Check in updated layout tests.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7862 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by kocienda

More cleanup of block layout.  Eliminates the separate step for tables that dont fit on a line with a float
and consolidates it with clearing.

Also patch dom_textimpl.cpp to reduce further the # of RenderTexts created.

        * khtml/rendering/render_block.cpp:
        (khtml::getInlineRun):
        (khtml::RenderBlock::layoutBlock):
        (khtml::RenderBlock::adjustFloatingBlock):
        (khtml::RenderBlock::collapseMargins):
        (khtml::RenderBlock::clearFloatsIfNeeded):
        (khtml::RenderBlock::estimateVerticalPosition):
        (khtml::RenderBlock::layoutBlockChildren):
        (khtml::RenderBlock::markAllDescendantsWithFloatsForLayout):
        (khtml::RenderBlock::getClearDelta):
        (khtml::RenderBlock::calcBlockMinMaxWidth):
        * khtml/rendering/render_block.h:
        * khtml/rendering/render_frames.cpp:
        (RenderFrameSet::layout):
        * khtml/xml/dom_textimpl.cpp:
        (TextImpl::rendererIsNeeded):

        Fix for 3841060, regression with * in frames.  Reviewed by kocienda.

* layout-tests/fast/frames/002-expected.txt: Added.
        * layout-tests/fast/frames/002.html: Added.

 2004-10-19  Darin Adler  <darin@apple.com>

        Reviewed by Maciej.

        - follow-on to my fix yesterday, which broke a layout test because I rolled out a fix that Maciej had done

        * khtml/html/htmltokenizer.cpp: (khtml::HTMLTokenizer::write): Need to check the actual queue of external
        scripts being loaded here. If the current code being run is the external script itself, then we don't want
        to defer parsing. But loadingExtScript has to stay true until after the script runs. The old code would
        assume that any time we're running a script there's no need to look at loadingExtScript, but that was also
        wrong since there can be a script loading in that case too. Layout tests check for both problems.

        * layout-tests/fast/tokenizer/external-script-document-write-expected.txt: Added.
        * layout-tests/fast/tokenizer/external-script-document-write.html: Added.
        * layout-tests/fast/tokenizer/resources/external-script-document-write.js: Added.

        * layout-tests/fast/tokenizer/004.html: Fixed line endings (were CR, should be LF).

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7861 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Change suggested by Maciej during code review.

        * WebCoreSupport.subproj/WebTextRenderer.m: Changed rounding hack table to be const so it can be in shared instead
        of private memory, and doesn't require an initialization function.
        (+[WebTextRenderer initialize]): Removed initialization.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7860 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Maciej.

        - fixed <rdar://problem/3838934> Safari stops loading pages after rangeOfCharacterFromSet nil argument exception
        - fixed <rdar://problem/3843951> REGRESSION (166-167): Safari crashes in widthForNextCharacter (belkin.com, at startup for others)
        - fixed <rdar://problem/3841049> REGRESSION (109-110): control characters render as square boxes

        * WebCoreSupport.subproj/WebTextRenderer.m:
        (isSpace): Merged in isAlternateSpace, never used.
        (setupRoundingHackCharacterTable): Fixed size of table, was 1 entry too short. Got rid of unneeded call to bzero,
        since globals start out zeroed automatically.
        (isRoundingHackCharacter): Fixed backwards logic causing the crash in widthForNextCharacter.
        Also removed explicit compare with 1; check for non-zero is just fine.
        (fontContainsString): Change code so we'll just skip the font if the covered character set returns nil rather than
        throwing an exception like the old version did. This should make bug 3838934 go away, although perhaps covering up
        the underlying problem.
        (-[WebTextRenderer _convertCharacters:length:toGlyphs:]): Removed unused skipControlCharacters: parameter and also
        the unnecessary code to copy the buffer to change newline characters and non-break spaces to spaces.
        (-[WebTextRenderer _convertUnicodeCharacters:length:toGlyphs:]): Removed unused local.
        (-[WebTextRenderer _extendCharacterToGlyphMapToInclude:]): Added code to set up special cases for control characters,
        \n and non-break spaces.
        (-[WebTextRenderer _createATSUTextLayoutForRun:]): Added comment about the cases this code does not handle that
        are handled by the CG case.
        (widthForNextCharacter): Call isSpace instead of checking specifically for the space character here. The old code
        would not handle cases with '\n' coming across from WebCore properly.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7859 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add image for lists test.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7858 268f45cc-cd09-0410-ab3c-d52691b4dbfc

 2004-10-19  Darin Adler  <darin@apple.com>

        Reviewed by Maciej.

        - follow-on to my fix yesterday, which broke a layout test because I rolled out a fix that Maciej had done

        * khtml/html/htmltokenizer.cpp: (khtml::HTMLTokenizer::write): Need to check the actual queue of external
        scripts being loaded here. If the current code being run is the external script itself, then we don't want
        to defer parsing. But loadingExtScript has to stay true until after the script runs. The old code would
        assume that any time we're running a script there's no need to look at loadingExtScript, but that was also
        wrong since there can be a script loading in that case too. Layout tests check for both problems.

        * layout-tests/fast/tokenizer/external-script-document-write-expected.txt: Added.
        * layout-tests/fast/tokenizer/external-script-document-write.html: Added.
        * layout-tests/fast/tokenizer/resources/external-script-document-write.js: Added.

        * layout-tests/fast/tokenizer/004.html: Fixed line endings (were CR, should be LF).

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7857 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Dave Hyatt.

        - fixed <rdar://problem/3807234> REGRESSION (152-153): can't get element by ID that was just written with document.write in separate JS file (lacoccinelle.net)

        * khtml/html/htmltokenizer.cpp:
        (khtml::HTMLTokenizer::scriptHandler): Use !isEmpty instead of count != 0, since it's cheaper.
        (khtml::HTMLTokenizer::write): Roll back to the old version of the check here. The fix is that
        we only want to look at loadingExtScript if m_executingScript is 0.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7854 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        Fix for this bug:

        <rdar://problem/3840907> textedit doesn't render italic or bold text in html documents

        * khtml/css/css_valueimpl.cpp:
        (CSSPrimitiveValueImpl::getStringValue): This function did not return string values for idents.
        Also changed the return value to be DOMString, rather than DOMStringImpl, to deal with the
        lifecycle issues associated with creating a string to be returned in the ident case.
        * khtml/css/css_valueimpl.h: Change getStringValue to return DOMString rather than DOMStringImpl.
        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::applyProperty): Two calls of getStringValue needed updating.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7852 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3770135> hang loading page with EMBED tag pointing to same page (tridentantennas.co.uk)

        Reviewed by kocienda.

        * khtml/rendering/render_frames.cpp:
        (RenderPartObject::updateWidget): use completeURL before comparing the URL of the plug-in with the base URL of the document when avoiding frame recursion

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7851 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3840916> GC: -[WebNetscapePluginPackage initWithPath:] leaks an NSURL

        Reviewed by kevin.

        * Plugins.subproj/WebNetscapePluginPackage.m:
        (-[WebNetscapePluginPackage initWithPath:]): use executablePath on NSBundle instead of CFBundleCopyExecutableURL

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7850 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * DOM.subproj/DOMPrivate.h: change to copied header that was never committed

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7849 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3810183> Make WebHTMLView respect return value of webView:doCommandBySelector:

        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView doCommandBySelector:]):
        only do default action if delegate returns NO; this works with Mail as of Tiger 8A275.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7848 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3841774> would like to get NSColor from DOM-CSS

        Reviewed by john.

        * kwq/DOM-CSS.mm:
        (-[DOMRGBColor _color]): new, returns getNSColor on KWQColor
        * kwq/DOMPrivate.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7847 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        * khtml/rendering/bidi.cpp:
        (khtml::RenderBlock::skipNonBreakingSpace): Also need to forego the
        skipping after a clean line break, in addition to the cases already
        checked for.
        * layout-tests/editing/inserting/insert-br-007-expected.txt: Added.
        * layout-tests/editing/inserting/insert-br-007.html: Added.
        * layout-tests/editing/inserting/insert-br-008-expected.txt: Added.
        * layout-tests/editing/inserting/insert-br-008.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7846 268f45cc-cd09-0410-ab3c-d52691b4dbfc

versioning for TOT, Safari 2.0 (168u)

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7845 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Safari-167 stamp

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7843 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John

        Fix for this bug:

        <rdar://problem/3839989> REGRESSION (Mail): Left arrow does nothing after inserting attachment

        * khtml/editing/visible_position.cpp:
        (khtml::VisiblePosition::deepEquivalent): Remove code that attempted to bridge old-style
        position code to new-style VisiblePosition code. In retrospect, this code was misguided.
        Since we do a good job of insulating external code from the internal workings of
        VisiblePosition, the "hop ahead" being done here was not doing anyone any real good, and
        in the case of this bug, was doing harm. Simply removing this code makes the bug
        go away and does not cause any editing layout test regresssions.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7842 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        * khtml/rendering/bidi.cpp:
        (khtml::skipNonBreakingSpace): New helper.
        (khtml::RenderBlock::skipWhitespace): Do not skip non-breaking spaces that are
        at the start of a block. This was preventing users from typing spaces in empty
        documents.
        * layout-tests/editing/inserting/insert-space-in-empty-doc-expected.txt: Added.
        * layout-tests/editing/inserting/insert-space-in-empty-doc.html: Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7841 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin and Ken.

fix for <rdar://problem/3821070> null de-ref in DelectSelectionCommand::doApply()

        This change shifts some code around so that the code that determines what typing style
        is in effect is called before deleteUnrenderedText is called.  Two asserts are also added
        to ensure that start and end nodes of the selection are in the document.

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

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7840 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken

This change makes these three functions virtual so that the work is being done in KWQHTMLPart
        instead of khtml_part, eliminating the need for the "#if APPLE_CHANGES" statements in the khtml code.

