtml::RebalanceWhitespaceCommand::RebalanceWhitespaceCommand): New command.
        (khtml::RebalanceWhitespaceCommand::~RebalanceWhitespaceCommand): Ditto.
        (khtml::RebalanceWhitespaceCommand::doApply): Ditto.
        (khtml::RebalanceWhitespaceCommand::doUnapply): Ditto.
        (khtml::RebalanceWhitespaceCommand::preservesTypingStyle): Ditto.
        (khtml::ReplaceSelectionCommand::completeHTMLReplacement): Ditto.
        * khtml/editing/htmlediting.h: Ditto.
        (khtml::RebalanceWhitespaceCommand::): Ditto.

2004-12-05  Darin Adler  <darin@apple.com>

        - fixed small problem in my check-in from yesterday

        * kwq/KWQEvent.mm:
        (positionForEvent): Get location from event without raising exception if it's the wrong type.
        (clickCountForEvent): Same, for clickCount.
        (QMouseEvent::QMouseEvent): Use the new helper functions so this can be constructed even with
        the wrong type of NSEvent. Required for cases where a keyboard event causes a "click" and we need
        to synthesize a QMouseEvent for KHTML internal use, using the key down NSEvent.

2004-12-04  Darin Adler  <darin@apple.com>

        Reviewed by John.

        - fixed <rdar://problem/3878329> REGRESSION (169-170): colors are wrong for my.yahoo.com due to CSS background parsing changes

        * khtml/css/cssparser.cpp: (CSSParser::parseBackgroundShorthand): Changed function so it doesn't rely on the position
        attribute being at the end of the array and then moved position attribute before color attribute so it takes precedence.
        Since "0" can be both the X coordinate of a position and a legal color (meaning black), we need to do position first.

        - fixed <rdar://problem/3760869> click events for input type=button or type=checkbox don't have flags like shiftKey set

        * khtml/rendering/render_form.h: Remove unused RenderFormElement fields.
        * khtml/rendering/render_form.cpp:
        (RenderFormElement::RenderFormElement): Take out code to set a bunch of unused fields.
        (RenderFormElement::slotClicked): Change to create the QMouseEvent from the actual mouse event rather than
        creating it with all the flags set to 0, using the new QMouseEvent constructor that does so.

        * kwq/KWQEvent.h: Added constructor that takes no parameters which uses the current event from AppKit.
        Made the click count getter const and added an isDoubleClick that matches the logic used elsewhere.
        Added a fixState helper method so the constructors can save code.
        * kwq/KWQEvent.mm:
        (QMouseEvent::QMouseEvent): Factored out the state-fixing code that was in the two existing constructors
        and added a third constructor that uses the "current event" from AppKit (used above).
        (QMouseEvent::fixState): Compute state and click count based on event type.

        - fixed first symptom of <rdar://problem/3830936> REGRESSION (125-165): crash due to null font family, hang at changeforamerica.com

        * kwq/KWQFontFamily.mm: (KWQFontFamily::getNSFamily): Handle empty strings specially so we don't run into trouble when
        the family name is a null string. This prevents the crash, but there are still other problems that may have the same
        underlying cause in CSS.

        - fixed <rdar://problem/3829808> Safari crashes when adding a DOM node that was removed from an XMLHTTP request result

        * khtml/html/html_baseimpl.cpp: (HTMLBodyElementImpl::insertedIntoDocument):
        Added nil check before calling scheduleRelayout. This is new code so the nil-dereference is a recent regression.

2004-12-03  Chris Blumenberg  <cblu@apple.com>

        New fixes for:
        <rdar://problem/3685766> WebDataSource is missing subresources when they use cached WebCore data
        <rdar://problem/3722434> REGRESSION?: Assertion failure trying to drag image in iframe (itapema.sc.gov.br)
        <rdar://problem/3903173> REGRESSION (172-TOT): assertion failure and crash in slotAllData logging into hotmail account
        <rdar://problem/3902749> REGRESSION (Tiger): missing image symbol does not appear

        Reviewed by darin, rjw, kocienda.

        * khtml/misc/loader.cpp:
        (CachedObject::~CachedObject):
        (CachedCSSStyleSheet::checkNotify):
        (Loader::servePendingRequests):
        (Loader::slotFinished):
        (Loader::slotReceivedResponse):
        (Cache::requestImage):
        (Cache::requestScript):
        * khtml/misc/loader.h:
        (khtml::CachedObject::CachedObject):
        (khtml::CachedObject::response):
        (khtml::CachedObject::allData):
        * kwq/KWQKJobClasses.h:
        * kwq/KWQKJobClasses.mm:
        (KIO::TransferJobPrivate::TransferJobPrivate):
        (KIO::TransferJobPrivate::~TransferJobPrivate):
        (KIO::TransferJob::TransferJob):
        (KIO::TransferJob::assembleResponseHeaders):
        (KIO::TransferJob::retrieveCharset):
        (KIO::TransferJob::emitResult):
        (KIO::TransferJob::emitReceivedResponse):
        * kwq/KWQLoader.h:
        * kwq/KWQLoader.mm:
        (KWQHeaderStringFromDictionary):
        (KWQCheckCacheObjectStatus):
        (KWQIsResponseURLEqualToURL):
        (KWQResponseURL):
        (KWQResponseMIMEType):
        (KWQCacheObjectExpiresTime):
        (khtml::CachedObject::setResponse):
        (khtml::CachedObject::setAllData):
        * kwq/KWQPixmap.h:
        * kwq/KWQPixmap.mm:
        (QPixmap::QPixmap):
        * kwq/KWQResourceLoader.mm:
        (-[KWQResourceLoader finishJobAndHandle:]):
        (-[KWQResourceLoader cancel]):
        (-[KWQResourceLoader reportError]):
        (-[KWQResourceLoader finishWithData:]):
        * kwq/KWQSignal.h:
        * kwq/KWQSignal.mm:
        (KWQSignal::call):
        * kwq/KWQSlot.h:
        * kwq/KWQSlot.mm:
        (KWQSlot::KWQSlot):
        (KWQSlot::call):
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreResourceLoader.h:

2004-12-04  Darin Adler  <darin@apple.com>

        Reviewed by John.

        - fixed <rdar://problem/3876093> REGRESSION (166-167): Setting slider control's value from JavaScript has no effect (breaks RSS)

        * khtml/rendering/render_form.cpp:
        (RenderSlider::updateFromElement): Call setValue to update the value of the DOM element rather than
        modifying the m_value data member directly. We don't use m_value at all for sliders now, and in fact
        the code relies on the fact that m_value is null. Setting m_value to a non-null value was causing the bug.
        (RenderSlider::slotSliderValueChanged): Ditto.

2004-12-03  John Sullivan  <sullivan@apple.com>

        Reviewed by Ken.
        
        - fixed <rdar://problem/3889411> REGRESSION (125-172): repro crash in 
        khtml::BackgroundLayer::cullEmptyLayers

        * khtml/rendering/render_style.cpp:
        (BackgroundLayer::cullEmptyLayers):
        added missing nil check

=== Safari-173 ===

2004-12-03  Ken Kocienda  <kocienda@apple.com>

        Reviewed by me

        Roll out some recent changes by Chris that caused a performance regression.
        Fix is in hand, but it is a little risky this close to a submission. So,
        we have decided to roll back the change with the regression and roll in
        the new code after we submit.

        * khtml/css/cssproperties.c:
        (hash_prop):
        (findProp):
        * khtml/css/cssvalues.c:
        (hash_val):
        (findValue):
        * khtml/misc/htmlattrs.c:
        (hash_attr):
        (findAttr):
        * khtml/misc/htmltags.c:
        (hash_tag):
        (findTag):
        * khtml/misc/loader.cpp:
        (CachedObject::~CachedObject):
        (CachedObject::setResponse):
        (CachedCSSStyleSheet::checkNotify):
        (Loader::servePendingRequests):
        (Loader::slotFinished):
        (Loader::slotReceivedResponse):
        (Cache::requestImage):
        (Cache::requestScript):
        * khtml/misc/loader.h:
        (khtml::CachedObject::CachedObject):
        (khtml::CachedObject::response):
        * kwq/KWQKJobClasses.h:
        * kwq/KWQKJobClasses.mm:
        (KIO::TransferJobPrivate::TransferJobPrivate):
        (KIO::TransferJobPrivate::~TransferJobPrivate):
        (KIO::TransferJob::TransferJob):
        (KIO::TransferJob::assembleResponseHeaders):
        (KIO::TransferJob::retrieveCharset):
        (KIO::TransferJob::emitResult):
        (KIO::TransferJob::emitReceivedResponse):
        * kwq/KWQLoader.h:
        * kwq/KWQLoader.mm:
        (KWQHeaderStringFromDictionary):
        (KWQCheckCacheObjectStatus):
        (KWQRetainResponse):
        (KWQReleaseResponse):
        (KWQIsResponseURLEqualToURL):
        (KWQResponseURL):
        (KWQResponseMIMEType):
        (KWQResponseTextEncodingName):
        (KWQResponseHeaderString):
        (KWQCacheObjectExpiresTime):
        (KWQLoader::KWQLoader):
        * kwq/KWQPixmap.h:
        * kwq/KWQPixmap.mm:
        (QPixmap::QPixmap):
        * kwq/KWQResourceLoader.mm:
        (-[KWQResourceLoader finishJobAndHandle]):
        (-[KWQResourceLoader cancel]):
        (-[KWQResourceLoader reportError]):
        (-[KWQResourceLoader finish]):
        * kwq/KWQSignal.h:
        * kwq/KWQSignal.mm:
        (KWQSignal::call):
        * kwq/KWQSlot.h:
        * kwq/KWQSlot.mm:
        (KWQSlot::KWQSlot):
        (KWQSlot::call):
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreResourceLoader.h:

2004-12-03  John Sullivan  <sullivan@apple.com>

        Reviewed by Chris.
        
        - fixed <rdar://problem/3903990> can't tab to all items on www.google.com any more (other pages too?)

        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::nextKeyViewInFrameHierarchy):
        when checking whether we moved the focus to another view, make sure we didn't "move" it to
        our documentView, because that's no move at all.

2004-12-03  Darin Adler  <darin@apple.com>

        Reviewed by Vicki.

        - fixed <rdar://problem/3901109> REGRESSION (171-172): repro crash in DOM::NodeImpl::setChanged at chick-fil-a.com

        * khtml/css/css_valueimpl.cpp: (DOM::CSSMutableStyleDeclarationImpl::CSSMutableStyleDeclarationImpl):
        Added missing initialization for base class and node pointer.

        - fixed a few places that could leave dangling node pointers

        * khtml/html/html_baseimpl.cpp: (HTMLBodyElementImpl::~HTMLBodyElementImpl):
        Clear out the node pointer when the node is destroyed.
        * khtml/html/html_elementimpl.cpp: (HTMLElementImpl::~HTMLElementImpl): Ditto.

2004-12-03  Chris Blumenberg  <cblu@apple.com>

        Fix for performance regression. My original patch added a signal for passing the data of a resource to its WebCore cache object. This patch passes the data with the preexisting "finished" symbol so we make less calls. 
        Fixed: <rdar://problem/3903173> REGRESSION (172-TOT): assertion failure and crash in slotAllData logging into hotmail account

        Reviewed by rjw.

        * khtml/misc/loader.cpp:
        (Loader::servePendingRequests): pass data param to slotFinished, removed allData signal
        (Loader::slotFinished): take data param
        * khtml/misc/loader.h:
        * kwq/KWQKJobClasses.h:
        * kwq/KWQKJobClasses.mm:
        (KIO::TransferJob::TransferJob): have m_result take a data param, removed m_allData
        (KIO::TransferJob::emitResult): take data param and pass it
        * kwq/KWQResourceLoader.mm:
        (-[KWQResourceLoader finishJobAndHandle:]): take data param and pass it
        (-[KWQResourceLoader cancel]): pass nil for data
        (-[KWQResourceLoader reportError]): ditto
        (-[KWQResourceLoader finishWithData:]): pass data
        * kwq/KWQSlot.mm:
        (KWQSlot::KWQSlot): pass data param to slotFinished
        (KWQSlot::call): added support for slotFinished_Loader, removed slotAllData

2004-12-03  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Did some clean up in the Position class as a result of trying to write some new layout
        tests and discovering a bug along the way.

        I removed these three functions from the Position class:
        
        1. bool isFirstRenderedPositionOnLine() const;
        2. bool isLastRenderedPositionOnLine() const;
        3. static bool renderersOnDifferentLine(RenderObject *r1, long o1, RenderObject *r2, long o2);
        4. bool inFirstEditableInRootEditableElement() const;

        The first two have replacements in the VisiblePosition class, and some code has been
        moved to use these new variants. The third function was a helper used only by these
        first two function, and can be removed as well. The fourth function was not used by anyone.

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertTextCommand::input): Change over to use VisiblePosition isFirstVisiblePositionOnLine().
        * khtml/editing/visible_position.cpp:
        (khtml::visiblePositionsOnDifferentLines): Added an additional check for blocks to this function.
        Incorrect results were being returned when asking about positions at the starts of blocks.
        * khtml/xml/dom_position.cpp:
        (DOM::Position::previousCharacterPosition): Change over to use VisiblePosition isFirstVisiblePositionOnLine().
        (DOM::Position::nextCharacterPosition): Change over to use VisiblePosition isLastVisiblePositionOnLine().
        (DOM::Position::rendersInDifferentPosition): Removed use of #3 helper in a log message. We can live without it.
        * khtml/xml/dom_position.h: Update header for deletions.

