Context Navigation

WebView.subproj/WebController.m: (-[WebController userAgentForURL:]): Find the suffix of the host name that contains exactly one dot, lower-case it, and look it up in the user-agent spoof table. For now, anything in the table pretends to be MacIE. Don't fret that this table is simple. We will complicate it as needed. The technique is flexible, even though it may not look it now.

WebView.subproj/WebControllerPrivate.h: Add userAgentWhenPretendingToBeMacIE field.
WebView.subproj/WebControllerPrivate.m: (-[WebControllerPrivate dealloc]): Release userAgentWhenPretendingToBeMacIE. (-[WebController _defaultsDidChange]): Release and nil userAgentWhenPretendingToBeMacIE.

Makefile.am: Added rules to build WebUserAgentSpoofTable.c using gperf.
WebView.subproj/WebUserAgentSpoofTable.c: Added. Generated file.
WebView.subproj/WebUserAgentSpoofTable.gperf: Added. Table

English.lproj/StringsNotToBeLocalized.txt: Update for above changes.

fixed a crash I saw in a simplistic way

WebView.subproj/WebHTMLViewPrivate.m: (-[WebHTMLView _elementAtPoint:]): Don't assert if the webFrame is nil, just return a partial dictionary.

9:00 PM Changeset in webkit [3072] by mjs

3 edits in trunk/WebCore

Tools:

Reviewed by Darin.

Scripts/set-module-version: Accept `u' in source version.

WebCore:

Reviewed by Darin.

WebCore.pbproj/project.pbxproj: Updated source version to 37u2.

8:14 PM Changeset in webkit [3071] by darin

3 edits in trunk/WebCore

Reviewed by Trey.

fixed regression where clicked-on buttons were getting keyboard focus

kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::khtmlMousePressEvent): Check clickCount here the way NSWindow does. See below for rationale. (KWQKHTMLPart::passWidgetMouseDownEventToWidget): I didn't read the code in [NSWindow sendEvent:] that I was stealing carefully enough. The old needsPanelToBecomeKey flag is used to decide whether something should really get first responder, and I need to do check that too. And while I was in here, did the same clickCount check that NSWindow does for good measure.

7:05 PM Changeset in webkit [3070] by mjs

7 edits in trunk/WebCore

Reviewed by Darin.

fixed 3128383 - DOM nodes should always use the same JS object

khtml/ecma/kjs_binding.h:
khtml/ecma/kjs_binding.cpp: (ScriptInterpreter::ScriptInterpreter): Set auto-delete to true on document dictionary. (ScriptInterpreter::getDOMObjectForDocument): New function - gets DOM object from per-document dictionary. (ScriptInterpreter::putDOMObjectForDocument): New function - put DOM object into per-document dictionary. (ScriptInterpreter::deleteDOMObjectsForDocument): Clear all objects from specific per-document dictionary. (ScriptInterpreter::mark): Mark per-document objects. (ScriptInterpreter::forgetDOMObjectsForDocument): Clear objects for this document from all interpreters.
khtml/ecma/kjs_dom.h:
khtml/ecma/kjs_dom.cpp: (KJS::getDOMNode): Use per-document rather than general DOM object dictionary.
khtml/xml/dom_docimpl.cpp: (DocumentImpl::~DocumentImpl): Call forgetDOMObjectsForDocument.

6:42 PM Changeset in webkit [3069] by hyatt

3 edits in trunk/WebCore

Fix for 3057974 and 3128713. I needed to make frames check to
see if their parentNode had a renderer to complete the FOUC
fix. Completing the FOUC fix finally solves 3057974.

Reviewed by darin

khtml/html/html_baseimpl.cpp: (HTMLFrameElementImpl::attach):

5:51 PM Changeset in webkit [3068] by hyatt

4 edits in trunk/WebCore

Fix for 3109226, epinions page is mostly blank. Add a hack
to disallow > but only inside attribute values for which there
is no corresponding attribute name.

Also fix attribute values with no name so that the value is
used as the name so that e.g., <option "selected"> works.

Reviewed by darin

khtml/html/htmltokenizer.cpp: (HTMLTokenizer::HTMLTokenizer):
khtml/html/htmltokenizer.h:

5:03 PM Changeset in webkit [3067] by darin

2 edits in trunk/WebKit

Reviewed by Trey.

fixed 3128260 -- REGRESSION: context menus in frames are always the generic page menu

WebView.subproj/WebHTMLViewPrivate.m: (-[WebHTMLView hitTest:]): Check the control key in the mouse down event. If it's down, then do no magic.