        * khtml/khtml_part.cpp:
        (KHTMLPart::shouldBeginEditing):
        (KHTMLPart::shouldEndEditing):
        (KHTMLPart::isContentEditable):
        * khtml/khtml_part.h:
        * kwq/KWQKHTMLPart.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7839 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

        Reviewed by John

        Final fix for these bugs:

        <rdar://problem/3806306> HTML editing puts spaces at start of line
        <rdar://problem/3814252> HTML editing groups space with word causing wrapping

        This change sets some new CSS properties that have been added to WebCore to
        enable whitespace-handling and line-breaking features that make WebView work
        more like a text editor.

        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::applyProperty): Add and remove special editing CSS properties
        based on property value.
        * khtml/html/html_elementimpl.cpp:
        (HTMLElementImpl::setContentEditable): Add and remove special editing CSS properties
        based on attribute value.
        * khtml/khtml_part.cpp:
        (KHTMLPart::applyEditingStyleToBodyElement): New helper. Calls applyEditingStyleToElement on
        body element.
        (KHTMLPart::removeEditingStyleFromBodyElement): New helper. Calls removeEditingStyleFromElement on
        body element.
        (KHTMLPart::applyEditingStyleToElement): Adds special editing CSS properties to passed in element.
        (KHTMLPart::removeEditingStyleFromElement): Removes special editing CSS properties from passed in element.
        * khtml/khtml_part.h: Add new declarations.
        * kwq/WebCoreBridge.h: Ditto.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge applyEditingStyleToBodyElement]): Call through to similarly-named function on KHTMLPart.
        (-[WebCoreBridge removeEditingStyleFromBodyElement]): Ditto.
        (-[WebCoreBridge applyEditingStyleToElement:]): Ditto.
        (-[WebCoreBridge removeEditingStyleFromElement:]): Ditto.

WebKit:

        Reviewed by John

        Final fix for these bugs:

        <rdar://problem/3806306> HTML editing puts spaces at start of line
        <rdar://problem/3814252> HTML editing groups space with word causing wrapping

        This change sets some new CSS properties that have been added to WebCore to
        enable whitespace-handling and line-breaking features that make WebView work
        more like a text editor.

        * WebView.subproj/WebHTMLRepresentation.m:
        (-[WebHTMLRepresentation finishedLoadingWithDataSource:]): Turn on special editing
        CSS properties when loading an HTML document into a WebView that is editable.
        * WebView.subproj/WebView.m:
        (-[WebView setEditable:]): Add and remove special editing CSS properties in current
        document being displayed.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7838 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3840052> Crash in removeBlockPlaceholderIfNeeded attaching file to empty document

        * khtml/editing/htmlediting.cpp:
        (khtml::CompositeEditCommand::removeBlockPlaceholderIfNeeded):
        needed a nil check to handle empty document case

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7837 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fixed <rdar://problem/3823026> making isRoundingHackCharacter use -O3 and an 8-bit lookup-table will speed "XBS" test up by 3% (actually < 1%)

        Careful testing shows a small performance gain on very large text files.
        I saw large variations in timings, but taking the lowest PLT timing
        with and without this change showed a 0.9% gain.  Note the cvs-base showed
        no improvement.  The improvement was for the large page attached to the
        bug.

        Reviewed by Ken.

        * WebCoreSupport.subproj/WebTextRenderer.m:
        (setupRoundingHackCharacterTable):
        (isRoundingHackCharacter):
        (+[WebTextRenderer initialize]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7836 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by me

        Fix build breakage. These three functions need to return the values from their
        calls to WebCGColorSpaceCreateXXX.

        * WebCoreSupport.subproj/WebGraphicsBridge.m:
        (-[WebGraphicsBridge createRGBColorSpace])
        (-[WebGraphicsBridge createGrayColorSpace])
        (-[WebGraphicsBridge createCMYKColorSpace])

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7835 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Addressed concerns in <rdar://problem/3803117> RESP: High complexity in icu uidna_IDNToASCII called by [NSString(WebNSURLExtras) _web_mapHostNameWithRange:encode:makeString:]

In practice I saw NO improvement in performance.  Although,
special-case tests could possibly show improvement.  Anyway,
the changes don't hurt performance.

        Reviewed by Maciej.

        * Misc.subproj/WebNSURLExtras.m:
        (-[NSString _web_mapHostNameWithRange:encode:makeString:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7834 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

        Reviewed by Ken.

<rdar://problem/3824626> Change to do colormatching for DeviceRGB colorspace causes ~11% Safari slowdown

- I fixed this by turning off all colormatching for WebKit
content. We might turn it back on later. For now, it's possible to
turn it on temporarily by defining COLORMATCH_EVERYTHING.

* WebCorePrefix.h:
        * khtml/ecma/kjs_html.cpp:
        (KJS::Context2DFunction::tryCall):
        (Context2D::colorRefFromValue):
        (Gradient::getShading):
        * khtml/rendering/render_canvasimage.cpp:
        (RenderCanvasImage::createDrawingContext):
        * kwq/KWQColor.mm:
        (QColor::getNSColor):
        * kwq/KWQPainter.h:
        * kwq/KWQPainter.mm:
        (CGColorFromNSColor):
        (QPainter::selectedTextBackgroundColor):
        (QPainter::rgbColorSpace):
        (QPainter::grayColorSpace):
        (QPainter::cmykColorSpace):
        * kwq/WebCoreGraphicsBridge.h:
        * kwq/WebCoreGraphicsBridge.m:
        (-[WebCoreGraphicsBridge createRGBColorSpace]):
        (-[WebCoreGraphicsBridge createGrayColorSpace]):
        (-[WebCoreGraphicsBridge createCMYKColorSpace]):

WebKit:

        Reviewed by Ken.

<rdar://problem/3824626> Change to do colormatching for DeviceRGB colorspace causes ~11% Safari slowdown

- I fixed this by turning off all colormatching for WebKit
content. We might turn it back on later. For now, it's possible to
turn it on temporarily by defining COLORMATCH_EVERYTHING.

* WebCoreSupport.subproj/WebGraphicsBridge.m:
        (-[WebGraphicsBridge setFocusRingStyle:radius:color:]):
        (-[WebGraphicsBridge additionalPatternPhase]):
        (-[WebGraphicsBridge createRGBColorSpace]):
        (-[WebGraphicsBridge createGrayColorSpace]):
        (-[WebGraphicsBridge createCMYKColorSpace]):
        * WebCoreSupport.subproj/WebImageData.m:
        * WebCoreSupport.subproj/WebImageRenderer.h:
        * WebCoreSupport.subproj/WebImageRenderer.m:
        (-[WebImageRenderer _adjustSizeToPixelDimensions]):
        (-[WebImageRenderer incrementalLoadWithBytes:length:complete:]):
        (-[WebImageRenderer _adjustColorSpace]):
        (-[WebImageRenderer drawClippedToValidInRect:fromRect:]):
        (-[WebImageRenderer tileInRect:fromPoint:context:]):
        (_createImageRef):
        (WebCGColorSpaceCreateRGB):
        (WebCGColorSpaceCreateGray):
        (WebCGColorSpaceCreateCMYK):
        * WebKitPrefix.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7833 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        * khtml/css/css_valueimpl.cpp:
        (CSSStyleDeclarationImpl::merge): A little cleanup. Also, make sure m_lstValues
        is non-null before appending.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7832 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by me

        Update expected results for improved behavior as a result of fix to 3816768.

        * layout-tests/editing/deleting/delete-3775172-fix-expected.txt
        * layout-tests/editing/deleting/delete-3800834-fix-expected.txt
        * layout-tests/editing/deleting/delete-block-merge-contents-002-expected.txt

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7831 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Richard

        * khtml/css/css_computedstyle.cpp:
        (DOM::CSSComputedStyleDeclarationImpl::getPropertyCSSValue): Add support to computed style
        for getting -khtml-line-break and -khml-nbsp-mode.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7830 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John

        Fix for this bug:

        <rdar://problem/3816768> REGRESSION (Mail): Deleting last character in block incorrectly
        moves caret out of block.

        The issue here is that an empty block with no explicit height set by style collapses
        to zero height, and does so immediately after the last bit of content is removed from
        it (as a result of deleting text with the delete key for instance). Since zero-height
        blocks are not eligible caret positions, the caret jumped to the closest eligible spot.

        The fix is to detect when a block has not been removed itself, but has had all its
        contents removed. In this case, a BR element is placed in the block, one that is
        specially marked as a placeholder. Later, if the block ever receives content, this
        placeholder is removed.

        * khtml/editing/htmlediting.cpp:
        (khtml::blockPlaceholerClassString): String which acts as a placeholder marker class.
        (khtml::CompositeEditCommand::insertBlockPlaceholderIfNeeded): Adds a placeholder BR if needed.
        (khtml::CompositeEditCommand::removeBlockPlaceholderIfNeeded): Removes a placeholder BR if needed.
        (khtml::DeleteSelectionCommand::moveNodesAfterNode): Call removeBlockPlaceholderIfNeeded.
        Also, do some cleanup on some old, crufty code in the move logic that is just so clearly wrong
        (it's very clear that we needs to be able to move more than just text nodes). This may expose
        bugs, but these bugs needs to be filed and fixed, not ducked. Besides, undoing this silliness
        made the test case in the bug work.
        (khtml::DeleteSelectionCommand::doApply): Call insertBlockPlaceholderIfNeeded and
        removeBlockPlaceholderIfNeeded.
        (khtml::InputTextCommand::input): Call removeBlockPlaceholderIfNeeded.
        (khtml::ReplaceSelectionCommand::doApply): Call removeBlockPlaceholderIfNeeded.
        * khtml/editing/htmlediting.h: Declare new functions.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7829 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Don't fill background with transparency unless debug flag
is enabled.

        Reviewed by Hyatt.

        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView drawRect:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7828 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Moved boolean checks prior to NSNumber checks.  booleans are
NSNumbers.

Follow on to <rdar://problem/3821515> binding layer needs to convert NSNumber-bools to js type boolean not number.

        Reviewed by John.

        * bindings/objc/objc_utility.mm:
        (KJS::Bindings::convertObjcValueToValue):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7827 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Added support for -apple-dashboard-region:none.  And fixed
a few computed style problems.