2004-12-03  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Terminology change in execCommand command identifiers. Specifically, the name of 
        "InsertNewline" command has been changed to "InsertLineBreak". This matches the 
        terminology used by AppKit. It is also more accurate, since the insertion of a
        "br" element is what the command does. The inspiration for this change is so the 
        -insertNewline AppKit method can be mapped to insert a new "div" element in 
        a document and avoid ambiguity with what the javascript editing command does.
        
        * khtml/editing/jsediting.cpp
        * layout-tests/editing/deleting/delete-tab-004.html
        * layout-tests/editing/editing.js
        * layout-tests/editing/inserting/insert-3654864-fix.html
        * layout-tests/editing/inserting/insert-3659587-fix.html
        * layout-tests/editing/inserting/insert-3775316-fix.html
        * layout-tests/editing/inserting/insert-3800346-fix.html
        * layout-tests/editing/inserting/insert-br-001.html
        * layout-tests/editing/inserting/insert-br-002.html
        * layout-tests/editing/inserting/insert-br-003.html
        * layout-tests/editing/inserting/insert-br-004.html
        * layout-tests/editing/inserting/insert-br-005.html
        * layout-tests/editing/inserting/insert-br-006.html
        * layout-tests/editing/inserting/insert-br-007.html
        * layout-tests/editing/inserting/insert-br-008.html
        * layout-tests/editing/inserting/insert-tab-004.html
        * layout-tests/editing/inserting/insert-text-with-newlines.html
        * layout-tests/editing/pasteboard/paste-text-010.html

2004-12-02  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3786362> REGRESSION (Mail): pasted text loses one newline

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertLineBreakCommand::doApply): Added check for strict mode before adding an extra br element
        at the end of a block. This is only necessary in quirks mode. Also, lower-case "br" used to make element.
        (khtml::ReplaceSelectionCommand::doApply): If the replacement adds a br element as the last element
        in a block and the document is in quirks mode, add an additional br to make the one in the 
        replacement content show up. This turns out to be much the same logic as is done in InsertLineBreakCommand.
        * layout-tests/editing/inserting/insert-3786362-fix-expected.txt: Added.
        * layout-tests/editing/inserting/insert-3786362-fix.html: Added.

2004-12-02  Richard Williamson   <rjw@apple.com>

        Fixed <rdar://problem/3841332> REGRESSION (125.9-167u): repro crash in -[KWQPageState invalidate] involving .Mac images

        Ensure that the document is cleared when leaving a non-HTML page.  This ensures that
        the b/f cache won't incorrectly trash the previous state when restoring.

        Reviewed by John.

        * kwq/WebCoreBridge.h:
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge openURL:reload:contentType:refresh:lastModified:pageCache:]):
        (-[WebCoreBridge canCachePage]):
        (-[WebCoreBridge clear]):

2004-12-02  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Richard

        Fix for this bug:
        
        <rdar://problem/3857775> 8A293: Mail.app crashes converting copy-pasted text into plain text

        * khtml/xml/dom2_rangeimpl.cpp:
        (DOM::RangeImpl::commonAncestorContainer): Return the document element if no common ancestor container
        was found. This can happen in cases where the DOM was built from malformed markup (as in the case
        of this bug where there is content after the body tag). Did a little code clean up as well.
        (DOM::RangeImpl::compareBoundaryPoints): Made code more robust by adding some null checks.

2004-12-02  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Chris

        Fix for this bug:
        
        <rdar://problem/3668157> REGRESSION (Mail): shift-click deselects when selection was created right-to-left

        * khtml/khtml_part.cpp:
        (KHTMLPart::handleMousePressEventSingleClick): Use RangeImpl::compareBoundaryPoints
        to figure out which end of the selection to extend.

2004-12-02  David Harrison  <harrison@apple.com>

        Reviewed by Ken Kocienda.

                <rdar://problem/3834917> REGRESSION (Mail): double-clicking blank line selects end of previous line
                Fixed originally reported bug plus the case of double-clicking whitespace at the beginning of a line, which has a similar result.

        * khtml/editing/visible_text.cpp:
        (khtml::SimplifiedBackwardsTextIterator::SimplifiedBackwardsTextIterator):
        (khtml::SimplifiedBackwardsTextIterator::handleTextNode):
        (khtml::SimplifiedBackwardsTextIterator::handleReplacedElement):
        (khtml::SimplifiedBackwardsTextIterator::handleNonTextNode):
        (khtml::SimplifiedBackwardsTextIterator::emitCharacter):
        Distinguish BR from whitespace.
        * khtml/editing/visible_text.h:
        Distinguish BR from whitespace.
        * khtml/editing/visible_units.cpp:
        (khtml::previousWordBoundary):
        Use UPSTREAM visible position now that SimplifiedBackwardsTextIterator distinguishes BR from whitespace.  Otherwise, double-clicking at end of line would result in caret selection at start of next line. 

2004-12-02  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3900996> Crash dragging past end of contentEditable DIV, at DOM::RangeImpl::pastEndNode() const + 24

        * khtml/xml/dom_position.cpp:
        (DOM::Position::equivalentRangeCompliantPosition): Fixed this function so that it constrains the offset
        of the position to be >= 0 and <= number of kids of its node. Not doing this constraining led to a DOM
        exception trying to use a Position returned from this function to set the boundary point of a Range (which
        eventually led to the crash). Since this crash happened, it seems like this function was failing in its
        contract to return a range-compliant position, hence the need for this fix.

2004-12-01  Ken Kocienda  <kocienda@apple.com>

        Reviewed by me

        Moving code only.

        * khtml/editing/htmlediting.cpp: Move ReplaceSelectionCommand into alphabetical order with
        regard to other editing commands. The class had a name change ages ago, and it was never
        moved.
        * khtml/editing/htmlediting.h: Ditto.

2004-12-01  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Hyatt
        
        Some improvements for paste, including some new code to annotate
        whitespace when writing to the pasteboard to ensure that the meaning
        of the markup on the pasteboard is unambiguous.
        
        There is also new code for reading this annotated markup from the pasteboard,
        removing the nodes that were added only to prevent ambiguity.

        * WebCore.pbproj/project.pbxproj: Added html_interchange.h and html_interchange.cpp files.
        The header should have been added earlier, but I did not do so.
        * khtml/editing/html_interchange.cpp: Added.
        (convertHTMLTextToInterchangeFormat):
        * khtml/editing/html_interchange.h: Added some new constants for use with whitespace annotations.
        * khtml/editing/htmlediting.cpp:
        (khtml::ReplacementFragment::ReplacementFragment): Now looks for and removes annotations added for whitespace.
        Also fixed a bug in the code that counts blocks in a fragment.
        (khtml::ReplacementFragment::isInterchangeConvertedSpaceSpan): New helper. Recognizes annotation spans.
        (khtml::ReplacementFragment::insertNodeBefore): New helper.
        (khtml::ReplaceSelectionCommand::doApply): Fixed a bug in the code that sets the start position
        for the replacement after deleting. This was causing a bug when pasting at the end of a block.
        * khtml/editing/htmlediting.h: Add some new declarations.
        * khtml/xml/dom2_rangeimpl.cpp:
        (DOM::RangeImpl::toHTML): Calls to startMarkup now pass true for the new annotate flag.
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::stringValueForRange): New helper.
        (NodeImpl::renderedText): New helper to return only the rendered text in a node.
        (NodeImpl::startMarkup): Now takes an additional flag to control whether interchange annotations
        should be added. Called by the paste code.
        * khtml/xml/dom_nodeimpl.h: Added and modified function declarations.

        New test to check the khtml::ReplaceSelectionCommand::doApply fix.
        * layout-tests/editing/pasteboard/paste-text-010-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-010.html: Added.

2004-11-30  Chris Blumenberg  <cblu@apple.com>

        * ChangeLog: removed conflict marker

2004-11-30  Chris Blumenberg  <cblu@apple.com>

        Fixed:
        <rdar://problem/3685766> WebDataSource is missing subresources when they use cached WebCore data
        <rdar://problem/3722434> REGRESSION?: Assertion failure trying to drag image in iframe (itapema.sc.gov.br)

        Reviewed by darin.

        * khtml/misc/loader.cpp:
        (CachedObject::~CachedObject): release m_allData
        (CachedObject::setAllData): new
        (Loader::servePendingRequests): connect slotAllData
        (Loader::slotAllData): new
        (Cache::requestImage): tweak
        * khtml/misc/loader.h:
        (khtml::CachedObject::CachedObject): set allData to 0
        (khtml::CachedObject::allData): new
        * kwq/KWQKJobClasses.h:
        * kwq/KWQKJobClasses.mm:
        (KIO::TransferJob::TransferJob): set m_allData
        (KIO::TransferJob::emitAllData): new
        * kwq/KWQLoader.mm:
        (KWQCheckCacheObjectStatus): pass WebKit the data instead of the length of the resource
        * kwq/KWQResourceLoader.mm:
        (-[KWQResourceLoader finishWithData:]): renamed to pass all data for the resource
        * kwq/KWQSlot.mm:
        (KWQSlot::KWQSlot): support for slotAllData
        (KWQSlot::call):
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreResourceLoader.h:

2004-11-30  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by John.

2004-11-30  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by John.

        <rdar://problem/3805311> REGRESSION (159-163): onload in dynamically written document not called (causes blank search page at Japanese EPP site, many others)

        * khtml/khtml_part.cpp:
        (KHTMLPart::begin): call setParsing on document here after opening
        - from now on we'll only set parsing to true for a document open
        caused by page loading, not a programmatic one.
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::open): don't setParsing to true here any more.

2004-11-30  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by John.

        - fix recent regression from collection perf fixes.
        
        * khtml/html/html_miscimpl.cpp:
        (HTMLFormCollectionImpl::updateNameCache): Look up the name
        attribute in the name cache, not the id cache (d'oh!)

2004-11-30  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        - rolled in a KDE fix for a problem that may underlie a number of crashes

        * khtml/xml/dom2_rangeimpl.cpp: (RangeImpl::compareBoundaryPoints): Rolled in a change from
        the KDE guys to fix a subtle problem. Code said "n = n =".

        - rolled in a KDE fix for a containingBlock crash

        * khtml/rendering/render_object.cpp: Roll in a change from KDE that adds frameset to the list of
        elements that can not be a containingBlock. They said this fixes a crash, although I did not look
        into that further.

        - fixed <rdar://problem/3884660> 8A305: Repro crash in QScrollBar::setValue (affects Safari RSS)

        * kwq/KWQButton.mm:
        (-[KWQButton initWithQButton:]): Set up target and action here instead of in caller.
        (-[KWQButton detachQButton]): Added.
        (-[KWQButton sendConsumedMouseUpIfNeeded]): Check button for nil instead of checking target.
        (-[KWQButton mouseDown:]): Add calls to QWidget::beforeMouseDown/afterMouseDown.
        (-[KWQButton widget]): Added.
        (-[KWQButton becomeFirstResponder]): Added check to handle when button is 0.
        (-[KWQButton resignFirstResponder]): Ditto.
        (-[KWQButton canBecomeKeyView]): Ditto.
        (QButton::QButton): Remove target and action setup; handled in KWQButton now.
        (QButton::~QButton): Call detachQButton instead of setTarget:nil.

        * kwq/KWQComboBox.mm:
        (QComboBox::~QComboBox): Call detachQComboBox.
        (-[KWQPopUpButtonCell detachQComboBox]): Added.
        (-[KWQPopUpButtonCell trackMouse:inRect:ofView:untilMouseUp:]): Handle case where box is 0.
        (-[KWQPopUpButtonCell setHighlighted:]): Ditto.
        (-[KWQPopUpButton action:]): Ditto.
        (-[KWQPopUpButton widget]): Tweaked.
        (-[KWQPopUpButton mouseDown:]): Added. Calls QWidget::beforeMouseDown/afterMouseDown.
        (-[KWQPopUpButton becomeFirstResponder]): Handle case where widget is 0.
        (-[KWQPopUpButton resignFirstResponder]): Ditto.
        (-[KWQPopUpButton canBecomeKeyView]): Ditto.

        * kwq/KWQLineEdit.mm: (QLineEdit::~QLineEdit): Updated to use new detachQLineEdit name.

        * kwq/KWQListBox.mm:
        (-[KWQTableView mouseDown:]): Added. Calls QWidget::beforeMouseDown/afterMouseDown.

        * kwq/KWQScrollBar.h: Removed m_scroller field.
        * kwq/KWQScrollBar.mm:
        (-[KWQScrollBar initWithQScrollBar:]): Rearranged a little bit.
        (-[KWQScrollBar detachQScrollBar]): Added.
        (-[KWQScrollBar widget]): Added.
        (-[KWQScrollBar mouseDown:]): Added.  Calls QWidget::beforeMouseDown and afterMouseDown.
        (QScrollBar::QScrollBar): Changed to no longer set m_scroller.
        (QScrollBar::~QScrollBar): Changed to call detachQScrollBar. No longer calls removeFromSuperview.
        (QScrollBar::setValue): Chagned to use getView instad of m_scrollBar.
        (QScrollBar::setKnobProportion): Ditto.
        (QScrollBar::scrollbarHit): Ditto.