4:42 PM Changeset in webkit [3066] by darin

5 edits in trunk/WebCore

Reviewed by Dave.

fixed 3127900 -- REGRESSION: text fields select all their text when clicked on with the mouse

kwq/KWQKHTMLPart.h: Add _firstResponderAtMouseDownTime.
kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::khtmlMousePressEvent): Don't call makeFirstResponder: if it's already set. (KWQKHTMLPart::passWidgetMouseDownEventToWidget): Using _firstResponderAtMouseDownTime, catch the case where an NSTextField's editor was just created and became first responder. In those cases, pass the mouseDown: to the text field, not the newly created editor. Even though it shouldn't make a difference it does. Also don't makeFirstResponder: if it's already set. (KWQKHTMLPart::mouseDown): Set _firstResponderAtMouseDownTime.

kwq/KWQWidget.mm: (QWidget::setFocus): Don't call makeFirstResponder: if it's already set.

4:42 PM Changeset in webkit [3065] by darin

2 edits in trunk/WebKit

Reviewed by Dave.

fixed 3128651 -- REGRESSION: Mouse wheeling is busted on frames pages

WebView.subproj/WebHTMLViewPrivate.m: (-[WebHTMLView hitTest:]): Only do the hitTest magic for left mouse downs; the rest of the clicks end up going to the correct view anyway, without our help. This makes this stop making trouble for the scroll wheel events.

3:42 PM Changeset in webkit [3064] by hyatt

7 edits in trunk/WebCore

Make sure framesets don't get built twice when FOUC delays
their loading.

Reviewed by gramps

khtml/css/cssstyleselector.cpp:
khtml/css/cssstyleselector.h:
khtml/html/html_baseimpl.cpp: (HTMLFrameSetElementImpl::attach):
khtml/rendering/render_style.cpp: (RenderStyle::operator==): (RenderStyle::isStyleAvailable):
khtml/rendering/render_style.h:

3:02 PM Changeset in webkit [3063] by darin

5 edits in trunk/WebCore

Reviewed by Dave.

fixed ==/= typo in QFontMetrics that caused a part of my recent speedup to not take effect
fixed 3128582 -- REGRESSION: Crash on planetquake.com

kwq/KWQFontMetrics.mm: (QFontMetricsPrivate::setFont): Fix a horrifying "== instead of =" mistake. (QFontMetrics::ascent): Give an ERROR message, don't crash, when this is called on a font metrics object that hasn't been updated to work for a particular font. (QFontMetrics::descent): Ditto. (QFontMetrics::lineSpacing): Ditto. (QFontMetrics::xHeight): Ditto. (QFontMetrics::width): Ditto. (QFontMetrics::floatWidth): Ditto. (QFontMetrics::floatCharacterWidths): Ditto.

khtml/html/html_inlineimpl.cpp: (HTMLBRElementImpl::attach): Eliminated what was causing us to hit the above. BR elements need to respect display() == NONE.

fixed the fix for operator==; needs to work even if the pseudo type was extracted already for one of the two, and not the other

khtml/css/cssparser.cpp: (CSSSelector::extractPseudoType): Only do work when the selector is Pseudo. (CSSSelector::operator==): Call pseudoType() instead of going right at _pseudoType.

2:34 PM Changeset in webkit [3062] by hyatt

3 edits in trunk/WebCore

Fix for 3128402, missing content on voodooextreme. Don't allow
object/embed inside the head.

Reviewed by cblu

khtml/html/dtd.cpp:

12:18 PM Changeset in webkit [3061] by darin

6 edits in trunk/WebCore

Reviewed by Ken.

fixed 3127898 -- Crash going to altavista.com after visiting sony.com

khtml/ecma/kjs_window.cpp: (Window::get): Change lots of properties to return "undefined" rather than crashing when the view is NULL. I'm not sure that undefined is the perfect thing to return in this case, but it's better than crashing the program. We can refine later as needed.