        Fixed <rdar://problem/3833532> -apple-dashboard-region: none; is needed
        Reviewed by Hyatt.

        * khtml/css/css_computedstyle.cpp:
        (DOM::CSSComputedStyleDeclarationImpl::getPropertyCSSValue):
        * khtml/css/css_valueimpl.cpp:
        (CSSPrimitiveValueImpl::cssText):
        * khtml/css/cssparser.cpp:
        (CSSParser::parseValue):
        (CSSParser::parseDashboardRegions):
        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::applyProperty):
        * khtml/rendering/render_style.cpp:
        (RenderStyle::noneDashboardRegions):
        * khtml/rendering/render_style.h:
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::dashboardRegionsDictionary):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7826 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Make the review fixes from kocienda's review.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7825 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Rework block layout to clean it up and simplify it (r=kocienda).

Also fixing the style sharing bug (r=mjs).

        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::MarginInfo::MarginInfo):
        (khtml::RenderBlock::layoutBlock):
        (khtml::RenderBlock::adjustPositionedBlock):
        (khtml::RenderBlock::adjustFloatingBlock):
        (khtml::RenderBlock::handleSpecialChild):
        (khtml::RenderBlock::handleFloatingOrPositionedChild):
        (khtml::RenderBlock::handleCompactChild):
        (khtml::RenderBlock::insertCompactIfNeeded):
        (khtml::RenderBlock::handleRunInChild):
        (khtml::RenderBlock::collapseMargins):
        (khtml::RenderBlock::clearFloatsIfNeeded):
        (khtml::RenderBlock::estimateVerticalPosition):
        (khtml::RenderBlock::determineHorizontalPosition):
        (khtml::RenderBlock::setCollapsedBottomMargin):
        (khtml::RenderBlock::adjustChildIfOverhangingFloatsExist):
        (khtml::RenderBlock::handleBottomOfBlock):
        (khtml::RenderBlock::layoutBlockChildren):
        (khtml::RenderBlock::getAbsoluteRepaintRectIncludingFloats):
        (khtml::RenderBlock::addOverHangingFloats):
        * khtml/rendering/render_block.h:
        (khtml::RenderBlock::maxTopMargin):
        (khtml::RenderBlock::maxBottomMargin):
        (khtml::RenderBlock::CompactInfo::compact):
        (khtml::RenderBlock::CompactInfo::block):
        (khtml::RenderBlock::CompactInfo::matches):
        (khtml::RenderBlock::CompactInfo::clear):
        (khtml::RenderBlock::CompactInfo::set):
        (khtml::RenderBlock::CompactInfo::CompactInfo):
        (khtml::RenderBlock::MarginInfo::setAtTopOfBlock):
        (khtml::RenderBlock::MarginInfo::setAtBottomOfBlock):
        (khtml::RenderBlock::MarginInfo::clearMargin):
        (khtml::RenderBlock::MarginInfo::setSelfCollapsingBlockClearedFloat):
        (khtml::RenderBlock::MarginInfo::setTopQuirk):
        (khtml::RenderBlock::MarginInfo::setBottomQuirk):
        (khtml::RenderBlock::MarginInfo::setDeterminedTopQuirk):
        (khtml::RenderBlock::MarginInfo::setPosMargin):
        (khtml::RenderBlock::MarginInfo::setNegMargin):
        (khtml::RenderBlock::MarginInfo::setPosMarginIfLarger):
        (khtml::RenderBlock::MarginInfo::setNegMarginIfLarger):
        (khtml::RenderBlock::MarginInfo::setMargin):
        (khtml::RenderBlock::MarginInfo::atTopOfBlock):
        (khtml::RenderBlock::MarginInfo::canCollapseWithTop):
        (khtml::RenderBlock::MarginInfo::canCollapseWithBottom):
        (khtml::RenderBlock::MarginInfo::canCollapseTopWithChildren):
        (khtml::RenderBlock::MarginInfo::canCollapseBottomWithChildren):
        (khtml::RenderBlock::MarginInfo::selfCollapsingBlockClearedFloat):
        (khtml::RenderBlock::MarginInfo::quirkContainer):
        (khtml::RenderBlock::MarginInfo::determinedTopQuirk):
        (khtml::RenderBlock::MarginInfo::topQuirk):
        (khtml::RenderBlock::MarginInfo::bottomQuirk):
        (khtml::RenderBlock::MarginInfo::posMargin):
        (khtml::RenderBlock::MarginInfo::negMargin):
        (khtml::RenderBlock::MarginInfo::margin):
        * khtml/rendering/render_box.cpp:
        (RenderBox::calcAbsoluteVertical):
        * khtml/rendering/render_box.h:
        (khtml::RenderBox::marginTop):
        (khtml::RenderBox::marginBottom):
        (khtml::RenderBox::marginLeft):
        (khtml::RenderBox::marginRight):
        * khtml/rendering/render_image.cpp:
        (RenderImage::setImage):
        * khtml/rendering/render_object.cpp:
        (RenderObject::sizesToMaxWidth):
        * khtml/rendering/render_object.h:
        (khtml::RenderObject::collapsedMarginTop):
        (khtml::RenderObject::collapsedMarginBottom):
        (khtml::RenderObject::maxTopMargin):
        (khtml::RenderObject::maxBottomMargin):
        (khtml::RenderObject::marginTop):
        (khtml::RenderObject::marginBottom):
        (khtml::RenderObject::marginLeft):
        (khtml::RenderObject::marginRight):
        * khtml/rendering/render_text.h:
        (khtml::RenderText::marginLeft):
        (khtml::RenderText::marginRight):
        * khtml/xml/dom_elementimpl.cpp:
        (ElementImpl::recalcStyle):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7824 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John

        Fix for this bug:

        <rdar://problem/3836158> REGRESSION (Mail): command-right-arrow moves to beginning of next line

        * khtml/editing/selection.cpp:
        (khtml::endOfLastRunAt): Do not let the end of the last run on a line be a BR.
        This will make it seem like the run ends on the next line.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7823 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        Fix for this bug:

        <rdar://problem/3836986> Delete code removes elements of table structure; can result in very broken-looking web pages

        * khtml/editing/htmlediting.cpp:
        (khtml::isTableStructureNode): New helper. Determines whether a node is a table cell,
        row, section, or column.
        (khtml::CompositeEditCommand::removeFullySelectedNode): New helper that recurses into elements
        of table structure when doing deletes, rather than deleting the structure elements themselves.
        (khtml::DeleteSelectionCommand::moveNodesAfterNode): Do not move content between elements
        of table structure. We may want to revisit this some day, but this seems like the best behavior
        to me now.
        (khtml::DeleteSelectionCommand::doApply): Call removeFullySelectedNode instead of removeNode
        where needed.
        * khtml/editing/htmlediting.h: Add declarations for new functions.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7822 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebKit:
        Fixed <rdar://problem/3829705> Need to remove filling w/ transparency when not drawing backgroundy.

        Reviewed by Ken.

        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView _transparentBackground]):
        (-[WebHTMLView _setTransparentBackground:]):
        (-[WebHTMLView drawRect:]):
        * WebView.subproj/WebHTMLViewInternal.h:
        * WebView.subproj/WebHTMLViewPrivate.h:

WebBrowser:
As part of fix <rdar://problem/3829705> Need to remove filling w/ transparency when not drawing backgroundy, we can no longer rely on WebHTMLView filling with transparency, so added additional flag to indicate that we should.

        * Debug/DebugUtilities.m:
        (-[BrowserDocument toggleTransparentWindow:]):
        * LocationChangeHandler.m:
Removed #define to prevent console spew.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7821 268f45cc-cd09-0410-ab3c-d52691b4dbfc

JavaScriptCore
        Fixed access to DOM object via WebScriptObject API.
        The execution context for DOM objects wasn't being found.
        <rdar://problem/3831372> The valueForKey method for @"offsetLeft" on a paragraph element causes a crash.

        Reviewed by Chris.

        * bindings/objc/WebScriptObject.mm:
        (_didExecute):
        (-[WebScriptObject KJS::Bindings::]):
        (-[WebScriptObject callWebScriptMethod:withArguments:]):
        (-[WebScriptObject evaluateWebScript:]):
        (-[WebScriptObject setValue:forKey:]):
        (-[WebScriptObject valueForKey:]):
        (-[WebScriptObject stringRepresentation]):
        * bindings/objc/WebScriptObjectPrivate.h:

WebCore:
        Fixed access to DOM object via WebScriptObject API.
        The execution context for DOM objects wasn't being found.
        <rdar://problem/3831372> The valueForKey method for @"offsetLeft" on a paragraph element causes a crash.
        Reviewed by Chris

        Fixed <rdar://problem/3831063> regions use left offset instead of top offset
Reviewed by John

        * khtml/khtml_part.h:
        * khtml/rendering/render_object.cpp:
        (RenderObject::addDashboardRegions):
        * kwq/DOM.mm:
        (-[DOMNode isContentEditable]):
        (-[DOMNode KJS::Bindings::]):
        * kwq/KWQKHTMLPart.h:
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::executionContextForDOM):

WebBrowser:
        Added debugging code (disable) to assist in debugging DOM/WebScriptObjects.

        Reviewed by Chris.

        * LocationChangeHandler.m:
        (-[Logger logMessage:]):
        (-[LocationChangeHandler webView:locationChangeDone:forDataSource:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7820 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        Fix for this bug:

        <rdar://problem/3834779> Mail crashes when editing HTML message - khtml::Selection::layout()

        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::styleForSelectionStart): Table code seems to be more robust when the call to
        insert our style-checking node is done with an appendChild rather than an insertBefore. Note
        that this table-related problem was exposed by fixing Selection::layout(), which I did
        yesterday. This change simply improves things even more so that we do not crash in the
        scenario described in the bug.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7819 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John

        This is a partial fix to this bug:

        <rdar://problem/3832886> increase quote level on new mail document leads to immediate
        crash in caret painting code

        To eliminate the bad behavior for good, I have done some investigations in Mail code,
        and I have sent a suggested code change on to Grant. Basically, Mail can't add empty
        blocks (like blockquote elements used for quoting) to documents without giving those
        blocks some content (so they have a height).