        * kwq/KWQScrollView.mm:
        (QScrollView::addChild): Changed to call QWidget to add to superview to accomodate the
        hack where we don't remove right away when doing mouse tracking.
        (QScrollView::removeChild): Changed to call QWidget to remove from superview to accomodate
        the hack where we don't add right away when doing mouse tracking.

        * kwq/KWQSlider.h: Added destructor.
        * kwq/KWQSlider.mm:
        (-[KWQSlider initWithQSlider:]): Tweaked a little.
        (-[KWQSlider detachQSlider]): Added.
        (-[KWQSlider mouseDown:]): Added call to QWidget::beforeMouseDown/afterMouseDown.
        (-[KWQSlider widget]): Added.
        (QSlider::~QSlider): Added. Calls detachQSlider.

        * kwq/KWQTextArea.h: Added detachQTextEdit method.
        * kwq/KWQTextArea.mm:
        (-[KWQTextArea detachQTextEdit]): Added.
        (-[KWQTextArea textDidChange:]): Added check for widget of 0.
        (-[KWQTextArea becomeFirstResponder]): Ditto.
        (-[KWQTextArea nextKeyView]): Ditto.
        (-[KWQTextArea previousKeyView]): Ditto.
        (-[KWQTextArea drawRect:]): Ditto.
        (-[KWQTextAreaTextView insertTab:]): Ditto.
        (-[KWQTextAreaTextView insertBacktab:]): Ditto.
        (-[KWQTextAreaTextView shouldDrawInsertionPoint]): Ditto.
        (-[KWQTextAreaTextView selectedTextAttributes]): Ditto.
        (-[KWQTextAreaTextView mouseDown:]): Ditto.
        (-[KWQTextAreaTextView keyDown:]): Ditto.
        (-[KWQTextAreaTextView keyUp:]): Ditto.

        * kwq/KWQTextEdit.h: Added ~QTextEdit.
        * kwq/KWQTextEdit.mm: (QTextEdit::~QTextEdit): Added. Calls detachQTextEdit.

        * kwq/KWQTextField.h: Changed invalidate to detachQLineEdit.
        * kwq/KWQTextField.mm: (-[KWQTextFieldController detachQLineEdit]): Changed.

        * kwq/KWQWidget.h: Added addToSuperview/removeFromSuperview for use from QScrollView.
        Added beforeMouseDown and afterMouseDown for use in widget implementations.
        Removed unused hasMouseTracking function.
        * kwq/KWQWidget.mm:
        (QWidget::QWidget): Initialize two new fields.
        (QWidget::~QWidget): Added code to remove view when widget is destroyed.
        (QWidget::getOuterView): Remove unneeded exception blocking since we're just caling superview.
        (QWidget::addToSuperview): Added.
        (QWidget::removeFromSuperview): Added.
        (QWidget::beforeMouseDown): Added.
        (QWidget::afterMouseDown): Added.

        * khtml/rendering/render_layer.cpp:
        (RenderLayer::setHasHorizontalScrollbar): Remove parent parameter; let addChild call addSubview:.
        (RenderLayer::setHasVerticalScrollbar): Ditto.

2004-11-30  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3863031> REGRESSION (Mail): caret continues flashing while mouse is down

        * khtml/khtml_part.cpp:
        (KHTMLPart::timerEvent): Add a check for whether the mouse is down. Keep the caret drawn
        with no blink if it is.

2004-11-30  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3861602> cursor gets lost trying to backspace to delete a form control

        * khtml/khtml_part.cpp:
        (KHTMLPart::setFocusNodeIfNeeded): This function would clear the selection if a <button>
        or <input type=image> was checked for focus since these elements are keyboard-focusable,
        but not mouse focusable. Also, this function did not work hard enough to set the focused
        node, and was content to clear it if the first element checked failed the test, rather
        than looking more at parents. This would have the effect of clearing, then resetting the
        focus on a DIV containing a button or image with content on either side of it in the
        process of arrowing over such content.

2004-11-30  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplaceSelectionCommand::doApply): Fix smart replace, which I (knowingly) broke with yesterday's checkin.
        Also, call updateLayout() in one more place to prevent stale information being returned from caretMaxOffset().
        * khtml/khtml_part.cpp:
        (KHTMLPart::isCharacterSmartReplaceExempt): Make this virtual and always return true. This gets rid of an
        ugly APPLE_CHANGES block and use of KWQ(part) in ReplaceSelectionCommand.
        * khtml/khtml_part.h: To help out with the isCharacterSmartReplaceExempt cleanup, add declaration.
        * kwq/KWQKHTMLPart.h: To help out with the isCharacterSmartReplaceExempt cleanup, make 
        isCharacterSmartReplaceExempt virtual.

2004-11-30  Ken Kocienda  <kocienda@apple.com>

        Reviewed by me

        * khtml/editing/htmlediting.cpp:
        (khtml::ReplacementFragment::mergeEndNode): Fixed one-line coding mistake that created an endless loop.
        Seemed simple enough to land without review.

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Chris
    
        Rewrite of paste code (specifically the ReplaceSelectionCommand class). Many more cases
        are handled correctly now, including selections that span multiple blocks, and cases
        where content on the pasteboard ends in newlines (or what appear to be newlines to a
        user, really block ends or BRs). I also made one small, but important change in the 
        copy code to annotate the markup written to the pasteboard to support these selections 
        ending in newlines.

        New header that defines a couple of constants used in copying and pasting.

        * ForwardingHeaders/editing/html_interchange.h: Added.
        * khtml/editing/html_interchange.h: Added.
        
        Rewrite of the ReplaceSelectionCommand. There are several new helper functions, as well
        as a new helper class, ReplacementFragment, which encapsulates information and functions
        pertaining to a document fragment that is being inserted into a document.
        
        * khtml/editing/htmlediting.cpp:
        (khtml::ReplacementFragment::ReplacementFragment):
        (khtml::ReplacementFragment::~ReplacementFragment):
        (khtml::ReplacementFragment::firstChild): Simple accessor.
        (khtml::ReplacementFragment::lastChild): Ditto.
        (khtml::ReplacementFragment::mergeStartNode): Looks at the nodes in a fragment and determines
        the starting node to use for merging into the block containing the start of the selection.
        (khtml::ReplacementFragment::mergeEndNode): Same as above, but for the end of the selection.
        (khtml::ReplacementFragment::pruneEmptyNodes): Simple helper.
        (khtml::ReplacementFragment::isInterchangeNewlineComment): Determines if a node is the
        special annotation comment added in by the copy code.
        (khtml::ReplacementFragment::removeNode): Simple helper.
        (khtml::isComment): Simple helper.
        (khtml::isProbablyBlock): Determines if a node is of a type that is usually rendered as a block.
        I would like to do better than this some day, but this check will hold us until I can do better.
        (khtml::ReplaceSelectionCommand::ReplaceSelectionCommand):
        (khtml::ReplaceSelectionCommand::~ReplaceSelectionCommand):
        (khtml::ReplaceSelectionCommand::doApply):
        (khtml::ReplaceSelectionCommand::completeHTMLReplacement): Figures out the right ending selection.
        * khtml/editing/htmlediting.h: Declarations for the new ReplacementFragment class.
        (khtml::ReplacementFragment::root):
        (khtml::ReplacementFragment::type):
        (khtml::ReplacementFragment::isEmpty):
        (khtml::ReplacementFragment::isSingleTextNode):
        (khtml::ReplacementFragment::isTreeFragment):
        (khtml::ReplacementFragment::hasMoreThanOneBlock):
        (khtml::ReplacementFragment::hasLogicalNewlineAtEnd):
        
        This smaller set of changes markup generation to add the newline annotation described in the
        comment at the start of this entry.
        
        * khtml/xml/dom2_rangeimpl.cpp:
        (DOM::RangeImpl::addCommentToHTMLMarkup): Simple helper.
        (DOM::RangeImpl::toHTML): Added new EAnnotateForInterchange default argument to control whether
        comment annotations are added to the markup generated.
        * khtml/xml/dom2_rangeimpl.h: Add some new declarations.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge markupStringFromRange:nodes:]): Request that markup resulting from call to 
        DOM::RangeImpl::toHTML uses annotations when generating.
        
        New tests.        

        * layout-tests/editing/pasteboard/paste-text-001-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-001.html: Added.
        * layout-tests/editing/pasteboard/paste-text-002-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-002.html: Added.
        * layout-tests/editing/pasteboard/paste-text-003-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-003.html: Added.
        * layout-tests/editing/pasteboard/paste-text-004-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-004.html: Added.
        * layout-tests/editing/pasteboard/paste-text-005-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-005.html: Added.
        * layout-tests/editing/pasteboard/paste-text-006-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-006.html: Added.
        * layout-tests/editing/pasteboard/paste-text-007-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-007.html: Added.
        * layout-tests/editing/pasteboard/paste-text-008-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-008.html: Added.
        * layout-tests/editing/pasteboard/paste-text-009-expected.txt: Added.
        * layout-tests/editing/pasteboard/paste-text-009.html: Added.

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Made two small changes that make it possible for comments to have DOM nodes made for them
        when pasting. This relies on some earlier work I did some days ago.

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::startMarkup): Get the string from the comment.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge documentFragmentWithMarkupString:baseURLString:]): Did some very minor
        rearranging. Now passes a flag when creating a contextual fragment, requesting that comments
        be included in the DOM.

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Added some new helpers to the VisiblePosition class. I will begin to use these when I check in
        my improved paste code.

        * khtml/editing/visible_position.cpp:
        (khtml::blockRelationship)
        (khtml::visiblePositionsInDifferentBlocks)
        (khtml::isFirstVisiblePositionInBlock)
        (khtml::isFirstVisiblePositionInNode)
        (khtml::isLastVisiblePositionInBlock)
        * khtml/editing/visible_position.h

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        * khtml/xml/dom_position.cpp:
        (DOM::Position::downstream): Fix a bug in downstream that prevented a call with DoNotStayInBlock
        specified from obeying that directive. The old code would stop at an outer block boundary in
        the case where that block had a block as its first child. The correct behavior is to drill into
        that inner block (and continue on drilling down, if possible), to find the correct position.

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Small improvements to the node-display debugging helpers.

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::displayTree): Make the rootNode be this if there is no rootEditableElement.
        * khtml/xml/dom_nodeimpl.h: Make displayNode take a default argument of "" for its string.

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::handleGeneralDelete): The downstream position in this function 
        may need to be adjusted when deleting text off the front part of a text node. This fixes a problem 
        I discovered while improving the paste command, where the insertion poitn wound up in the wrong
        place after the delete. 

2004-11-29  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Add a new helper function to insert a paragraph separator. Will be used in my
        upcoming paste improvments.

        * khtml/editing/htmlediting.cpp: Added function
        (khtml::CompositeEditCommand::insertParagraphSeparator)
        * khtml/editing/htmlediting.h: Ditto.

2004-11-23  David Harrison  <harrison@apple.com>

        Added various comments.

        * khtml/editing/htmlediting.cpp:
        (khtml::StyleChange::init):
        (khtml::ApplyStyleCommand::doApply):
        (khtml::ApplyStyleCommand::applyBlockStyle):
        (khtml::ApplyStyleCommand::applyInlineStyle):

2004-11-23  David Hyatt  <hyatt@apple.com>

        Hit testing in table cells with top/bottom space from vertical alignment didn't work.  I forgot about the
        super-secret yPos() lie that table cells do.  Use m_y instead of yPos().

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::nodeAtPoint):

2004-11-22  David Hyatt  <hyatt@apple.com>

        Make sure you can use document.createElement to make a <canvas> element.

        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::createHTMLElement):

2004-11-22  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Dave.

        <rdar://problem/3492044> performing JavaScript operations on form elements is slower than WinIE (HTMLFormCollection)
        <rdar://problem/3489679> selecting an item on the Apache bugzilla query page is very slow (HTMLFormCollection)
        <rdar://problem/3477810> checking 80 check boxes with JavaScript is 10x slower than in IE (HTMLFormCollection)
        <rdar://problem/3760962> JavaScript that toggles checkboxes is slow (HTMLCollection,HTMLFormCollection)
        
        * khtml/ecma/kjs_html.cpp:
        (KJS::HTMLDocument::tryGet):
        * khtml/html/html_formimpl.cpp:
        (DOM::HTMLFormElementImpl::HTMLFormElementImpl):
        (DOM::HTMLFormElementImpl::~HTMLFormElementImpl):
        * khtml/html/html_formimpl.h:
        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::HTMLCollectionImpl):
        (HTMLCollectionImpl::~HTMLCollectionImpl):
        (HTMLCollectionImpl::CollectionInfo::CollectionInfo):
        (HTMLCollectionImpl::CollectionInfo::reset):
        (HTMLCollectionImpl::resetCollectionInfo):
        (HTMLCollectionImpl::checkForNameMatch):
        (appendToVector):
        (HTMLCollectionImpl::updateNameCache):
        (HTMLCollectionImpl::namedItems):
        (HTMLFormCollectionImpl::HTMLFormCollectionImpl):
        (HTMLFormCollectionImpl::~HTMLFormCollectionImpl):
        (HTMLFormCollectionImpl::item):
        (HTMLFormCollectionImpl::updateNameCache):
        * khtml/html/html_miscimpl.h:

2004-11-22  David Hyatt  <hyatt@apple.com>

        Improve the WebCore cache so that the maximum cacheable object size is scaled based off the total cache
        size.
        