        I added some other crash protections below.

        * khtml/editing/selection.cpp:
        (khtml::Selection::layout): Check for non-null position after calls to VisiblePosition, since
        the VisiblePosition constructors may fail to find a visible spot in the document. Also, add
        a couple position-has-renderer assertion checks.
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::styleForSelectionStart): Take out pos.isNotNull() assertion since VisiblePosition
        may not yield a position. This assertion is a holdover from before we had VisiblePosition.
        (KWQKHTMLPart::fontForSelection): Rearrange the code a little to deal with possible null results from
        calls to helpers.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7818 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3834230> empty table can result in division by 0

        * khtml/rendering/render_table.cpp: (RenderTableSection::layoutRows):
        Added 0 check; rolled in from KDE.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7817 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3802039> 8A259: Can't use Grab services to grab selection from screen

        Reviewed by john.

        * WebView.subproj/WebHTMLView.m:
        (+[WebHTMLView initialize]): register service "return types" which are types that can be inserted into a WebView
        (-[WebHTMLView writeSelectionToPasteboard:types:]): service protocol method, be sure to only write specified types
        (-[WebHTMLView readSelectionFromPasteboard:]): new, service protocol method, insert types
        (-[WebHTMLView validRequestorForSendType:returnType:]): moved, handle return types

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7816 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Added layout test for <input> value bug I just fixed.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7815 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3818712> form checkbox value property is read only

        The underlying problem was that we were storing two separate values for all
        form elements; one for the value property (JavaScript) and the other for the
        value attribute (DOM). This is a good idea for text input, but not for other types.

        * khtml/html/html_formimpl.h: Changed setValue to take a const DOMString reference.
        Added private storesValueSeparateFromAttribute function.
        * khtml/html/html_formimpl.cpp:
        (DOM::HTMLInputElementImpl::setType): Handle type changes, including detaching and re-attaching
        if type changed, and moving value from m_value to ATTR_VALUE and vice versa.
        (DOM::HTMLInputElementImpl::type): Added a case for ISINDEX and moved the default out of the
        switch so that we will get a warning if a type is left out.
        (DOM::HTMLInputElementImpl::parseHTMLAttribute): Tweaked comment format.
        (DOM::HTMLInputElementImpl::reset): Changed to only nuke the value if the value property is stored
        separately from the attribute. Otherwise, we just want to lave it alone
        (DOM::HTMLInputElementImpl::value): Changed to always use m_value if it's not null, then fall back
        on the attribute, and finally fall back to the "on" for the checkbox only if both are null.
        (DOM::HTMLInputElementImpl::setValue): Changed to set the attribute unless the value property is
        supposed to be stored separate from the attribute.
        (DOM::HTMLInputElementImpl::storesValueSeparateFromAttribute): Added. Returns true for text-type
        input elements, and false for the others.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7814 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3834130> nil-object-in-dictionary exception seen in -[WebView _elementAtWindowPoint:]

        * WebView.subproj/WebView.m: (-[WebView _elementAtWindowPoint:]): Added a check for nil frame.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7813 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3296652> checkbox input type does not respond to onchange

        * khtml/rendering/render_form.cpp:
        (RenderFormElement::updateFromElement): Some new code, commented out, for form element colors.
        (RenderCheckBox::slotStateChanged): Added call to onChange.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7812 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3834166> <input type=file> sends onchange even when the same file is chosen twice

        * WebCoreSupport.subproj/WebFileButton.m: (-[WebFileButton chooseFilename:]):
        Do nothing if filename is the same as before.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7811 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

        Reviewed by Darin

        Finish selection affinity implementation. This includes code to set the
        affinity correctly when clicking with the mouse, and clearing the
        affinity when altering the selection using any of the Selection object
        mutation functions.

        Each instance of the positionForCoordinates, inlineBox and caretRect
        functions have been changed to include an EAffinity argument to give results
        which take this bit into account.

        * khtml/editing/selection.cpp:
        (khtml::Selection::init): Default affinity is now UPSTREAM, to match AppKit.
        (khtml::Selection::modifyAffinity): New function to compute affinity based on
        modification constants.
        (khtml::Selection::moveTo): Reset affinity to UPSTREAM.
        (khtml::Selection::modifyExtendingRightForward): Ditto.
        (khtml::Selection::modifyMovingRightForward): Ditto.
        (khtml::Selection::modifyExtendingLeftBackward): Ditto.
        (khtml::Selection::modifyMovingLeftBackward): Ditto.
        (khtml::Selection::modify): Support saving, restoring, and then calculating new
        affinity value as needed.
        (khtml::Selection::xPosForVerticalArrowNavigation):
        (khtml::Selection::clear): Reset affinity to UPSTREAM.
        (khtml::Selection::setBase): Ditto.
        (khtml::Selection::setExtent): Ditto.
        (khtml::Selection::setBaseAndExtent): Ditto.
        (khtml::Selection::layout): Pass affinity to caretRect().
        (khtml::Selection::validate): Pass along affinity parameter to new functions that
        require it.
        (khtml::startOfFirstRunAt): Changed the way that the y-coordinate search is done, to
        keep this code working with changes made in selectionForLine().
        (khtml::endOfLastRunAt): Ditto.
        (khtml::selectionForLine): Make this function work for all renderers, not just text
        renderers.
        * khtml/editing/selection.h:
        (khtml::operator==): Consider affinity in equality check.
        * khtml/editing/visible_units.cpp:
        (khtml::previousLinePosition): Pass affinity argument to function so it can take this
        information into account while processing.
        (khtml::nextLinePosition): Ditto.
        (khtml::previousParagraphPosition): Ditto.
        (khtml::nextParagraphPosition): Ditto.
        * khtml/editing/visible_units.h: Ditto, for each of the functions listed.
        * khtml/khtml_events.cpp:
        (khtml::MouseEvent::offset): Rework code to remove dependence on NodeImpl::positionForCoordinates,
        as this function is being removed.
        * khtml/khtml_part.cpp:
        (KHTMLPart::isPointInsideSelection): Ditto.
        (KHTMLPart::selectClosestWordFromMouseEvent): Ditto.
        (KHTMLPart::handleMousePressEventTripleClick): Ditto.
        (KHTMLPart::handleMousePressEventSingleClick): Ditto. Plus, pass affinity argument in
        call to positionForCoordinates, and set resulting affinity on the selection.
        (KHTMLPart::handleMouseMoveEventSelection): Rework code to remove dependence on
        NodeImpl::positionForCoordinates, as this function is being removed.
        (KHTMLPart::khtmlMouseReleaseEvent): Ditto.
        * khtml/rendering/render_block.cpp:
        (khtml::RenderBlock::positionForCoordinates): Now takes an affinity argument.
        * khtml/rendering/render_block.h:
        * khtml/rendering/render_box.cpp:
        (RenderBox::caretRect): Ditto.
        * khtml/rendering/render_box.h:
        * khtml/rendering/render_br.cpp:
        (RenderBR::positionForCoordinates): Ditto.
        (RenderBR::caretRect): Ditto.
        (RenderBR::inlineBox): Ditto.
        * khtml/rendering/render_br.h:
        * khtml/rendering/render_container.cpp:
        (RenderContainer::positionForCoordinates): Ditto.
        * khtml/rendering/render_container.h:
        * khtml/rendering/render_flow.cpp:
        (RenderFlow::caretRect): Ditto.
        * khtml/rendering/render_flow.h:
        * khtml/rendering/render_inline.cpp:
        (RenderInline::positionForCoordinates): Ditto.
        * khtml/rendering/render_inline.h:
        * khtml/rendering/render_object.cpp:
        (RenderObject::caretRect): Ditto.
        (RenderObject::positionForCoordinates): Ditto.
        (RenderObject::inlineBox): Ditto.
        * khtml/rendering/render_object.h:
        * khtml/rendering/render_replaced.cpp:
        (RenderReplaced::positionForCoordinates): Ditto.
        * khtml/rendering/render_replaced.h:
        * khtml/rendering/render_text.cpp:
        (RenderText::positionForCoordinates): Ditto.
        (firstRendererOnNextLine): New helper used by caretRect().
        (RenderText::caretRect): Now takes an affinity argument.
        (RenderText::inlineBox): Ditto.
        * khtml/rendering/render_text.h:
        * khtml/xml/dom_nodeimpl.cpp: Remove positionForCoordinates helper.
        * khtml/xml/dom_nodeimpl.h: Ditto.
        * khtml/xml/dom_position.cpp:
        (DOM::Position::previousLinePosition): Now takes an affinity argument.
        (DOM::Position::nextLinePosition): Ditto.
        * khtml/xml/dom_position.h:
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge caretRectAtNode:offset:affinity:]): Ditto.
        (-[WebCoreBridge setSelectedDOMRange:affinity:]): Ditto.
        (-[WebCoreBridge _positionForPoint:]): Rework code to remove dependence on
        NodeImpl::positionForCoordinates, as this function is being removed.

WebKit:

        Reviewed by Darin

        * WebView.subproj/WebHTMLView.m:
        (-[WebTextCompleteController doCompletion]): bridge call to get caret rect at a node
        now takes an affinity: caretRectAtNode:offset:affinity:.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7810 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3670280> scroll position on overflowed textareas resets when leaving the tab

        * kwq/KWQTextArea.mm: (-[KWQTextAreaTextView becomeFirstResponder]):
        Scroll to reveal the text area, don't scroll to reveal the text view itself.
        Scrolling the text view ended up putting it at the top left, regardless of
        where the insertion point is.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7809 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3831546> More text is copied than is visually selected

        The bug here is that upstream was moving a position too far.