        Reviewed by mjs

        * khtml/misc/loader.cpp:
        (CachedObject::finish):
        (Cache::flush):
        (Cache::setSize):
        * khtml/misc/loader.h:
        (khtml::Cache::maxCacheableObjectSize):

2004-11-22  David Hyatt  <hyatt@apple.com>

        Fix for 3673381, huge directory listing so slow it seems like a hang.  Rework painting and hit testing so that
        it crawls the line box tree instead of the render tree.  This allows more precise intersection/containment testing
        that lets us short circuit earlier when painting and hit testing.
        
        Reviewed by mjs

        * khtml/khtml_part.cpp:
        (KHTMLPart::isPointInsideSelection):
        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::paint):
        (khtml::RenderBlock::paintChildren):
        (khtml::RenderBlock::paintObject):
        (khtml::RenderBlock::paintFloats):
        (khtml::RenderBlock::nodeAtPoint):
        * khtml/rendering/render_block.h:
        * khtml/rendering/render_box.cpp:
        (RenderBox::nodeAtPoint):
        * khtml/rendering/render_box.h:
        * khtml/rendering/render_br.h:
        * khtml/rendering/render_canvas.cpp:
        (RenderCanvas::paint):
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::paintLines):
        (RenderFlow::hitTestLines):
        (RenderFlow::caretRect):
        (RenderFlow::addFocusRingRects):
        (RenderFlow::paintFocusRing):
        (RenderFlow::paintOutlines):
        (RenderFlow::paintOutlineForLine):
        * khtml/rendering/render_flow.h:
        * khtml/rendering/render_frames.cpp:
        (RenderFrameSet::nodeAtPoint):
        * khtml/rendering/render_frames.h:
        * khtml/rendering/render_image.cpp:
        (RenderImage::nodeAtPoint):
        * khtml/rendering/render_image.h:
        * khtml/rendering/render_inline.cpp:
        (RenderInline::paint):
        (RenderInline::nodeAtPoint):
        * khtml/rendering/render_inline.h:
        * khtml/rendering/render_layer.cpp:
        (RenderLayer::paintLayer):
        (RenderLayer::hitTest):
        (RenderLayer::hitTestLayer):
        * khtml/rendering/render_layer.h:
        * khtml/rendering/render_line.cpp:
        (khtml::InlineBox::paint):
        (khtml::InlineBox::nodeAtPoint):
        (khtml::InlineFlowBox::flowObject):
        (khtml::InlineFlowBox::nodeAtPoint):
        (khtml::InlineFlowBox::paint):
        (khtml::InlineFlowBox::paintBackgrounds):
        (khtml::InlineFlowBox::paintBackground):
        (khtml::InlineFlowBox::paintBackgroundAndBorder):
        (khtml::InlineFlowBox::paintDecorations):
        (khtml::EllipsisBox::paint):
        (khtml::EllipsisBox::nodeAtPoint):
        (khtml::RootInlineBox::paintEllipsisBox):
        (khtml::RootInlineBox::paint):
        (khtml::RootInlineBox::nodeAtPoint):
        * khtml/rendering/render_line.h:
        (khtml::InlineRunBox::paintBackgroundAndBorder):
        * khtml/rendering/render_object.cpp:
        (RenderObject::hitTest):
        (RenderObject::setInnerNode):
        (RenderObject::nodeAtPoint):
        * khtml/rendering/render_object.h:
        (khtml::RenderObject::PaintInfo::PaintInfo):
        (khtml::RenderObject::PaintInfo::~PaintInfo):
        (khtml::RenderObject::paintingRootForChildren):
        (khtml::RenderObject::shouldPaintWithinRoot):
        * khtml/rendering/render_table.cpp:
        (RenderTable::layout):
        (RenderTable::paint):
        * khtml/rendering/render_text.cpp:
        (simpleDifferenceBetweenColors):
        (correctedTextColor):
        (InlineTextBox::nodeAtPoint):
        (InlineTextBox::paint):
        (InlineTextBox::selectionStartEnd):
        (InlineTextBox::paintSelection):
        (InlineTextBox::paintMarkedTextBackground):
        (InlineTextBox::paintDecoration):
        (RenderText::posOfChar):
        * khtml/rendering/render_text.h:
        (khtml::RenderText::paint):
        (khtml::RenderText::layout):
        (khtml::RenderText::nodeAtPoint):
        * khtml/xml/dom2_eventsimpl.cpp:
        (MouseEventImpl::computeLayerPos):
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::prepareMouseEvent):
        * kwq/KWQAccObject.mm:
        (-[KWQAccObject accessibilityHitTest:]):
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::scrollOverflowWithScrollWheelEvent):
        (KWQKHTMLPart::eventMayStartDrag):
        (KWQKHTMLPart::khtmlMouseMoveEvent):
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge elementAtPoint:]):
        (-[WebCoreBridge _positionForPoint:]):

2004-11-22  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Dave.

        <rdar://problem/3890961> selecting an item on the Apache bugzilla query page can be sped up 10% (HTMLFormCollection)
        <rdar://problem/3890958> JavaScript that toggles checkboxes can be improved 73% (HTMLCollection,HTMLFormCollection)

        This avoids the O(N^2) penalty for named item traversal for form collections.

        It also combines the item traversal logic for all non-form
        collection operations into a single traverseNextItem
        function. This avoids having 5 copies of the big switch statement
        for this.

        Also fixed a bug that prevented the last form element from being removed properly.
        
        * khtml/html/html_formimpl.cpp:
        (DOM::removeFromVector):
        * khtml/dom/html_misc.cpp:
        (HTMLCollection::namedItems):
        * khtml/dom/html_misc.h:
        * khtml/ecma/kjs_html.cpp:
        (KJS::HTMLCollection::getNamedItems):
        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::traverseNextItem):
        (HTMLCollectionImpl::calcLength):
        (HTMLCollectionImpl::length):
        (HTMLCollectionImpl::item):
        (HTMLCollectionImpl::nextItem):
        (HTMLCollectionImpl::checkForNameMatch):
        (HTMLCollectionImpl::namedItem):
        (HTMLCollectionImpl::namedItems):
        (HTMLCollectionImpl::nextNamedItem):
        (HTMLFormCollectionImpl::calcLength):
        (HTMLFormCollectionImpl::namedItem):
        (HTMLFormCollectionImpl::nextNamedItem):
        (HTMLFormCollectionImpl::namedItems):
        * khtml/html/html_miscimpl.h:

2004-11-22  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Change around the way we block the Javascript "Paste" command identifier from
        being available. Formerly, this was done with an ifdef we never compiled in.
        Now, this is done with a couple of cheap runtime checks. The advantage is that
        we can now compile this command into development builds, and still yet switch
        on the command in deployment builds through the use of WebCore SPI so we can
        write and run layout tests with all of our builds.

        * khtml/editing/jsediting.cpp:
        (DOM::JSEditor::queryCommandSupported): Checks state of paste command in case
        command being queried is the paste command. 
        (DOM::JSEditor::setSupportsPasteCommand): New SPI to turn on paste command.
        * khtml/editing/jsediting.h: Ditto.
        * khtml/khtml_part.cpp:
        (KHTMLPart::pasteFromPasteboard): Added.
        (KHTMLPart::canPaste): Added.
        * kwq/KWQKHTMLPart.mm:
        (KHTMLPart::canPaste): Added.
        * kwq/KWQRenderTreeDebug.cpp:
        (externalRepresentation): Turn on paste command.
        * kwq/WebCoreBridge.h: Add canPaste call so WebKit can fill in the answer.

2004-11-21  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Richard.

        <rdar://problem/3889655> HTMLCollectionImpl should use traverseNextNode to improve speed and save recursion
        
        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::calcLength):
        (HTMLCollectionImpl::getItem):
        (HTMLCollectionImpl::item):
        (HTMLCollectionImpl::nextItem):
        (HTMLCollectionImpl::getNamedItem):
        (HTMLCollectionImpl::namedItem):
        (HTMLCollectionImpl::nextNamedItemInternal):
        (HTMLFormCollectionImpl::nextNamedItemInternal):

2004-11-19  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Darin.

        <rdar://problem/3482935> JavaScript so slow it seems like a hang (hrweb.apple.com) (HTMLCollection?)
        <rdar://problem/3759149> PeopleSoft page in Safari twice as slow as Mozilla engine (HTMLFormCollection)
        <rdar://problem/3888368> selecting an item on the Apache bugzilla query page can be improved 95% (HTMLFormCollection)

        Many optimizations to HTMLFormCollection. Iterating it should not
        be N^2 any more, though finding items by name could still be.
        
        * khtml/html/html_formimpl.cpp:
        (DOM::HTMLFormElementImpl::~HTMLFormElementImpl):
        (DOM::HTMLFormElementImpl::length):
        (DOM::HTMLFormElementImpl::submitClick):
        (DOM::HTMLFormElementImpl::formData):
        (DOM::HTMLFormElementImpl::submit):
        (DOM::HTMLFormElementImpl::reset):
        (DOM::HTMLFormElementImpl::radioClicked):
        (DOM::appendToVector):
        (DOM::removeFromVector):
        (DOM::HTMLFormElementImpl::registerFormElement):
        (DOM::HTMLFormElementImpl::removeFormElement):
        (DOM::HTMLFormElementImpl::makeFormElementDormant):
        (DOM::HTMLFormElementImpl::registerImgElement):
        (DOM::HTMLFormElementImpl::removeImgElement):
        * khtml/html/html_formimpl.h:
        * khtml/html/html_miscimpl.cpp:
        (HTMLFormCollectionImpl::FormCollectionInfo::FormCollectionInfo):
        (void::HTMLFormCollectionImpl::FormCollectionInfo::reset):
        (HTMLFormCollectionImpl::resetCollectionInfo):
        (HTMLFormCollectionImpl::calcLength):
        (HTMLFormCollectionImpl::item):
        (HTMLFormCollectionImpl::getNamedItem):
        (HTMLFormCollectionImpl::getNamedFormItem):
        (HTMLFormCollectionImpl::firstItem):
        (HTMLFormCollectionImpl::nextItem):
        * khtml/html/html_miscimpl.h:
        (DOM::HTMLFormCollectionImpl::~HTMLFormCollectionImpl):
        * khtml/xml/dom_elementimpl.cpp:
        (ElementImpl::setAttribute):
        (ElementImpl::setAttributeMap):
        * kwq/KWQPtrVector.h:
        (QPtrVector::findRef):
        * kwq/KWQVectorImpl.h:
        * kwq/KWQVectorImpl.mm:
        (KWQVectorImpl::findRef):
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge elementWithName:inForm:]):
        (-[WebCoreBridge controlsInForm:]):

2004-11-19  David Harrison  <harrison@apple.com>

        Reviewed by Ken and Darin.

        <rdar://problem/3856215> Cannot remove bold from the beginning of a message

        Problem is that KHTMLPart::computeAndSetTypingStyle always looked upstream
        for the existing style, but in this case (hitting cmd-B with caret at top of
        file) there is nothing upstream.  Changed this to use the VisiblePosition
        deepEquivalent instead.
        
        * khtml/khtml_part.cpp:
        (KHTMLPart::computeAndSetTypingStyle):

=== Safari-172 ===

2004-11-19  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Darin.

        <rdar://problem/3864151> REGRESSION (125-167): Chrysler.com never stops loading
        
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::close): Don't fire the onload handler if there is a
        redirect pending. This is a very long-standing bug that was masked
        by our previously incorrect redirect logic. It used to be that an
        older redirect would always win. Recently we changed things so
        that a newer redirect would win, but a script that causes a
        redirect would stop parsing once complete (so if there are two
        redirects in the same script, the latter wins). However, we should
        have also prevented onload in this case. Testing with other
        browsers shows that onload handlers do not run at all when there
        is a pending redirect.

2004-11-19  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Fix some object lifetime issues in these two commands. This fixes some crashes
        I am seeing in some new code I am working on, but have not yet reproduced otherwise.

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertParagraphSeparatorCommand::~InsertParagraphSeparatorCommand): No longer deref nodes
        in the ancestor list. They are not ref'ed when put on list. D'uh.
        (khtml::InsertParagraphSeparatorCommand::doApply): Ref all cloned nodes that are created by the command
        before putting them on the cloned nodes list. This are still deref'ed in the destructor.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::~InsertParagraphSeparatorInQuotedContentCommand): Ditto
        destructor comment.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::doApply): Ditto doApply comment.