        * khtml/xml/dom_position.cpp:
        (DOM::Position::upstream): Use the "deep equivalent" node rather than the original node passed
        in for various checks. Also use local variables a bit more for slightly more efficiency.
        (DOM::Position::downstream): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7808 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Removed stray conflict marker

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7807 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Chris

        Fix for this bug:

        <rdar://problem/3814236> REGRESSION (Mail): Can't set the color of text in Mail compose window using drag/drag from color panel

        * WebView.subproj/WebHTMLView.m:
        (+[WebHTMLView _insertablePasteboardTypes]): Add NSColorPboardType to list.
        (-[WebHTMLView _isNSColorDrag:]): New helper. Determines if drag is an NSColor drag.
        (-[WebHTMLView draggingUpdatedWithDraggingInfo:actionMask:]): Add a case for NSColor drags,
        else do what we did before.
        (-[WebHTMLView concludeDragForDraggingInfo:actionMask:]): Add a case for NSColor drags, which creates
        a CSS style containing color info and calls the bridge to apply the style. Otherwise, do what we did before.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7806 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3833841> adding an event listener for keypress events does not work

        * khtml/xml/dom2_eventsimpl.h: Added numEventIds and made typeToId take a const DOMString &.
        * khtml/xml/dom2_eventsimpl.cpp:
        (EventImpl::typeToId): Changed to use table. Added "keypress", otherwise, the same as before.
        (EventImpl::idToType): Changed to use table.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7805 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3833848> REGRESSION (133-134): each keydown event is getting sent multiple times

        * WebView.subproj/WebHTMLView.m: (-[WebHTMLView performKeyEquivalent:]):
        Don't send an event through WebCore if it has already been through once.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7804 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Ken.

        - fixed <rdar://problem/3777253> Crash in redirect mechanism trying to display error page for bad scheme

        * WebView.subproj/WebMainResourceClient.m:
        (-[WebMainResourceClient connection:willSendRequest:redirectResponse:]):
        add retain/autorelease to the request returned from call to super. In this case, the return value
        was being dealloc'ed before being returned.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7803 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    - fixed <rdar://problem/3664375> repro crash in -[KWQAccObject accessibilityAttributeNames]

        (-[KWQAccObject accessibilityActionNames]):
        check for nil m_renderer

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7802 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin.

        - fixed <rdar://problem/3828147> REGRESSION: textareas with wrap="off" show their contents in a thin vertical line of text

        * kwq/KWQTextArea.h: Added setTextColor and setBackgroundColor methods. This is really for another fix
        I'm landing later, but it does no harm to add these now.
        * kwq/KWQTextArea.mm:
        (-[KWQTextArea _configureTextViewForWordWrapMode]): Set the container size after changing the flag that
        determines if the width tracks the text view. Otherwise, we won't successfully set the width in the case
        where we don't want it to track the text view. This caused the bug.
        (-[KWQTextArea _createTextView]): Remove unneeded call to setMaxSize. The above method already does that.
        (-[KWQTextArea setTextColor:]): Added.
        (-[KWQTextArea setBackgroundColor:]): Added.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7801 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Adele.

        - fixed <rdar://problem/3829452> REGRESSION (156-157): onload handler doesn't run on page with meta refresh of 0 duration (new Apple start page)

        The fix for <rdar://problem/3773150> made it so <meta> redirects prevent tokenizing the rest of the page.
        This is incorrect; the reason the JavaScript-triggered loads prevent tokenizing is that they take place
        "right away" in other browsers, but that is not true of <meta> redirect. We fixed this by using a separate
        call for <meta> redirect and not preventing tokenizing when that's in effect.

        * khtml/khtml_part.h: Removed userGesture parameter from scheduleRedirection. Renamed
        isImmediateRedirectPending to isScheduledLocationChangePending. Added scheduleLocationChange.
        * khtml/khtml_part.cpp:
        (KHTMLPart::openURL): Updated for new constant name.
        (KHTMLPart::scheduleRedirection): Removed now-unneeded userGesture parameter, and removed code that
        does the special case for redirection during load; a <meta> refresh can never be one of those special
        redirects during a load because it redirects the frame itself, not another frame. Also tightened up
        the logic by always stopping the redirect timer even if we aren't restarting it.
        (KHTMLPart::scheduleLocationChange): Added. Like scheduleRedirection, but with a different constant
        so we can tell it apart and always a delay of 0. The "redirection during load" case was moved in here
        and renamed to locationChangeScheduledDuringLoad.
        (KHTMLPart::isScheduledLocationChangePending): Renamed from isImmediateRedirectPending. This now
        returns true only for location changes and history navigation, not <meta> redirects.
        (KHTMLPart::scheduleHistoryNavigation): Tightened up logic to do the stop() outside the if as above,
        and got rid of a silly timer delay computation that always resulted in 0.

        * khtml/khtmlpart_p.h: Added a new value to the RedirectionScheduled enum for scheduleLocationChange
        and also renamed one of the existing values.

        * khtml/html/htmltokenizer.cpp: (HTMLTokenizer::write): Changed to use isScheduledLocationChangePending instead
        of isImmediateRedirectPending, because we do want to continue tokenizing if it's actually a redirect.

        * khtml/ecma/kjs_html.cpp: (KJS::HTMLDocument::putValue): Changed to call the new scheduleLocationChange
        instead of calling scheduleRedirection with delay of 0.
        * khtml/ecma/kjs_window.cpp:
        (Window::put): Ditto.
        (WindowFunc::tryCall): Ditto.
        (Location::put): Ditto.
        (LocationFunc::tryCall): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7800 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin.

        - fixed <rdar://problem/3658277> REGRESSION (1.1-1.2): form submission should either not simulate a click at all or use (0,0) the way Mozilla does

        * khtml/html/html_elementimpl.cpp: (HTMLElementImpl::click): Use 0,0 for the coordinates.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7799 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin.

        - fixed <rdar://problem/3804661> REGRESSION: JavaScriptCore framework now has two init routines

        * bindings/NP_jsobject.cpp: Fixed unnecessarily-complex globals set up that was
        creating an init routine.

        * kjs/ustring.cpp: Changed around the UString::Rep::empty construction to not
        require a global constructor that creates an init routine.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7798 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin.

        - fixed <rdar://problem/3804665> REGRESSION: WebCore framework now has many init routines

        * khtml/xml/dom_nodeimpl.h: Changed anyQName declaration to not use the inline function
        makeId. Surprisingly, the inline function was not "constant-folded" and we ended up with
        a copy of the function in each file as an init routine for the framework.

        * khtml/ecma/kjs_html.cpp: (Gradient::colorStops): Get rid of initialized ColorStop
        globals; their constructors were showing up as init routines for the framework.

        * khtml/rendering/render_style.h: Got rid of inline initialDashboardRegions function.
        * khtml/rendering/render_style.cpp: (RenderStyle::initialDashboardRegions): Made this
        a normal function. When it was an inline function, the constructors for the per-file
        copies of the globals were showing up as init routines for the framework.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7797 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Added layout test for just-fixed JavaScriptCore bug.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7796 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Forgot to include that I also fixed <rdar://problem/3397658> scroll wheel does not work to scroll overflow:auto/scroll/overlay areas (RSS)

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7795 268f45cc-cd09-0410-ab3c-d52691b4dbfc

WebCore:

Fixed:
<rdar://problem/3625352> up and down arrow and page up/down keys don't work to scroll overflow:auto/scroll/overlay areas
<rdar://problem/3822027> REGRESSION (Mail): When selection moves out of visible area, should center as NSText does

        Reviewed by hyatt, kocienda.

        * khtml/rendering/render_layer.cpp:
        (RenderLayer::scroll): new
        * khtml/rendering/render_layer.h:
        * khtml/rendering/render_object.cpp:
        (RenderObject::scroll): new
        * khtml/rendering/render_object.h:
        * kwq/KWQKHTMLPart.h:
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::scrollOverflow): new
        (KWQKHTMLPart::scrollOverflowWithScrollWheelEvent): new
        (KWQKHTMLPart::khtmlMousePressEvent): store pressed node so we know where the focus is
        * kwq/KWQScrollBar.h:
        * kwq/KWQScrollBar.mm:
        (QScrollBar::setValue): return a bool
        (QScrollBar::scrollbarHit): ditto
        (QScrollBar::scroll): new
        * kwq/WebCoreBridge.h:
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge scrollOverflowInDirection:granularity:]): new
        (-[WebCoreBridge scrollOverflowWithScrollWheelEvent:]): new
        (-[WebCoreBridge ensureSelectionVisible]): visually center the extent of the selection

WebKit:

Fixed: <rdar://problem/3625352> up and down arrow and page up/down keys don't work to scroll overflow:auto/scroll/overlay areas

        Reviewed by hyatt.

        * Plugins.subproj/WebBaseNetscapePluginStream.m:
        (-[WebBaseNetscapePluginStream initWithRequestURL:pluginPointer:notifyData:sendNotification:]): fixed typo in comment
        * Plugins.subproj/WebNetscapePluginStream.m:
        (-[WebNetscapePluginStream initWithRequest:pluginPointer:notifyData:sendNotification:]): ditto
        * WebView.subproj/WebFramePrivate.h:
        * WebView.subproj/WebFrameView.m:
        (-[WebFrameView _bridge]): new
        (-[WebFrameView scrollToBeginningOfDocument:]): call the bridge to scroll, if that fails, scroll the document view
        (-[WebFrameView scrollToEndOfDocument:]): ditto
        (-[WebFrameView _pageVertically:]): ditto
        (-[WebFrameView _pageHorizontally:]): ditto
        (-[WebFrameView _scrollLineVertically:]): ditto
        (-[WebFrameView _scrollLineHorizontally:]): ditto
        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView scrollWheel:]): call the bridge to scroll, if that fails, pass to next responder

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7794 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Kevin.

        - fixed <rdar://problem/3822618> REGRESSION (164-165): expedia.com's popup help doesn't work

        * kjs/reference.cpp: (Reference::putValue): Change so that references not found in any object
        work with the window object of the page the function is in, not the page of the caller. This
        is what all other browsers do. This code was hidden before by the "everything is defined on
        window object" hack in WebCore.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7793 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed build error.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7787 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Added simple JavaScript call tracing.  Very useful for
        debugging complex pages.

Tracing is only available in development builds and is
enabled by:

(gdb) set traceJavaScript = 1

or programatically

setTraceJavaScript(true)

Function, args, and return values are printed to console.  Very
verbose.