2004-11-19  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Fix for this bug:
        
        <rdar://problem/3655241> setTypingStyle: does not set the real typing style, and typingStyle does not return it

        * khtml/khtml_part.cpp:
        (KHTMLPart::computeAndSetTypingStyle): New helper that does the work of reducing a passed-in style
        declaration given the current selection, and then sets the minimum necessary style as the typing
        style on the part.
        (KHTMLPart::applyStyle): Call new computeAndSetTypingStyle. The guts of computeAndSetTypingStyle used
        to be here in the selection-as-caret case. But now [WebCoreBridge setTypingStyle:] needs this code
        as well.
        * khtml/khtml_part.h: Declare new computeAndSetTypingStyle() function.
        * kwq/WebCoreBridge.h: Declare new typingStyle and setTypingStyle: methods.
        * kwq/WebCoreBridge.mm: 
        (-[WebCoreBridge typingStyle]): Calls through to the part to retrieve the typing style.
        (-[WebCoreBridge setTypingStyle:]): Calls through to the part to set the typing style.

2004-11-18  David Harrison  <harrison@apple.com>

        Reviewed by Darin.
        
        Back out part of Darin's fix for <rdar://problem/3885729>, because the new exception gets triggered
        by Mail.app. Filed <rdar://problem/3886832> against Mail.app.

        * kwq/DOM-CSS.mm:
        (-[DOMCSSStyleDeclaration setProperty:::]):

2004-11-18  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3587481> Bug Reporter Login Page: Password AutoFill does not work reliably

        Reviewed by john.

        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::currentForm): just return the current form, don't scan the entire document looking for a form if there is no current form

2004-11-18  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Chris.

        - fix recursive item traversal, use traverseNextNode() instead of
        the buggy hand-rolled traversal.

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeListImpl::recursiveItem):

2004-11-17  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        - fixed <rdar://problem/3885744> crash with XMLHttpRequest test page (reported by KDE folks)

        * khtml/ecma/xmlhttprequest.cpp: (KJS::XMLHttpRequest::slotFinished):
        Rolled in fix from KDE; make sure to set job to 0 before calling changeState.

        - fixed <rdar://problem/3885729> attempting to modify a computed style does nothing, but should raise a DOM exception
        - fixed <rdar://problem/3885731> style declarations use too many malloc blocks; switch to QValueList
        - fixed <rdar://problem/3885739> DOM::NodeImpl accessor in DOM::Node class is hot; should be inlined
        - changed NodeImpl calls like replaceChild to always ref/deref the parameter; this is a better way to fix
          an entire category of leaks we have been fixing one by one recently
        - changed computed styles so they hold a reference to the DOM node; the old code could end up with a
          stale RenderObject pointer, although I never saw it do that in practice
        - implemented the length and item methods for computed styles
        - implemented querying additional properties in computed styles (29 more)

        * khtml/khtml_part.h: Update forward declaration of CSSMutableStyleDeclarationImpl since it's
        now a separate class rather than a typedef. Changed the parameter type of setTypingStyle to
        take a mutable style.
        * khtml/khtml_part.cpp:
        (KHTMLPart::setTypingStyle): Change parameter to take a mutable style.
        (KHTMLPart::applyStyle): Add code to make a mutable style in case we are passed
        a computed style; also change some types to mutable style.
        (updateState): Update iteration of CSSProperty objects in a style declaration to use
        the new valuesIterator interface.
        (KHTMLPart::selectionHasStyle): Add a call to makeMutable.
        (KHTMLPart::selectionStartHasStyle): Add call to makeMutable and update iteration.
        (editingStyle): Change type to mutable style, and simplify the style-creation calls,
        including accomodating the exception code that setCssText has now.
        (KHTMLPart::applyEditingStyleToElement): Change types to mutable style.
        (KHTMLPart::removeEditingStyleFromElement): Change code to call setChanged only if removing
        the style attributes really was a change, although it's not an important optimization it's
        good to do it right.

        * khtml/css/css_base.h: Remove unneeded setParsedValue method.
        * khtml/css/css_base.cpp: Remove unneeded setParsedValue method. All the places that were
        calling it were already removing the old property explicitly, so the code in here to remove
        the property again was redundant.

        * khtml/css/css_computedstyle.h: Updated virtual functions for changes to parameters in base class.
        Moved all the "set"-type functions so they are private. Store a node pointer instead of a renderer.
        * khtml/css/css_computedstyle.cpp:
        (DOM::CSSComputedStyleDeclarationImpl::CSSComputedStyleDeclarationImpl): Hold a reference to
        the node we compute style for, so we don't end up with a pointer to a deallocated RenderObject.
        Before we had no guarantee the object would outlast us.
        (DOM::CSSComputedStyleDeclarationImpl::setCssText): Add exception parameter, and set the
        exception to NO_MODIFICATION_ALLOWED_ERR.
        (DOM::CSSComputedStyleDeclarationImpl::getPropertyCSSValue): Update to use node pointer rather
        than renderer pointer. Added implementation for box-align, box-direction, box-flex, box-flex-group,
        box-lines, box-ordinal-group, box-orient, box-pack, caption-side, clear, cursor, direction,
        list-style-image, list-style-position, list-style-type, marquee-direction, marquee-repetition,
        marquee-style, user-modify, opacity, orphans, outline-style, page-break-after, page-break-before,
        page-break-inside, position, unicode-bidi, widows, z-index.
        (DOM::CSSComputedStyleDeclarationImpl::removeProperty): Add exception parameter, and set the
        exception to NO_MODIFICATION_ALLOWED_ERR.
        (DOM::CSSComputedStyleDeclarationImpl::setProperty): Ditto.
        (DOM::CSSComputedStyleDeclarationImpl::length): Implemented.
        (DOM::CSSComputedStyleDeclarationImpl::item): Implemented, calls getPropertyValue.
        (DOM::CSSComputedStyleDeclarationImpl::copyInheritableProperties): Changed return type to
        CSSMutableStyleDeclarationImpl.
        (DOM::CSSComputedStyleDeclarationImpl::copy): Added.
        (DOM::CSSComputedStyleDeclarationImpl::makeMutable): Added.

        * khtml/css/css_ruleimpl.h: Update forward declaration of CSSMutableStyleDeclarationImpl since it's
        now a separate class rather than a typedef.
        * khtml/css/cssparser.h: Ditto.

        * khtml/css/css_valueimpl.h: Refactor CSSStyleDeclarationImpl into two classes. New derived class
        CSSMutableStyleDeclarationImpl has the guts, and the base class has only some virtual functions.
        Removed a bunch of redundant stuff from other classes in this file too.
        (DOM::DashboardRegionImpl::setNext): Ref new before deref'ing old to handle the set-to-same case.
        (DOM::CSSProperty::CSSProperty): Added new overload so you can create a CSSProperty with initial values.
        (DOM::CSSProperty::operator=): Added.
        (DOM::CSSProperty::setValue): Use ref-before-deref pattern to simplify slightly.

        * khtml/css/css_valueimpl.cpp:
        (DOM::CSSStyleDeclarationImpl::CSSStyleDeclarationImpl): Remove uneeded things.
        (DOM::CSSStyleDeclarationImpl::isStyleDeclaration): Put here now that it's no longer inline.
        (DOM::CSSMutableStyleDeclarationImpl::CSSMutableStyleDeclarationImpl): Added.
        (DOM::CSSMutableStyleDeclarationImpl::operator=): Added.
        (DOM::CSSMutableStyleDeclarationImpl::~CSSMutableStyleDeclarationImpl): Updated.
        (DOM::CSSMutableStyleDeclarationImpl::getPropertyValue): Removed now-uneeded check.
        (DOM::CSSMutableStyleDeclarationImpl::get4Values): Moved here from base class.
        (DOM::CSSMutableStyleDeclarationImpl::getShortHandValue): Ditto.
        (DOM::CSSMutableStyleDeclarationImpl::getPropertyCSSValue): Update to use QValueList instead of QPtrList.
        (DOM::CSSMutableStyleDeclarationImpl::removeProperty): Added exception parameter, updated for QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::setChanged): Moved here from base class.
        (DOM::CSSMutableStyleDeclarationImpl::getPropertyPriority): Update to use QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::setProperty): Added more overloads to match new parameters.
        (DOM::CSSMutableStyleDeclarationImpl::setStringProperty): Update to use QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::setImageProperty): Ditto.
        (DOM::CSSMutableStyleDeclarationImpl::parseProperty): Remove unneeded initialization code due to QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::addParsedProperties): Added.
        (DOM::CSSMutableStyleDeclarationImpl::setLengthProperty): Moved here from base class.
        (DOM::CSSMutableStyleDeclarationImpl::length): Update to use QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::item): Moved here from base class.
        (DOM::CSSMutableStyleDeclarationImpl::cssText): Return empty string rather than null string when there are
        no styles in the list. Update to use QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::setCssText): Update to use QValueList and to take an exceptionCode
        parameter and set it.
        (DOM::CSSMutableStyleDeclarationImpl::merge): Update to use QValueList.
        (DOM::CSSStyleDeclarationImpl::diff): Update to use QValueList.
        (DOM::CSSMutableStyleDeclarationImpl::copyBlockProperties): Moved here from base class. Change return type.
        (DOM::CSSStyleDeclarationImpl::copyPropertiesInSet): Update to use QValueList and use stack, not new/delete.
        (DOM::CSSMutableStyleDeclarationImpl::makeMutable): Added.
        (DOM::CSSMutableStyleDeclarationImpl::copy): Added.

        * khtml/css/cssparser.cpp:
        (CSSParser::parseValue): Changed to use addParsedProperties.
        (CSSParser::parseDeclaration): Ditto.
        (CSSParser::createStyleDeclaration): Use new constructor to create declaration in a more efficient manner.

        * khtml/css/cssproperties.in: Removed unused font-size-adjust and -khtml-flow-mode.
        * khtml/css/cssproperties.c: Regenerated.
        * khtml/css/cssproperties.h: Regenerated.

        * khtml/css/cssstyleselector.cpp: (khtml::CSSStyleSelector::applyDeclarations):
        Updated to use QValueList interface to CSSMutableStyleDeclarationImpl.

        * khtml/dom/css_value.cpp:
        (DOM::CSSStyleDeclaration::cssText): Removed unneeded cast.
        (DOM::CSSStyleDeclaration::setCssText): Added exception code handling.
        (DOM::CSSStyleDeclaration::getPropertyValue): Changed to call getPropertyValue directly instead of
        first doing getPropertyCSSValue and then doing cssText.
        (DOM::CSSStyleDeclaration::getPropertyCSSValue): Removed unneeded cast.
        (DOM::CSSStyleDeclaration::removeProperty): Added exception code handling. 
        (DOM::CSSStyleDeclaration::setProperty): Added exception code handling.
        (DOM::CSSStyleDeclaration::length): Removed unneeded cast.
        (DOM::CSSStyleDeclaration::item): Removed unneeded cast.
        (DOM::CSSStyleDeclaration::parentRule): Removed unneeded cast.
        (DOM::CSSValue::setCssText): Removed strange non-implementation (still not implemented).

        * khtml/dom/dom_node.h: Made isNull and handle functions inline.
        * khtml/dom/dom_node.cpp: Ditto.

        * khtml/editing/htmlediting.h: Change some types to mutable style.
        * khtml/editing/htmlediting.cpp:
        (khtml::EditCommandPtr::typingStyle): Change return type to mutable style.
        (khtml::EditCommandPtr::setTypingStyle): Change parameter to mutable style.
        (khtml::StyleChange::init): Convert parameter to mutable style. Update to use QValueList.
        (khtml::EditCommand::assignTypingStyle): Change parameter to mutable type.
        (khtml::EditCommand::setTypingStyle): Ditto.
        (khtml::ApplyStyleCommand::ApplyStyleCommand): Convert parameter to mutable style.
        (khtml::ApplyStyleCommand::doApply): Change local variables to mutable style.
        (khtml::ApplyStyleCommand::applyBlockStyle): Change parameter to mutable style.
        (khtml::ApplyStyleCommand::applyInlineStyle): Ditto.
        (khtml::ApplyStyleCommand::isHTMLStyleNode): Ditto.
        (khtml::ApplyStyleCommand::removeCSSStyle): Ditto. Also update to use QValueList.
        (khtml::ApplyStyleCommand::removeBlockStyle): Change parameter to mutable style.
        (khtml::ApplyStyleCommand::removeInlineStyle): Ditto.
        (khtml::ApplyStyleCommand::addBlockStyleIfNeeded): Ditto.
        (khtml::ApplyStyleCommand::addInlineStyleIfNeeded): Ditto.
        (khtml::InsertLineBreakCommand::doApply): Convert locals to mutable style.
        (khtml::InsertTextCommand::prepareForTextInsertion): Ditto.
        (khtml::RemoveCSSPropertyCommand::RemoveCSSPropertyCommand): Convert parameter to mutable style.

        * khtml/editing/jsediting.cpp: Convert types to mutable styles where we create styles.
        * khtml/html/html_baseimpl.h: Change type to mutable style.
        * khtml/html/html_baseimpl.cpp: (HTMLBodyElementImpl::createLinkDecl): Ditto.

        * khtml/html/html_elementimpl.h: Make CSSMappedAttributeDeclarationImpl use the mutable style
        class as a base class, and change types to mutable style as needed.
        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::createInlineStyleDecl): Change type to mutable style.
        (HTMLElementImpl::parseHTMLAttribute): Call parseProperty method.
        (HTMLElementImpl::getInlineStyleDecl): Change type to mutable style.
        (HTMLElementImpl::additionalAttributeStyleDecl): Ditto.
        (HTMLElementImpl::createContextualFragment): Add ref/deref to fix potential node leak.
        (HTMLElementImpl::setInnerHTML): Remove ref/deref pair because this leak is now fixed by changes
        to appendChild.
        (HTMLElementImpl::setOuterHTML): Remove ref/deref pair because this leak is now fixed by changes
        to replaceChild.