        Reviewed by Ken.

        * kjs/function_object.cpp:
        (FunctionProtoFuncImp::call):
        * kjs/object.cpp:
        (KJS::Object::call):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7786 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Back out style sharing perf fix.

        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::locateCousinList):
        (khtml::CSSStyleSelector::canShareStyleWithElement):
        (khtml::CSSStyleSelector::locateSharedStyle):
        * khtml/css/cssstyleselector.h:
        * khtml/html/html_elementimpl.h:
        (DOM::HTMLElementImpl::inlineStyleDecl):
        * khtml/xml/dom_elementimpl.cpp:
        (ElementImpl::recalcStyle):
        * khtml/xml/dom_elementimpl.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7785 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Really bring the new code in line with the old style sharing code.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7784 268f45cc-cd09-0410-ab3c-d52691b4dbfc

reversion JSC, for the -Wno-long-double changes

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7782 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Add -Wno-long-double back to our warning flags.  Without it, the x86 build fails.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7781 268f45cc-cd09-0410-ab3c-d52691b4dbfc

versioning for TOT, Safari 2.0 (v167u)

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7780 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Safari-166 stamp for everything except WebBrowser.  In these projects, CFBundleShortVersionString matches CFBundleVersion (166 for both).

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7779 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix a bug in the iteration of locateCousinList and clean up the style sharing stats code.

        Reviewed by NOBODY (OOPS!).

        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::locateCousinList):
        (khtml::CSSStyleSelector::elementsCanShareStyle):
        (khtml::CSSStyleSelector::locateSharedStyle):
        (khtml::CSSStyleSelector::styleForElement):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7778 268f45cc-cd09-0410-ab3c-d52691b4dbfc

JavaScriptCore:
        Fixed <rdar://problem/3819234> NPN_SetException (and throwException:) isn't implemented

        Reviewed by Chris.

        * bindings/NP_jsobject.cpp:
        (_NPN_SetException):
        * bindings/npruntime.cpp:
        (_NPN_SetExceptionWithUTF8):
        * bindings/objc/WebScriptObject.mm:
        (+[WebScriptObject throwException:]):
        * kjs/internal.h:
        (KJS::InterpreterImp::context):

WebBrowser:
Added debugging support for:

        Fixed <rdar://problem/3819234> NPN_SetException (and throwException:) isn't implemented

        Logging is not enabled in either development or deployment build.

        Reviewed by Chris.

        * LocationChangeHandler.m:
        (+[Logger isSelectorExcludedFromWebScript:]):
        (-[Logger logMessage:]):
        (-[LocationChangeHandler webView:windowScriptObjectAvailable:]):
        (-[LocationChangeHandler webView:locationChangeDone:forDataSource:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7777 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        * khtml/rendering/bidi.cpp:
        (khtml::RenderBlock::computeHorizontalPositionsForLine): Fix coding mistake that
        broke layout tests involving compacts.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7776 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix style sharing optimization to be fast again.  Go back to using pointer comparisons when looking for
cousins to share.

        Reviewed by darin

        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::locateCousinList):
        * khtml/rendering/render_object.cpp:
        (RenderObject::setStyleInternal):
        * khtml/rendering/render_object.h:
        * khtml/xml/dom_elementimpl.cpp:
        (ElementImpl::recalcStyle):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7775 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3827002> assertion failure in WebBaseNetscapePluginStream on abc.go.com

        Reviewed by rjw.

        * Plugins.subproj/WebBaseNetscapePluginStream.m:
        (-[WebBaseNetscapePluginStream initWithRequestURL:pluginPointer:notifyData:sendNotification:]): avoid assertion failure in dealloc by temporarily setting isTerminated to YES in case we are released in this method
        * Plugins.subproj/WebNetscapePluginStream.m:
        (-[WebNetscapePluginStream initWithRequest:pluginPointer:notifyData:sendNotification:]): ditto

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7774 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix lists so that they properly participate in line layout as though they are text (when text bullets are
used) and as images (when image bullets are used).

        Reviewed by kocienda

        * khtml/rendering/render_list.cpp:
        (RenderListMarker::createInlineBox):
        * khtml/rendering/render_list.h:
        (khtml::ListMarkerBox:::InlineBox):
        (khtml::ListMarkerBox::isText):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7773 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * WebCoreSupport.subproj/WebBridge.m:
        (-[WebBridge pluginViewWithPackage:attributeNames:attributeValues:baseURL:]):
        initialize "arguments" var to nil to satisfy compiler on deployment build.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7772 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fixed <rdar://problem/3821515> binding layer needs to convert NSNumber-bools to js type boolean not number

        Reviewed by Ken.

        * bindings/objc/objc_utility.mm:
        (KJS::Bindings::convertObjcValueToValue):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7771 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Fixed <rdar://problem/3825442> first click lost for Dashboard
Allow dashboard to force acceptsFirstMouse:

        Reviewed by Chris.

        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView acceptsFirstMouse:]):
        * WebView.subproj/WebView.m:
        (-[WebView _dashboardBehavior:]):
        * WebView.subproj/WebViewInternal.h:
        * WebView.subproj/WebViewPrivate.h:

        * WebCoreSupport.subproj/WebImageRenderer.h:
        Comment change only

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7770 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Back out the assertion.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7769 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3760920> Need to record plugin view instances

        Reviewed by rjw.

        * Plugins.subproj/WebPluginController.h:
        * Plugins.subproj/WebPluginController.m:
        (+[WebPluginController plugInViewWithArguments:fromPluginPackage:]): new, creates plug-in view and adds it to global list
        (+[WebPluginController isPlugInView:]): new, checks if the plug-in view is in the global list
        (-[WebPluginController destroyAllPlugins]): remove the plug-in from the global list
        * Plugins.subproj/WebPluginDocumentView.m:
        (-[WebPluginDocumentView setDataSource:]): call [WebPluginController plugInViewWithArguments:fromPluginPackage:]
        * WebCoreSupport.subproj/WebBridge.m:
        (-[WebBridge pluginViewWithPackage:attributeNames:attributeValues:baseURL:]): ditto
        * WebView.subproj/WebFrame.m:
        (-[WebFrame _reloadForPluginChanges]): call [WebPluginController isPlugInView:]
        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView addSubview:]): ditto

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7768 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin

        Recent checkin adding upstreamDeepEquivalent had it backwards. The helper
        we want is downstreamDeepEquivalent, as the deepEquivalent function returns
        an upstream position.

        * khtml/editing/selection.cpp:
        (khtml::Selection::layout): DOWNSTREAM case now uses downstreamDeepEquivalent.
        UPSTREAM uses deepEquivalent.
        * khtml/editing/visible_position.cpp:
        (khtml::VisiblePosition::downstreamDeepEquivalent): New helper, replacing
        upstreamDeepEquivalent.
        * khtml/editing/visible_position.h

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7767 268f45cc-cd09-0410-ab3c-d52691b4dbfc

New selection gap-filling architecture.  Makes the gap-filling much more like NSTextView and puts the responsibility
for gap-filling in the block.  Fixes numerous bugs with selection drawing including bidi issues, incorrect old
horizontal gap filling, and selection performance issues.

        Reviewed by kocienda

        * khtml/html/html_imageimpl.cpp:
        (HTMLImageLoader::notifyFinished):
        * khtml/misc/khtmllayout.h:
        (khtml::GapRects::left):
        (khtml::GapRects::center):
        (khtml::GapRects::right):
        (khtml::GapRects::uniteLeft):
        (khtml::GapRects::uniteCenter):
        (khtml::GapRects::uniteRight):
        (khtml::GapRects::unite):
        (khtml::GapRects::operator QRect):
        (khtml::GapRects::operator==):
        (khtml::GapRects::operator!=):
        * khtml/rendering/font.cpp:
        (Font::drawHighlightForText):
        * khtml/rendering/font.h:
        * khtml/rendering/render_block.cpp:
        (khtml:::RenderFlow):
        (khtml::RenderBlock::removeChild):
        (khtml::RenderBlock::paintObject):
        (khtml::RenderBlock::paintEllipsisBoxes):
        (khtml::RenderBlock::setSelectionState):
        (khtml::RenderBlock::shouldPaintSelectionGaps):
        (khtml::RenderBlock::isSelectionRoot):
        (khtml::RenderBlock::selectionGapRects):
        (khtml::RenderBlock::paintSelection):
        (khtml::RenderBlock::fillSelectionGaps):
        (khtml::RenderBlock::fillInlineSelectionGaps):
        (khtml::RenderBlock::fillBlockSelectionGaps):
        (khtml::RenderBlock::fillHorizontalSelectionGap):
        (khtml::RenderBlock::fillVerticalSelectionGap):
        (khtml::RenderBlock::fillLeftSelectionGap):
        (khtml::RenderBlock::fillRightSelectionGap):
        (khtml::RenderBlock::getHorizontalSelectionGapInfo):
        (khtml::RenderBlock::leftSelectionOffset):
        (khtml::RenderBlock::rightSelectionOffset):
        * khtml/rendering/render_block.h:
        (khtml::RenderBlock::hasSelectedChildren):
        (khtml::RenderBlock::selectionState):
        (khtml::RenderBlock::BlockSelectionInfo::BlockSelectionInfo):
        (khtml::RenderBlock::BlockSelectionInfo::rects):
        (khtml::RenderBlock::BlockSelectionInfo::state):
        (khtml::RenderBlock::BlockSelectionInfo::block):
        (khtml::RenderBlock::selectionRect):
        * khtml/rendering/render_box.cpp:
        (RenderBox::position):
        * khtml/rendering/render_br.cpp:
        (RenderBR::inlineBox):
        * khtml/rendering/render_br.h:
        (khtml::RenderBR::selectionRect):
        (khtml::RenderBR::paint):
        * khtml/rendering/render_canvas.cpp:
        (RenderCanvas::selectionRect):
        (RenderCanvas::setSelection):
        * khtml/rendering/render_canvasimage.cpp:
        (RenderCanvasImage::paint):
        * khtml/rendering/render_image.cpp:
        (RenderImage::paint):
        * khtml/rendering/render_image.h:
        * khtml/rendering/render_line.cpp:
        (khtml::InlineBox::nextLeafChild):
        (khtml::InlineBox::prevLeafChild):
        (khtml::InlineBox::selectionState):
        (khtml::InlineFlowBox::addToLine):
        (khtml::InlineFlowBox::firstLeafChild):
        (khtml::InlineFlowBox::lastLeafChild):
        (khtml::InlineFlowBox::firstLeafChildAfterBox):
        (khtml::InlineFlowBox::lastLeafChildBeforeBox):
        (khtml::InlineFlowBox::selectionState):
        (khtml::RootInlineBox::fillLineSelectionGap):
        (khtml::RootInlineBox::setHasSelectedChildren):
        (khtml::RootInlineBox::selectionState):
        (khtml::RootInlineBox::firstSelectedBox):
        (khtml::RootInlineBox::lastSelectedBox):
        (khtml::RootInlineBox::selectionTop):
        (khtml::RootInlineBox::block):
        * khtml/rendering/render_line.h:
        (khtml::RootInlineBox::RootInlineBox):
        (khtml::RootInlineBox::hasSelectedChildren):
        (khtml::RootInlineBox::selectionHeight):
        * khtml/rendering/render_object.cpp:
        (RenderObject::selectionColor):
        * khtml/rendering/render_object.h:
        (khtml::RenderObject::):
        (khtml::RenderObject::selectionState):
        (khtml::RenderObject::setSelectionState):
        (khtml::RenderObject::selectionRect):
        (khtml::RenderObject::canBeSelectionLeaf):
        (khtml::RenderObject::hasSelectedChildren):
        (khtml::RenderObject::hasDirtySelectionState):
        (khtml::RenderObject::setHasDirtySelectionState):
        (khtml::RenderObject::shouldPaintSelectionGaps):
        (khtml::RenderObject::SelectionInfo::SelectionInfo):
        * khtml/rendering/render_replaced.cpp:
        (RenderReplaced::RenderReplaced):
        (RenderReplaced::shouldPaint):
        (RenderReplaced::selectionRect):
        (RenderReplaced::setSelectionState):
        (RenderReplaced::selectionColor):
        (RenderWidget::paint):
        (RenderWidget::setSelectionState):
        * khtml/rendering/render_replaced.h:
        (khtml::RenderReplaced::canBeSelectionLeaf):
        (khtml::RenderReplaced::selectionState):
        * khtml/rendering/render_text.cpp:
        (InlineTextBox::checkVerticalPoint):
        (InlineTextBox::isSelected):
        (InlineTextBox::selectionState):
        (InlineTextBox::selectionRect):
        (InlineTextBox::paintSelection):
        (InlineTextBox::paintMarkedTextBackground):
        (RenderText::paint):
        (RenderText::setSelectionState):
        (RenderText::selectionRect):
        * khtml/rendering/render_text.h:
        (khtml::RenderText::canBeSelectionLeaf):
        * kwq/KWQPainter.h:
        * kwq/KWQPainter.mm:
        (QPainter::drawHighlightForText):
        * kwq/KWQPtrDict.h:
        (QPtrDictIterator::toFirst):
        * kwq/KWQRect.mm:
        (QRect::unite):
        * kwq/WebCoreTextRenderer.h:
        * kwq/WebCoreTextRendererFactory.mm:
        (WebCoreInitializeEmptyTextGeometry):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7766 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fix to make selection more like NSTextView.  All gap painting is now done by WebCore, so WebKit no longer
needs to try to fill gaps around text.

        Reviewed by kocienda

        * WebCoreSupport.subproj/WebTextRenderer.m:
        (-[WebTextRenderer _CG_drawHighlightForRun:style:geometry:]):
        (-[WebTextRenderer _ATSU_drawHighlightForRun:style:geometry:]):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7765 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin

        Use the new CSS properties I added with my previous check-in. Also makes
        some changes to caret positioning and drawing to make the proper editing
        end-of-line behavior work correctly.

        * khtml/editing/selection.cpp:
        (khtml::Selection::layout): Caret drawing now takes affinity into account
        when deciding where to paint the caret (finally!).
        * khtml/editing/visible_position.cpp:
        (khtml::VisiblePosition::previousVisiblePosition): Move off Position::rendersInDifferentPosition
        to determine the result. Use a simpler test involving comparisons between
        downstream positions while iterating. This is cheaper to do and easier to understand.
        (khtml::VisiblePosition::nextVisiblePosition): Ditto.
        * khtml/rendering/bidi.cpp:
        (khtml::BidiIterator::current): Do not return non-breaking spaces for empty
        text renderers and for non-text renderers. Return a null Qchar instead. Returning
        non-breaking spaces was causing errors when the new -khtml-nbsp-mode was set to "space".
        (khtml::RenderBlock::computeHorizontalPositionsForLine): Shrink line boxes that
        contain with more spaces than can fit on the end of a line.
        (khtml::RenderBlock::skipWhitespace): Factor this out from findNextLineBreak.
        (khtml::RenderBlock::findNextLineBreak): Use new skipWhitespace function. Add
        in code to check and use new CSS properties.
        * khtml/rendering/break_lines.cpp:
        (khtml::isBreakable): Consider a non-breaking space a breakable character based
        on setting of new -khtml-nbsp-mode property.
        * khtml/rendering/break_lines.h: Ditto.
        * khtml/rendering/render_block.h: Declare skipWhitespace function.
        * khtml/rendering/render_text.cpp:
        (RenderText::caretRect): Do not draw the caret beyond the right edge of the
        window when in white-space normal mode.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7764 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Darin

        Fix for these bugs:

        In this patch, I add two new CSS properties and their associated behavior.
        This is to support end-of-line and word-wrapping features that match the
        conventions of text editors.

        There are also some other small changes here which begin to lay the groundwork
        for using these new properties to bring about the desired editing behavior.

        * khtml/css/cssparser.cpp:
        (CSSParser::parseValue): Add support for new CSS properties.
        * khtml/css/cssproperties.c: Generated file.
        * khtml/css/cssproperties.h: Ditto.
        * khtml/css/cssproperties.in: Add new properties.
        * khtml/css/cssstyleselector.cpp:
        (khtml::CSSStyleSelector::applyProperty): Add support for new CSS properties.
        * khtml/css/cssvalues.c: Generated file.
        * khtml/css/cssvalues.h: Ditto.
        * khtml/css/cssvalues.in:  Add support for new CSS properties.
        * khtml/editing/visible_position.cpp:
        (khtml::VisiblePosition::upstreamDeepEquivalent): Added new helper.
        * khtml/editing/visible_position.h:
        * khtml/rendering/render_box.cpp:
        (RenderBox::deleteLineBoxWrapper): Zero out inlineBoxWrapper.
        * khtml/rendering/render_replaced.cpp:
        (RenderWidget::detach): Zero out inlineBoxWrapper.
        * khtml/rendering/render_style.cpp:
        (StyleCSS3InheritedData):
        (StyleCSS3InheritedData::operator==): Add support for new CSS properties.
        (RenderStyle::diff): Ditto.
        * khtml/rendering/render_style.h:
        (khtml::RenderStyle::nbspMode): Ditto.
        (khtml::RenderStyle::khtmlLineBreak): Ditto.
        (khtml::RenderStyle::setNBSPMode): Ditto.
        (khtml::RenderStyle::setKHTMLLineBreak): Ditto.
        (khtml::RenderStyle::initialNBSPMode): Ditto.
        (khtml::RenderStyle::initialKHTMLLineBreak): Ditto.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7763 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3577255> custom file icon shows up upside down in <input type=file>

        * WebCoreSupport.subproj/WebFileButton.m: (-[WebFileButton setFilename:]): Added a call to
        setFlipped that fixes the problem, even though I don't know why.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7762 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3673150> Pasting string from clipboard that is longer than input box will accept fails rather than truncating

        * kwq/KWQTextField.mm:
        (-[KWQTextFieldFormatter isPartialStringValid:proposedSelectedRange:originalString:originalSelectedRange:errorDescription:]):
        Wrote a new version of this method that truncates incoming strings rather than rejecting them
        out of hand.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7761 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Maciej.

        - fixed <rdar://problem/3814237> REGRESSION (Mail): Copy/paste style does not set color in Mail compose window

        * WebView.subproj/WebHTMLView.m:
        (-[WebHTMLView _selectionStartFontAttributesAsRTF]): Changed to call new bridge method
        named fontAttributesForSelectionStart, deleted the method this used to use, and renamed
        this to have the word "start" in it.
        (-[WebHTMLView copyFont:]): Updated for name change.

        * English.lproj/StringsNotToBeLocalized.txt: Updated for recent changes.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7760 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Maciej.

        - fixed <rdar://problem/3826343> crash in KHTMLParser::setCurrent parsing document fragment (happens in Calendar widget)

        * khtml/html/htmlparser.cpp:
        (KHTMLParser::KHTMLParser): Initialized currentIsReferenced to false (fixes bug).
        (KHTMLParser::reset): Use doc() to make code easier to read.
        (KHTMLParser::setCurrent): Ditto.
        (KHTMLParser::parseToken): Ditto.
        (KHTMLParser::insertNode): Ditto.
        (KHTMLParser::getElement): Ditto.
        (KHTMLParser::popOneBlock): Ditto.

        - fixed <rdar://problem/3814237> REGRESSION (Mail): Copy/paste style does not set color in Mail compose window

        * kwq/KWQKHTMLPart.h: Added fontAttributesForSelectionStart.
        * kwq/KWQKHTMLPart.mm:
        (KWQKHTMLPart::styleForSelectionStart): Factored out most of fontForSelection.
        (KWQKHTMLPart::fontAttributesForSelectionStart): Added.
        (KWQKHTMLPart::fontAttributesForSelectionStart): Added.
        (KWQKHTMLPart::registerCommandForUndo): Updated for name change (see below).
        (KWQKHTMLPart::registerCommandForRedo): Ditto.