        * khtml/html/html_tableimpl.h: Change types to mutable style.
        * khtml/html/html_tableimpl.cpp:
        (HTMLTableElementImpl::additionalAttributeStyleDecl): Change type to mutable style.
        (HTMLTableElementImpl::getSharedCellDecl): Change type to mutable style.
        (HTMLTableCellElementImpl::additionalAttributeStyleDecl): Change type to mutable style.

        * khtml/html/htmlparser.cpp:
        (KHTMLParser::parseToken): Use a local variable to protect the node by ref'ing it. This is better
        than using an explicit delete to make the node go away, and is required for compatibility with the
        changes to the NodeImpl functions.
        (KHTMLParser::insertNode): Ditto.
        (KHTMLParser::createHead): Get rid of explicit delete, no longer needed because of changes to
        the NodeImpl functions.

        * khtml/xml/dom_docimpl.cpp: (DocumentImpl::createCSSStyleDeclaration): Call simpler constructor
        now that there's no need to make the property list explictly.

        * kwq/DOM-CSS.mm:
        (-[DOMCSSStyleDeclaration setCssText:]): Raise exception when appropriate.
        (-[DOMCSSStyleDeclaration removeProperty:]): Ditto.
        (-[DOMCSSStyleDeclaration setProperty:::]): Dito.

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::insertBefore): Always do a ref/deref, so callers don't have to worry about whether the
        function succeeded or not for ownership purposes.
        (NodeImpl::replaceChild): Ditto.
        (NodeImpl::appendChild): Ditto.
        (NodeBaseImpl::insertBefore): Ditto.
        (NodeBaseImpl::replaceChild): Ditto.
        (NodeBaseImpl::appendChild): Ditto.
        (NodeBaseImpl::addChild): Ditto.

        * WebCore-tests.exp: Removed CSSStyleDeclaration::length; not sure why it was in here.
        * WebCore-combined.exp: Regenerated.

2004-11-18  Maciej Stachowiak  <mjs@apple.com>

        still even more build fixing

        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::resetCollectionInfo):

2004-11-18  Maciej Stachowiak  <mjs@apple.com>

        more build fixing

        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::resetCollectionInfo):

2004-11-18  Maciej Stachowiak  <mjs@apple.com>

        Fixed build problem.

        * khtml/html/html_miscimpl.h:
        (DOM::HTMLCollectionImpl::CollectionInfo::CollectionInfo): it's haslength, not hasLenght.

2004-11-18  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Ken.

        - merged and cleaned up HTMLCollection and HTMLFormCollection speedups from konqueror

        <rdar://problem/3822992> VIP: Program listings pages at directv.com take a really long time to load [HTMLCollection]
        <rdar://problem/3701991> Safari unresponsive loading (www.maxim-ic.com) (HTMLCollection)

        This is also a start on fixing 5 other bugs, but those need additional work to make
        HTMLFormCollection fast.

        * khtml/html/html_documentimpl.h:
        (DOM::HTMLDocumentImpl::collectionInfo):
        * khtml/html/html_formimpl.cpp:
        (DOM::HTMLFormElementImpl::~HTMLFormElementImpl):
        (DOM::HTMLFormElementImpl::isURLAttribute):
        (DOM::HTMLFormElementImpl::registerImgElement):
        (DOM::HTMLFormElementImpl::removeImgElement):
        * khtml/html/html_formimpl.h:
        * khtml/html/html_imageimpl.cpp:
        (HTMLImageElementImpl::HTMLImageElementImpl):
        (HTMLImageElementImpl::~HTMLImageElementImpl):
        * khtml/html/html_imageimpl.h:
        * khtml/html/html_miscimpl.cpp:
        (HTMLCollectionImpl::HTMLCollectionImpl):
        (HTMLCollectionImpl::~HTMLCollectionImpl):
        (HTMLCollectionImpl::updateCollectionInfo):
        (HTMLCollectionImpl::length):
        (HTMLCollectionImpl::item):
        (HTMLCollectionImpl::firstItem):
        (HTMLCollectionImpl::nextItem):
        (HTMLCollectionImpl::namedItem):
        (HTMLCollectionImpl::nextNamedItemInternal):
        (HTMLFormCollectionImpl::getNamedFormItem):
        * khtml/html/html_miscimpl.h:
        (DOM::HTMLCollectionImpl::):
        (DOM::HTMLCollectionImpl::CollectionInfo::CollectionInfo):
        * khtml/html/htmlparser.cpp:
        (KHTMLParser::getElement):
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::DocumentImpl):
        * khtml/xml/dom_docimpl.h:
        (DOM::DocumentImpl::incDOMTreeVersion):
        (DOM::DocumentImpl::domTreeVersion):
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::attach):
        (NodeImpl::detach):

2004-11-18  Kevin Decker  <kdecker@apple.com>

        Reviewed by Chris.

        fixed: <rdar://problem/3841842> getPropertyID expensive

        * kwq/DOM-CSS.mm:
        (getPropertyID): avoid unnecessary memory allocations by using a fixed-sized stack based buffer.

2004-11-17  David Hyatt  <hyatt@apple.com>

        Improve responsiveness by being willing to break out of the tokenizer.  (This patch was landed already
        and subsequently backed out).
        
        Reviewed by kocienda

        * khtml/html/html_baseimpl.cpp:
        (HTMLBodyElementImpl::insertedIntoDocument):
        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::reset):
        (khtml::HTMLTokenizer::scriptHandler):
        (khtml::HTMLTokenizer::scriptExecution):
        (khtml::HTMLTokenizer::write):
        (khtml::HTMLTokenizer::continueProcessing):
        (khtml::HTMLTokenizer::timerEvent):
        (khtml::HTMLTokenizer::notifyFinished):
        * khtml/html/htmltokenizer.h:
        * khtml/khtmlview.cpp:
        (KHTMLViewPrivate::KHTMLViewPrivate):
        (KHTMLViewPrivate::reset):
        (KHTMLView::clear):
        (KHTMLView::layout):
        (KHTMLView::timerEvent):
        (KHTMLView::scheduleRelayout):
        (KHTMLView::layoutPending):
        (KHTMLView::haveDelayedLayoutScheduled):
        (KHTMLView::unscheduleRelayout):
        * khtml/khtmlview.h:
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::DocumentImpl):
        (DocumentImpl::close):
        (DocumentImpl::setParsing):
        (DocumentImpl::shouldScheduleLayout):
        (DocumentImpl::minimumLayoutDelay):
        (DocumentImpl::write):
        (DocumentImpl::finishParsing):
        (DocumentImpl::stylesheetLoaded):
        (DocumentImpl::updateStyleSelector):
        * khtml/xml/dom_docimpl.h:
        (DOM::DocumentImpl::parsing):
        * kwq/KWQDateTime.mm:
        (KWQUIEventTime::uiEventPending):

2004-11-17  David Harrison  <harrison@apple.com>

        Reviewed by Ken Kocienda.
        
        Make sure previousLineStart is non-null before calling compareBoundaryPoints.
        Treat null case as meaning no post-move merge is needed.

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::initializePositionData):

2004-11-17  David Harrison  <harrison@apple.com>
        
        Added displayNode and displayTree methods for debugging.  Fixed comment typo in dispatchChildRemovalEvents.
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::displayNode):
        (NodeImpl::displayTree):
        (NodeBaseImpl::dispatchChildRemovalEvents):
        * khtml/xml/dom_nodeimpl.h:

2004-11-16  John Sullivan  <sullivan@apple.com>

        Reviewed by Richard.
        
        - fixed <rdar://problem/3881929> 32 byte leak in editingStyle() in KHTMLPart (one-time only)

        * khtml/khtml_part.cpp:
        (editingStyle):
        delete the list we created when we're done with it

2004-11-16  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John
        
        It is unwise to use the QPtrList autodelete feature on shared objects like DOM nodes.
        Instead, I replaced this with a helper function that derefs DOM nodes stored in a 
        QPtrList when the list goes out of scope.

        * khtml/editing/htmlediting.cpp:
        (khtml::derefNodesInList): New helper to deref DOM nodes stored in a QPtrList.
        (khtml::InsertParagraphSeparatorCommand::InsertParagraphSeparatorCommand): No longer set lists to autodelete.
        (khtml::InsertParagraphSeparatorCommand::~InsertParagraphSeparatorCommand): Call new derefNodesInList helper.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::InsertParagraphSeparatorInQuotedContentCommand):  
        No longer set lists to autodelete.
        (khtml::InsertParagraphSeparatorInQuotedContentCommand::~InsertParagraphSeparatorInQuotedContentCommand):
         Call new derefNodesInList helper.
        * khtml/editing/htmlediting.h: Add virtual destructor for InsertParagraphSeparatorCommand. It had no need
        of one before, but now it does.

2004-11-15  David Harrison  <harrison@apple.com>

        Reviewed by Chris and Darin.
        
        <rdar://problem/3880304> Non-linear performance hit for style changes

        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::traverseNextNode):
        (NodeImpl::traverseNextSibling):
        (NodeImpl::traversePreviousNodePostOrder):
        Return 0 rather than traversing beyond stayWithin when this == stayWithin.
        Add asserts that stayWithin is an ancestor of the returned node.

2004-11-15  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        - fixed <rdar://problem/3880036> Many leaks from CSSComputedStyleDeclarationImpl::getPropertyCSSValue, seen in Mail and Blot

        * khtml/css/css_computedstyle.cpp:
        (DOM::CSSComputedStyleDeclarationImpl::getPropertyValue): Ref and deref the value returned from getPropertyCSSValue,
        since there's no guarantee it's already ref'd.
        * khtml/css/css_valueimpl.cpp:
        (CSSStyleDeclarationImpl::getPropertyValue): Wrap result in a CSSValue to ref/deref.
        (CSSStyleDeclarationImpl::get4Values): Ref/deref explicitly.
        (CSSStyleDeclarationImpl::getShortHandValue): Ditto.
        (CSSStyleDeclarationImpl::merge): Ditto.
        (CSSStyleDeclarationImpl::diff): Ditto.
        * khtml/editing/htmlediting.cpp:
        (khtml::StyleChange::currentlyHasStyle): Ditto.
        (khtml::ApplyStyleCommand::removeCSSStyle): Ditto.
        * khtml/html/html_baseimpl.cpp: (HTMLBodyElementImpl::parseHTMLAttribute): Ditto.
        * khtml/html/html_tableimpl.cpp: (HTMLTableElementImpl::parseHTMLAttribute): Ditto.

2004-11-15  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        Use separate mutable style and computed style types as appropriate.
        For now this should have no effect, but it prepares us for refactoring later.
        Also remove some unnecessary "DOM::" prefixes and in one case factor out
        some shared code.

        * khtml/khtml_part.cpp:
        (KHTMLPart::typingStyle):
        (KHTMLPart::setTypingStyle):
        (updateState):
        (KHTMLPart::selectionHasStyle):
        (KHTMLPart::selectionStartHasStyle):
        (KHTMLPart::selectionComputedStyle):
        * khtml/khtml_part.h:
        * khtml/khtmlpart_p.h:

        * khtml/css/css_base.h:
        * khtml/css/css_ruleimpl.cpp:
        (CSSStyleRuleImpl::setDeclaration):
        * khtml/css/css_ruleimpl.h:
        (DOM::CSSFontFaceRuleImpl::style):
        (DOM::CSSPageRuleImpl::style):
        (DOM::CSSStyleRuleImpl::style):
        (DOM::CSSStyleRuleImpl::declaration):
        * khtml/css/css_valueimpl.h:
        (DOM::CSSPrimitiveValueImpl::):
        * khtml/css/cssparser.cpp:
        (CSSParser::parseValue):
        (CSSParser::parseColor):
        (CSSParser::parseDeclaration):
        (CSSParser::createStyleDeclaration):
        * khtml/css/cssparser.h:
        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::addMatchedDeclaration):
        (khtml::CSSStyleSelector::matchRulesForList):
        (khtml::CSSStyleSelector::styleForElement):
        (khtml::CSSStyleSelector::applyDeclarations):
        * khtml/css/cssstyleselector.h:
        * khtml/css/parser.cpp:
        * khtml/css/parser.y:
        * khtml/dom/css_rule.h:
        * khtml/dom/css_stylesheet.h:
        * khtml/dom/css_value.h:
        * khtml/dom/dom2_views.cpp:
        * khtml/xml/dom2_viewsimpl.cpp:
        (DOM::AbstractViewImpl::getComputedStyle):
        * khtml/xml/dom_docimpl.cpp:
        (DocumentImpl::importNode):
        (DocumentImpl::setStyleSheet):
        * khtml/xml/dom_docimpl.h:
        * khtml/xml/dom_xmlimpl.cpp:
        (DOM::ProcessingInstructionImpl::setStyleSheet):
        * khtml/xml/dom_xmlimpl.h:

        * khtml/dom/css_value.cpp:
        (DOM::throwException): Added.
        (DOM::CSSStyleDeclaration::setCssText): Call throwException, but always on 0 for now.
        The real thing is coming with the next change to refactor.
        (DOM::CSSPrimitiveValue::setFloatValue): Call throwException.
        (DOM::CSSPrimitiveValue::setStringValue): Ditto.