        * kwq/WebCoreBridge.h: Added fontAttributesForSelectionStart.
        * kwq/WebCoreBridge.mm:
        (-[WebCoreBridge undoEditing:]): Updated for name change (see below).
        (-[WebCoreBridge redoEditing:]): Ditto.
        (-[WebCoreBridge replaceSelectionWithFragment:selectReplacement:smartReplace:]): Got rid of named temporary
        EditCommandPtr variable to make things slightly more terse.
        (-[WebCoreBridge moveSelectionToDragCaret:smartMove:]): Ditto.
        (-[WebCoreBridge deleteSelectionWithSmartDelete:]): Ditto.
        (-[WebCoreBridge fontAttributesForSelectionStart]): Added.

        - fix compile on Panther and other cleanup

        * khtml/khtml_part.cpp: Removed unneeded include.
        * kwq/KWQEditCommand.h: Got rid of use of "impl" when referring to EditCommand pointers.
        * kwq/KWQEditCommand.mm: Added include so we compile on Panther.
        (-[KWQEditCommand initWithEditCommand:]): Changed name.
        (-[KWQEditCommand dealloc]): Updated for m_impl change to m_command.
        (-[KWQEditCommand finalize]): Ditto.
        (+[KWQEditCommand commandWithEditCommand:]): Changed name.
        (-[KWQEditCommand command]): Changed name.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7759 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    * WebView.subproj/WebFrameInternal.h: removed constant declarations that I committed by mistake

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7758 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - did a more-robust version of the fix I just landed

        * khtml/html/htmlparser.h: Added currentIsReferenced boolean.
        * khtml/html/htmlparser.cpp:
        (KHTMLParser::KHTMLParser): Initializes currentIsReferenced.
        (KHTMLParser::setCurrent): Changed to respect and set currentIsReferenced.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7757 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by John.

        - fixed <rdar://problem/3824393> REGRESSION (165-TOT): Crash in KHTMLParser::popOneBlock closing window (bose.com)

        * khtml/html/htmlparser.cpp:
        (KHTMLParser::~KHTMLParser): Move call to setCurrent(0) after the call to freeBlock, since freeBlock doesn't
        work well when current is 0, and there's no reason we need to reset the current block first.
        (KHTMLParser::setCurrent): Don't ever hold a reference to the document. This prevents a situation where there
        would be a reference cycle. In the test case from the bug above, this cycle actually happened and resulted
        in a double-delete of the document, tokenizer, and parser.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7756 268f45cc-cd09-0410-ab3c-d52691b4dbfc

Fixed: <rdar://problem/3798948> NPP_URLNotify is not called if plug-in calls NPN_*URLNotfy
Fixed a number of FIXME's related to notifying plug-ins of loaded pages.

        Reviewed by rjw.

        * Plugins.subproj/WebBaseNetscapePluginStream.h:
- replaced URL ivar with requestURL and responseURL ivars since we need to pass both to plug-ins
         - added sendNotification boolean. Relying on notifyData not being NULL was not information to know whether to call NPP_URLNotify or not.
- added isTerminated boolean because determining whether or not stream.ndata is NULL is not enough to know if the stream has been cancelled.
* Plugins.subproj/WebBaseNetscapePluginStream.m:
        (+[WebBaseNetscapePluginStream reasonForError:]): new, factored out from receivedError:
        (-[WebBaseNetscapePluginStream initWithRequestURL:pluginPointer:notifyData:sendNotification:]): new
        (-[WebBaseNetscapePluginStream dealloc]): release new ivars
        (-[WebBaseNetscapePluginStream finalize]): added assert
        (-[WebBaseNetscapePluginStream setRequestURL:]): new
        (-[WebBaseNetscapePluginStream setResponseURL:]): new
        (-[WebBaseNetscapePluginStream startStreamResponseURL:expectedContentLength:lastModifiedDate:MIMEType:]): renamed, use responseURL as it basically did before
        (-[WebBaseNetscapePluginStream startStreamWithResponse:]): call renamed method
        (-[WebBaseNetscapePluginStream destroyStream]):
- do nothing if terminated
- call NPP_StreamAsFile and NPP_DestroyStream if stream.ndata is not NULL
- call NPP_URLNotify if sendNotification is YES regardless of value of notifyData
        (-[WebBaseNetscapePluginStream receivedError:]): call reasonForError
        (-[WebBaseNetscapePluginStream deliverData]): use renamed ivar
        * Plugins.subproj/WebBaseNetscapePluginView.h:
- added observingFrameLoadNotification boolean
- renamed dictionary ivar to pendingFrameLoads which has WebFrame keys and WebPluginRequest values
        * Plugins.subproj/WebBaseNetscapePluginView.m:
        (-[WebBaseNetscapePluginView addFrameLoadObserver]): new
        (-[WebBaseNetscapePluginView removeFrameLoadObserver]): new
        (-[WebBaseNetscapePluginView stop]): call removeFrameLoadObserver
        (-[WebBaseNetscapePluginView initWithFrame:]): use renamed pendingFrameLoads ivar
        (-[WebBaseNetscapePluginView dealloc]): ditto
        (-[WebBaseNetscapePluginView requestWithURLCString:]): set referrer on the request just as IE does
        (-[WebBaseNetscapePluginView evaluateJavaScriptPluginRequest:]):
- call NPP_URLNotify depending of value of sendNotification
- call new init method on WebBaseNetscapePluginStream rather then setting variables individually
        (-[WebBaseNetscapePluginView webFrame:didFinishLoadWithReason:]): new, calls NPP_URLNotify at the right time with the right value
        (-[WebBaseNetscapePluginView webFrame:didFinishLoadWithError:]): new, delegate method called from WebFrame
        (-[WebBaseNetscapePluginView loadPluginRequest:]): call addFrameLoadObserver
        (-[WebBaseNetscapePluginView loadRequest:inTarget:withNotifyData:sendNotification:]): take new sendNotification parameter and pass it
        (-[WebBaseNetscapePluginView getURLNotify:target:notifyData:]): pass YES for sendNotification
        (-[WebBaseNetscapePluginView getURL:target:]): pass NO for sendNotification
        (-[WebBaseNetscapePluginView _postURL:target:len:buf:file:notifyData:sendNotification:allowHeaders:]): take new sendNotification parameter and pass it
        (-[WebBaseNetscapePluginView postURLNotify:target:len:buf:file:notifyData:]): pass YES for sendNotification
        (-[WebBaseNetscapePluginView postURL:target:len:buf:file:]): pass NO for sendNotification
        (-[WebPluginRequest initWithRequest:frameName:notifyData:sendNotification:]): take new sendNotification parameter
        (-[WebPluginRequest sendNotification]): new
        * Plugins.subproj/WebBaseNetscapePluginViewPrivate.h:
        * Plugins.subproj/WebNetscapePluginEmbeddedView.m:
        (-[WebNetscapePluginEmbeddedView didStart]): set referrer on the request just as IE does
        * Plugins.subproj/WebNetscapePluginRepresentation.m:
        (-[WebNetscapePluginRepresentation receivedData:withDataSource:]): set the request URL on the stream
        * Plugins.subproj/WebNetscapePluginStream.h:
        * Plugins.subproj/WebNetscapePluginStream.m:
        (-[WebNetscapePluginStream initWithRequest:pluginPointer:notifyData:sendNotification:]): take new sendNotification parameter and pass it
        (-[WebNetscapePluginStream dealloc]): use renamed ivar
        (-[WebNetscapePluginStream start]): ditto
        * WebView.subproj/WebFrame.m:
        (-[WebFrame _setState:]): removed notification posting code. This was only used by WebBaseNetscapePluginView and it was the wrong notification to send.
        (-[WebFrame _checkLoadCompleteForThisFrame]): call internal load delegate to tell it that the load has finished
        (-[WebFrame _loadItem:withLoadType:]): ditto
        (-[WebFrame _continueFragmentScrollAfterNavigationPolicy:formState:]): ditto
        (-[WebFrame _setInternalLoadDelegate:]): new
        (-[WebFrame _internalLoadDelegate]): new
        * WebView.subproj/WebFrameInternal.h:
        * WebView.subproj/WebFramePrivate.h:

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7755 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Maciej.

        - fixed <rdar://problem/3825429> onclick handler called when mouse down on another element (affects Dashboard Movies widget, test case enclosed)

        * khtml/khtmlview.cpp:
        (KHTMLViewPrivate::KHTMLViewPrivate): Initialize the click node to 0.
        (KHTMLViewPrivate::~KHTMLViewPrivate): Deref the click node.
        (KHTMLViewPrivate::reset): Clear the click node.
        (KHTMLView::viewportMousePressEvent): Call invalidateClick when we pass the event to a subframe to
        reduce the chance that we'll hold on to an old click node for a long time. Set the click node to
        the node we we are clicking on.
        (KHTMLView::viewportMouseDoubleClickEvent): Only send a click even if the node is the same one from
        the original click. Call invalidateClick when done to reduce the chance that we'll hold on to an
        old click node for a long time.
        (KHTMLView::invalidateClick): Clear the click node.
        (KHTMLView::viewportMouseReleaseEvent): Only send a click even if the node is the same one from
        the original click. Call invalidateClick when done to reduce the chance that we'll hold on to an
        old click node for a long time.
        (KHTMLView::keyPressEvent): Remove code that sets the originalNode field, which is never used.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7754 268f45cc-cd09-0410-ab3c-d52691b4dbfc

    Reviewed by Hyatt

        Fix for this bug:

        <rdar://problem/3825289> REGRESSION (Mail): Crash in fontForSelection in empty window

        * khtml/editing/visible_position.cpp:
        (khtml::VisiblePosition::previousPosition): Switch to node iteration instead "leaf"
        iteration. I have been wanting to make this change for a long time, but couldn't
        since other code relied on the leaf behavior. That is no longer true. Plus, the
        bug fix requires the new behavior.
        (khtml::VisiblePosition::nextPosition): Ditto.
        (khtml::VisiblePosition::isCandidate): Empty blocks needed a height to be a candidate,
        but we make a special case for the body element. This fixes the bug.

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@7753 268f45cc-cd09-0410-ab3c-d52691b4dbfc