2004-11-15  Darin Adler  <darin@apple.com>

        Reviewed by Ken.

        - fixed <rdar://problem/3878489> REGRESSION: modifying attribute of <textarea> blows away edited text (breaks simplemachines.org forum)

        * khtml/xml/dom_nodeimpl.h: Added boolean "children changed" parameter to
        dispatchSubtreeModifiedEvent, so it can be called in cases where only the
        node's attributes changed without sending a misleading childrenChanged call,
        but the childrenChanged call can happen at the exact right moment.
        * khtml/xml/dom_nodeimpl.cpp: Removed some uneeded "DOM::".
        (NodeImpl::dispatchSubtreeModifiedEvent): Only call "children changed" if
        the boolean true is passed in.

        * khtml/xml/dom_elementimpl.cpp:
        (NamedAttrMapImpl::addAttribute): Pass false for "children changed".
        (NamedAttrMapImpl::removeAttribute): Ditto.

2004-11-15  John Sullivan  <sullivan@apple.com>

        Reviewed by Ken.
        
        - fixed <rdar://problem/3880075> leak in CSSStyleDeclarationImpl::copyPropertiesInSet, 
        seen often in Mail and Blot

        * khtml/css/css_valueimpl.cpp:
        (CSSStyleDeclarationImpl::copyPropertiesInSet):
        delete temporary list after we're done using it

2004-11-15  Richard Williamson   <rjw@apple.com>

        Fixed leak (3879883) that John found.  Early return leaked
        allocated instance.

        Reviewed by John.

        * khtml/css/css_computedstyle.cpp:
        (DOM::CSSComputedStyleDeclarationImpl::getPropertyCSSValue):

2004-11-15  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for this bug:
        
        <rdar://problem/3879569> Many leaks in EditCommand mechanism, seen in Mail

        Fixed a couple of object lifetime issues. The EditCommand class used to hold an
        EditCommandPtr to its parent, but this caused a a reference cycle in composite 
        commands as the children held a ref to their parent. Now, the parent variable
        is a non-retained reference to an EditCommand *. It would be nice to have a 
        weak reference to the parent or even override deref in composite commands (but I
        can't since deref() is not virtual). However, this should be OK since any
        dangling parent pointer is a sign of a bigger object lifetime problem that
        would need to be addressed anyway.

        * khtml/css/css_valueimpl.cpp:
        (CSSStyleDeclarationImpl::CSSStyleDeclarationImpl): Fix bug in constructor that takes a 
        QPtrList<CSSProperty> *. List values must be copied into newly-allocated list, rather than
        just assigning the list variable passed in to the local list variable, or the list will be 
        double-deleted.
        * khtml/editing/htmlediting.cpp:
        (khtml::EditCommand::setStartingSelection): No longer call get(). m_parent is no longer a smart pointer.
        (khtml::EditCommand::setEndingSelection): Ditto.
        (khtml::EditCommand::assignTypingStyle): Short-circuit if passed in style is identical to current style.
        Unrelated to the change, but saves some ref's and deref's.
        (khtml::EditCommand::setTypingStyle): No longer call get(). m_parent is no longer a smart pointer.
        * khtml/editing/htmlediting.h: Change m_parent to a EditCommand *. Was an EditCommandPtr. Using an
        EditCommandPtr caused a reference cycle in composite commands as the children held a ref to their parent.
        (khtml::EditCommand::parent): No longer call get(). m_parent is no longer a smart pointer.

2004-11-15  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Kevin.

        <rdar://problem/3807080> Safari so slow it seems like a hang accessing a page on an IBM website
        
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeListImpl::NodeListImpl): Initialize isItemCacheValid, renamed isCacheValid to
        isLengthCacheValid.
        (NodeListImpl::recursiveLength): Adjusted for rename.
        (NodeListImpl::recursiveItem): Cache the last item accessed and its offset.
        If the same offset is looked up again, just return it, otherwise, if looking up
        a later offset, start at the last item and proceed from there.
        (NodeListImpl::itemById): Apply the special document optimization to all
        nodes that are either a document or in a document - just walk up to make
        sure the node found by ID has the root node as an ancestor.
        (NodeListImpl::rootNodeSubtreeModified): Adjust both cache bits.
        * khtml/xml/dom_nodeimpl.h: Prototype new stuff.

2004-11-15  John Sullivan  <sullivan@apple.com>

        Reviewed by Ken.
        
        - fixed <rdar://problem/3879539> leak of NSString after pasting into editable HTML (e.g. Mail)

        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::documentFragmentWithText):
        release mutable copy of string after we're done using it

2004-11-14  Kevin Decker  <kdecker@apple.com>

        Reviewed by mjs.

        fixed: <rdar://problem/3823038> LEAK: huge leak in DOM::HTMLElementImpl::createContextualFragment(DOM::DOMString const&, bool)

        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::setInnerHTML): uses the ref counting system to deallocate fragments instead of explicitly invoking a destructor.
        (HTMLElementImpl::setOuterHTML): function is responsible for derefing the fragment prior to returning. Now it does.

2004-11-13  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Kevin.

        <rdar://problem/3878766> VIP: Program listings pages at directv.com takes 75% of time traversing NodeLists

        * khtml/dom/dom_node.cpp:
        (NodeList::itemById): New method, just forward to impl.
        * khtml/dom/dom_node.h: Prototype it.
        * khtml/ecma/kjs_dom.cpp:
        (DOMNodeList::tryGet): Instead of looping over the whole list to do by-id access,
        let the NodeList do it. The NodeList might be able to do it more efficiently.
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeListImpl::itemById): Optimize for the case where the NodeList
        covers the whole document. In this case, just use getElementById,
        then check that the element satisfies the list criteria.
        (ChildNodeListImpl::nodeMatches): Return true only if the node is our child.
        (TagNodeListImpl::TagNodeListImpl): Irrelevant change to reformat initializers.
        * khtml/xml/dom_nodeimpl.h:

2004-11-12  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Gramps.

        - fixed another bug in the last checkin, isCacheValid was unitialized, resulting in
        sometimes using a huge bogus length value.
        
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeListImpl::NodeListImpl): Initialize isCacheValid.

2004-11-12  Darin Adler  <darin@apple.com>

        Reviewed by Maciej.

        - fixed an infinite loop in that last check-in

        * khtml/xml/dom_nodeimpl.cpp: (NodeImpl::notifyLocalNodeListsSubtreeModified):
        Added a ++i to the loop so it won't get stuck on the first element in the list.

2004-11-12  Maciej Stachowiak  <mjs@apple.com>

        Reviewed by Kevin.

        - fixed <rdar://problem/3878183> Safari is 77% slower than it should be on a page on an IBM website due to NodeListImpl length
        
        I fixed this by changing NodeLists to cache their length, but
        invalidate it whenever there is a change in the DOM subtree at
        which they are rooted. This makes NodeListImpl::recursiveLength()
        drop completely off the profile, since we were repeatedly getting
        a length for the same NodeList over and over.
        
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::NodeImpl):
        (NodeImpl::~NodeImpl):
        (NodeImpl::registerNodeList):
        (NodeImpl::unregisterNodeList):
        (NodeImpl::notifyLocalNodeListsSubtreeModified):
        (NodeImpl::notifyNodeListsSubtreeModified):
        (NodeImpl::dispatchSubtreeModifiedEvent):
        (NodeListImpl::NodeListImpl):
        (NodeListImpl::~NodeListImpl):
        (NodeListImpl::recursiveLength):
        (NodeListImpl::recursiveItem):
        (NodeListImpl::rootNodeSubtreeModified):
        (ChildNodeListImpl::ChildNodeListImpl):
        (ChildNodeListImpl::length):
        (ChildNodeListImpl::item):
        (TagNodeListImpl::TagNodeListImpl):
        (TagNodeListImpl::length):
        (TagNodeListImpl::item):
        (NameNodeListImpl::NameNodeListImpl):
        (NameNodeListImpl::length):
        (NameNodeListImpl::item):
        * khtml/xml/dom_nodeimpl.h:

2004-11-12  Darin Adler  <darin@apple.com>

        Reviewed by Maciej.

        - various small cleanups

        * khtml/xml/dom_docimpl.h: Added policyBaseURL and setPolicyBaseURL.
        * khtml/html/html_documentimpl.h: Removed policyBaseURL and setPolicyBaseURL.

        * khtml/xml/xml_tokenizer.h: Marked isWaitingForScripts const.
        * khtml/xml/xml_tokenizer.cpp: (khtml::XMLTokenizer::isWaitingForScripts): Marked const.
        * khtml/html/htmltokenizer.h: Marked isWaitingForScripts const.
        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::isWaitingForScripts): Marked const.
        (khtml::HTMLTokenizer::setOnHold): Took out extraneous line of code.

        * khtml/khtml_part.h: Removed docImpl function.
        * khtml/khtml_part.cpp: Ditto.

        * khtml/xml/dom_docimpl.cpp: (DocumentImpl::close): Simplified code that implements
        the "redirect during onload" optimization. Now uses isScheduledLocationChangePending.

        * kwq/KWQKHTMLPart.h: Removed now-unused _firstResponderAtMouseDownTime.
        * kwq/KWQKHTMLPart.mm: Removed _firstResponderAtMouseDownTime (forgot to land this
        part of the change last time, which is why the build broke).
        (KWQKHTMLPart::updatePolicyBaseURL): Use xmlDocImpl instead of docImpl.
        (KWQKHTMLPart::setPolicyBaseURL): Ditto.
        (KWQKHTMLPart::keyEvent): Ditto.
        (KWQKHTMLPart::dispatchCPPEvent): Ditto.
        (KWQKHTMLPart::bodyBackgroundColor): Ditto.

2004-11-12  Chris Blumenberg  <cblu@apple.com>

        <rdar://problem/3843312> REGRESSION: Tabbing into content area puts insertion point at start, should go to where it last was

        Reviewed by rjw.

        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::nextKeyViewInFrameHierarchy): only blow away selection when another view is focused

=== Safari-171 ===

2004-11-12  Darin Adler  <darin@apple.com>

        Reviewed by Kevin.

        - fixed a couple places that would not work for XML documents

        * khtml/ecma/kjs_window.cpp:
        (Window::isSafeScript): Use xmlDocImpl instead of docImpl, since the function we're using
        is present on the base class.
        (WindowFunc::tryCall): More of the same.

2004-11-12  Darin Adler  <darin@apple.com>

        - land versions of these files generated by the newer gperf

        People building on Panther will continue to see these files modified.
        A workaround would be to install the newer gperf on our Tiger build machines.

        * khtml/css/cssproperties.c: Regenerated.
        * khtml/css/cssvalues.c: Regenerated.
        * khtml/html/doctypes.cpp: Regenerated.
        * khtml/html/kentities.c: Regenerated.
        * khtml/misc/htmlattrs.c: Regenerated.
        * khtml/misc/htmltags.c: Regenerated.
        * kwq/KWQColorData.c: Regenerated.

2004-11-11  Richard Williamson   <rjw@apple.com>
        
        Fix build horkage from previous checkin.
                
        * kwq/KWQKHTMLPart.h:

2004-11-11  Darin Adler  <darin@apple.com>

        Reviewed by John.

        - fixed <rdar://problem/3846152> REGRESSION (125-166): can't drag text out of <input type=text> fields

        * kwq/WebCoreBridge.h: Added wasFirstResponderAtMouseDownTime: method.

        * kwq/KWQKHTMLPart.h: Removed _firstResponderAtMouseDownTime.
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::passWidgetMouseDownEventToWidget): Use the new wasFirstResponderAtMouseDownTime:
        method on the bridge instead of _firstResponderAtMouseDownTime. This will return YES for the case where
        we started with the NSTextField as first responder, and then took focus away and gave it back, which
        makes dragging text work again.
        (KWQKHTMLPart::mouseDown): Removed code to set _firstResponderAtMouseDownTime.

2004-11-11  David Hyatt  <hyatt@apple.com>

        Disable the tokenizer deferral, since it hurts the PLT by 3% or so.
        
        Reviewed by vicki

        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::continueProcessing):

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Maciej

        * khtml/editing/htmlediting.cpp:
        (khtml::InsertLineBreakCommand::doApply): Use new isLastVisiblePositionInBlock() helper instead
        of old isLastInBlock() member function on VisiblePosition. This is a cosmetic change in keeping
        with the prevailing style for the VisiblePosition class.
        * khtml/editing/htmlediting.h: Move isLastVisiblePositionInNode() function to visible_position.[cpp|h] files. 
        * khtml/editing/visible_position.cpp: Removed isLastInBlock() helper. Renamed to isLastVisiblePositionInBlock().
        (khtml::visiblePositionsInDifferentBlocks): New helper method.
        (khtml::isLastVisiblePositionInBlock): Ditto.
        (khtml::isLastVisiblePositionInNode): Ditto.
        * khtml/editing/visible_position.h: Add declarations for new functions.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Hyatt

        * khtml/editing/htmlediting.cpp:
        (khtml::CompositeEditCommand::deleteInsignificantText): Call new compareBoundaryPoints convenience.
        (khtml::ApplyStyleCommand::removeBlockStyle): Ditto.
        (khtml::ApplyStyleCommand::removeInlineStyle): Ditto.
        (khtml::ApplyStyleCommand::nodeFullySelected): Ditto.
        (khtml::DeleteSelectionCommand::initializePositionData): Ditto.
        * khtml/xml/dom2_rangeimpl.cpp:
        (DOM::RangeImpl::compareBoundaryPoints): New convenience variant of this function which takes two Position objects.
        * khtml/xml/dom2_rangeimpl.h: Ditto.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Some improvements to deleting when complete lines are selected.

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::initializePositionData): Detect when the line containing
        the end of a selection is fully selected. Turn off block merging in this case.
        (khtml::DeleteSelectionCommand::handleSpecialCaseBRDelete): Fix a bug in the check for
        whether a BR immediately followed a block. The old code could erroneously skip nodes.
        (khtml::DeleteSelectionCommand::handleGeneralDelete): Add a case for when the entire
        start block is selected. This new code will now delete this block in one call, rather
        than iterating over each child.
        * khtml/editing/visible_position.cpp:
        (khtml::visiblePositionsOnDifferentLines): New helper called in initializePositionData()
        to do the work mentioned above in the comment for that function.
        (khtml::isFirstVisiblePositionOnLine): Ditto.
        (khtml::isLastVisiblePositionOnLine): Ditto.
        * khtml/editing/visible_position.h: Add new functions.
        * layout-tests/editing/deleting/delete-line-001-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-001.html: Added.
        * layout-tests/editing/deleting/delete-line-002-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-002.html: Added.
        * layout-tests/editing/deleting/delete-line-003-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-003.html: Added.
        * layout-tests/editing/deleting/delete-line-004-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-004.html: Added.
        * layout-tests/editing/deleting/delete-line-005-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-005.html: Added.
        * layout-tests/editing/deleting/delete-line-006-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-006.html: Added.
        * layout-tests/editing/deleting/delete-line-007-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-007.html: Added.
        * layout-tests/editing/deleting/delete-line-008-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-008.html: Added.
        * layout-tests/editing/deleting/delete-line-009-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-009.html: Added.
        * layout-tests/editing/deleting/delete-line-010-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-010.html: Added.
        * layout-tests/editing/deleting/delete-line-011-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-011.html: Added.
        * layout-tests/editing/deleting/delete-line-012-expected.txt: Added.
        * layout-tests/editing/deleting/delete-line-012.html: Added.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Adele

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::initializePositionData): Add some comments and a new piece of debugging output.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Fix for these bugs:

        <rdar://problem/3875618> REGRESSION (Mail): Hitting down arrow with full line selected skips line (br case)
        <rdar://problem/3875641> REGRESSION (Mail): Hitting down arrow with full line selected skips line (div case)

        * khtml/editing/selection.cpp:
        (khtml::Selection::modifyMovingRightForward): Fixed by juggling the position as the starting point for
        the next line position when necessary.
        * layout-tests/editing/selection/move-3875618-fix-expected.txt: Added.
        * layout-tests/editing/selection/move-3875618-fix.html: Added.
        * layout-tests/editing/selection/move-3875641-fix-expected.txt: Added.
        * layout-tests/editing/selection/move-3875641-fix.html: Added.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        Improved some function names, at John's urging. No changes to the
        functions themselves.
        
        canPerformSpecialCaseAllContentDelete() --> handleSpecialCaseAllContentDelete()
        canPerformSpecialCaseBRDelete() --> handleSpecialCaseBRDelete()
        performGeneralDelete() --> handleGeneralDelete()

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::handleSpecialCaseAllContentDelete)
        (khtml::DeleteSelectionCommand::handleSpecialCaseBRDelete)
        (khtml::DeleteSelectionCommand::handleGeneralDelete)
        (khtml::DeleteSelectionCommand::doApply)
        * khtml/editing/htmlediting.h

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John
        
        Updated some layout test results that changed as a result of my last checking.
        Added a new test that has been in my tree for a few days.

        * layout-tests/editing/deleting/delete-3775172-fix-expected.txt:
        * layout-tests/editing/deleting/delete-3800834-fix-expected.txt:
        * layout-tests/editing/inserting/insert-3851164-fix-expected.txt: Added.
        * layout-tests/editing/inserting/insert-3851164-fix.html: Added.

2004-11-11  Ken Kocienda  <kocienda@apple.com>

        Reviewed by John

        * khtml/editing/htmlediting.cpp:
        (khtml::debugNode): New debugging helper.
        (khtml::DeleteSelectionCommand::initializePositionData): No longer call obsoleted
        startPositionForDelete() and endPositionForDelete() functions. Just use the 
        m_selectionToDelete object to determine start and end positions for the delete.
        (khtml::DeleteSelectionCommand::canPerformSpecialCaseAllContentDelete): New
        function that creates a special case for deleting all the content in a root
        editable element.
        (khtml::DeleteSelectionCommand::doApply): Call canPerformSpecialCaseAllContentDelete()
        function before BR special case and the general case delete functions.
        * khtml/editing/htmlediting.h: Updated for changed functions.

2004-11-10  Kevin Decker  <kdecker@apple.com>

        Reviewed by mjs.
        
        Fixed <rdar://problem/3875011> DOMNodeList::tryGet() performs unnecessary (and expensive) dom tree traversals. Improved a loop from 2-n-squared to just n-squared.  

        * khtml/ecma/kjs_dom.cpp:
        (DOMNodeList::tryGet): Got rid of an unnecessary node traversal. 

2004-11-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Chris

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::initializePositionData): Move position adjustments for
        smart delete from the two functions below to here. There was an unnecessary double 
        calculation of the leading and trailing whitespace positions. Also refined the trailing
        case so it only acts when the leading position is null (which seems to match TextEdit in
        my tests). Also removed some unnecessary copying of Position objects.
        (khtml::DeleteSelectionCommand::startPositionForDelete): Move out smart delete adjustment
        from here.
        (khtml::DeleteSelectionCommand::endPositionForDelete): Ditto.

2004-11-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        (khtml::DeleteSelectionCommand::performGeneralDelete): Add some more comments to
        make things more clear.
        * khtml/editing/selection.cpp:
        (khtml::Selection::toRange): Fixed the upstream and downstream calls so that the 
        resulting positions do not cross block boundaries. This was a bug and caused some
        delete problems when whole blocks were selected. I will be addressing that issue
        more fully in upcoming changes.

2004-11-10  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Harrison

        Some cleanups and fixes in upstream and downstream functions. 
        
        Removed redundant checks for isBlockFlow() when calling enclosingBlockFlowElement(). 
        Blocks do not need to skip the call to enclosingBlockFlowElement() for fear that the 
        block's enclosing block will be returned. 
        
        Remove code from upstream that confined the serach to block boundaries outside of
        the code which runs in the StayInBlock case. This code was redundant, and caused
        incorrect results to be returned in the DoNotStayInBlock case.
        
        In downstream, the check for crossing into a new block should use the equivalentDeepPosition()
        node, not the the this pointer's node.

        * khtml/xml/dom_position.cpp:
        (DOM::Position::upstream)
        (DOM::Position::downstream)

2004-11-09  David Hyatt  <hyatt@apple.com>

        Fix for 3873234, Safari UI is unresponsive when parsing multiple HTML docs and 3873233, Safari hangs when
        loading large local files.
        
        Reviewed by mjs

        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::HTMLTokenizer):
        (khtml::HTMLTokenizer::reset):
        (khtml::HTMLTokenizer::write):
        (khtml::HTMLTokenizer::stopped):
        (khtml::HTMLTokenizer::processingData):
        (khtml::HTMLTokenizer::continueProcessing):
        (khtml::HTMLTokenizer::timerEvent):
        (khtml::HTMLTokenizer::allDataProcessed):
        (khtml::HTMLTokenizer::end):
        (khtml::HTMLTokenizer::finish):
        (khtml::HTMLTokenizer::notifyFinished):
        * khtml/html/htmltokenizer.h:
        * khtml/khtml_part.cpp:
        (KHTMLPart::slotFinished):
        (KHTMLPart::end):
        (KHTMLPart::stop):
        * khtml/khtml_part.h:
        (KHTMLPart::tokenizerProcessedData):
        * khtml/khtmlview.cpp:
        * khtml/xml/dom_docimpl.cpp:
        * khtml/xml/xml_tokenizer.h:
        (khtml::Tokenizer::stopped):
        (khtml::Tokenizer::processingData):
        * kwq/KWQDateTime.h:
        * kwq/KWQDateTime.mm:
        (QDateTime::secsTo):
        (KWQUIEventTime::uiEventPending):
        * kwq/KWQKHTMLPart.h:
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::tokenizerProcessedData):
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge stop]):
        (-[WebCoreBridge numPendingOrLoadingRequests]):
        (-[WebCoreBridge doneProcessingData]):

2004-11-09  David Harrison  <harrison@apple.com>

        Reviewed by Ken Kocienda.

                <rdar://problem/3865837> Wrong text style after delete to start of document
                
        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::saveTypingStyleState):
        Sample computedStyle of m_selectionToDelete.start instead of m_downstreamStart.

2004-11-09  Richard Williamson   <rjw@apple.com>

        Fixed <rdar://problem/3872440> NSTimer prematurely released.

        Reviewed by Darin.

        * kwq/KWQTimer.mm:
        (QTimer::fire):
        * kwq/KWQWidget.mm:
        (QWidget::paint):

2004-11-09  Chris Blumenberg  <cblu@apple.com>

        Reviewed by vicki.

        * WebCore.pbproj/project.pbxproj: explicitly link against libxml2.2.6.14.dylib since the version number has been bumped

2004-11-08  David Harrison  <harrison@apple.com>

        Reviewed by Ken Kocienda.

                <rdar://problem/3865854> Deleting first line deletes all lines
                
        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::performGeneralDelete):
        Problem was that the code that deletes fully selected m_downstreamEnd.node() by deleting one
        of its ancestors, failed to end the loop that deletes all fully selected nodes.  Also,
        fixed this code to clear m_trailingWhitespaceValid.  Also removed dead m_endingPosition
        update because it is handled in calculateEndingPosition now.
        * layout-tests/editing/deleting/delete-3865854-fix-expected.txt: Added.
        * layout-tests/editing/deleting/delete-3865854-fix.html: Added.

2004-11-08  Ken Kocienda  <kocienda@apple.com>

        Reviewed by Hyatt

        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::createContextualFragment): Now takes flag to control whether comments
        are added to the DOM.
        * khtml/html/html_elementimpl.h: Ditto.
        * khtml/html/htmlparser.cpp:
        (KHTMLParser::KHTMLParser): Ditto.
        (KHTMLParser::getElement): Remove ifdef for comment processing. Replace with flag check.
        * khtml/html/htmlparser.h: Add flag to constructor so callers can request comment nodes.
        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::HTMLTokenizer): Add flag to constructor so callers can request comment nodes.
        (khtml::HTMLTokenizer::parseComment): Fix code to handle parsing out comment text correctly.
        There were a couple of indexing errors that resulted in the comment text containing part of the 
        comment markers.
        (khtml::HTMLTokenizer::processToken): Don't let token id be reset to ID_TEXT if token is a comment.
        * khtml/html/htmltokenizer.h:  Add flag to constructor so callers can request comment nodes.

2004-11-08  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3870907> WebCore unnecessary links against JavaVM and Security

        Reviewed by darin.

        * WebCore.pbproj/project.pbxproj: stop unnecessary linking
        * khtml/html/html_objectimpl.h: don't unnecessarily include JavaVM header
        * kwq/KWQKHTMLPart.h: ditto

2004-11-08  Darin Adler  <darin@apple.com>

        Reviewed by John.

        - fixed <rdar://problem/3825966> 8A274 Safari crashes closing window: QTimer::fire() with MallocStackLogging and MallocScribble enabled

        * kwq/KWQTimer.mm: (QTimer::fire): Rearrange so we don't access the QTimer object after calling code
        that possibly deletes the QTimer.

2004-11-08  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3783904> Return key behavior is confusingly different between popup menus and autofill menus

        Reviewed by john.

        * kwq/KWQTextField.mm:
        (-[KWQTextFieldController textView:shouldHandleEvent:]): let the bridge have a crack at the event so that it can swallow the newline if it wants to
        * kwq/WebCoreBridge.h:

2004-11-08  David Harrison  <harrison@apple.com>

        Reviewed by Darin.
        
        Renamed NodeImpl::enclosingNonBlockFlowElement to NodeImpl::enclosingInlineElement, per Hyatt.

        * khtml/editing/htmlediting.cpp:
        (khtml::DeleteSelectionCommand::moveNodesAfterNode):
        * khtml/xml/dom_nodeimpl.cpp:
        (NodeImpl::enclosingInlineElement):
        * khtml/xml/dom_nodeimpl.h:

2004-11-05  Chris Blumenberg  <cblu@apple.com>

        Fixed: <rdar://problem/3838413> REGRESSION (Mail): "Smart" word paste adds spaces before/after special characters

        Reviewed by