fixed 3127157 -- REGRESSION: applet does not become first responder when clicked-on
fixed 3127934 -- Crash in KWQKHTMLPart::widgetWillReleaseView scrolling around

kwq/KWQKHTMLPart.h: Get rid of widgetWillReleaseView(). Add mouseDownViewIfStillGood().
kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::passWidgetMouseDownEventToWidget): Call makeFirstResponder: in the same way that the NSWindow does before delivering a mouseDown: message. This oversight caused the applet problem. (findViewInSubviews): Added. (KWQKHTMLPart::mouseDownViewIfStillGood): Added. Gets _mouseDownView, but then checks that it's still a subview before returning it. A little inefficient, but the only way to do it given that there's no general way to find out that an NSView is being removed from your view hierarchy or deallocated. This fixes the crash in widgetWillReleaseView. (KWQKHTMLPart::khtmlMouseMoveEvent): Use mouseDownViewIfStillGood() instead of _mouseDownView directly. (KWQKHTMLPart::khtmlMouseReleaseEvent): Ditto. (KWQKHTMLPart::passSubframeEventToSubframe): Ditto.
kwq/KWQWidget.mm: (QWidget::~QWidget): Remove call to widgetWillReleaseView(). (QWidget::setView): Ditto.

3:23 AM Changeset in webkit [3060] by hyatt

3 edits in trunk/WebCore

operator== on CSSSelectors needed to be updated to compare
_pseudoTypes. A nice subtle bug from the mega-patch.

Reviewed by mjs

khtml/css/cssparser.cpp: (CSSSelector::operator == ):

2:46 AM Changeset in webkit [3059] by hyatt

3 edits in trunk/WebCore

Fix for 3128429, sans-serif font got broken. Just a typo
from darin's mega-patch.

Reviewed by mjs

khtml/css/cssstyleselector.cpp:

1:15 AM Changeset in webkit [3058] by darin

34 edits
2 adds in trunk

JavaScriptCore:

Reviewed by Maciej.

JavaScriptCore.pbproj/project.pbxproj: Make dtoa.h visible as an SPI so I can use it inside QString.

WebFoundation:

Reviewed by Maciej.

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

WebCore:

Reviewed by Maciej and Dave.

optimizations that improve speed; total is 13.8% on cached cvs-base on my machine

khtml/css/css_valueimpl.h: Add isKonqBody() and genericFamilyType() to find out if a font is a standard one. This is much faster than doing string comparisons all the time.
khtml/css/css_valueimpl.cpp: (FontFamilyValueImpl::FontFamilyValueImpl): Check the font names against the standard ones on time when the FontFamilyValueImpl object is constructed rather than every time the stuff is matched.
khtml/css/cssparser.h: Add pseudoType() and extractPseudoType() to CSSSelector so we don't have to do string compares all the time.
khtml/css/cssparser.cpp: (StyleBaseImpl::parseValue): Use QConstString in a way that's effective (with an object that has a long enough lifetime), or don't use it at all. (StyleBaseImpl::parseContent): Ditto. (StyleBaseImpl::preprocess): Preprocess into a buffer, usually on the stack, rather than using QString += QChar over and over again. (CSSSelector::extractPseudoType): Convert the value string into a type, and get rid of it. Saves memory because the string goes away. And doing it once is much faster than string comparing each time.
khtml/css/cssstyleselector.cpp: (cleanPath): Restructure to save one find() in the common case. (checkPseudoState): Restructure to avoid some DOMString creation/destruction in the case where the tag is not an <A>. Also use QConstString correctly. (CSSStyleSelector::checkOneSelector): Use the new pseudoType() instead of string compares. (CSSStyleSelector::applyRule): Use isKonqBody() and genericFamilyType() instead of string comparisons.

khtml/dom/dom_string.h: Make the destructor non-virtual. This was a big win, and a mistake the way it was before. Also make the DOMString constructor with no parameters inline.
khtml/dom/dom_string.cpp: (DOMString::string): Don't use QConstString here, since it does no good. (DOM::operator==): Rewrite to do things faster and avoid calling strlen.

khtml/html/html_objectimpl.h: Don't use QConstString in places where it does harm and no good.
khtml/html/html_objectimpl.cpp: (HTMLEmbedElementImpl::parseAttribute): Use QConstString properly.

khtml/html/htmltokenizer.cpp: (tagMatch): Added. (HTMLTokenizer::parseSpecial): Change tokenizing so it doesn't construct temporary strings (with QConstString) just to compare small substrings.

khtml/khtml_part.cpp: (KHTMLPart::write): Don't use a decoder object when source is all ASCII.

kwq/KWQColor.mm: (hex2int): Don't bother with uppercase hex since we always lowercase anyway. (QColor::setNamedColor): Use a new gperf-based table instead of an NSDictionary to look up colors.
kwq/KWQColorData.c: Added this autogenerated file.
kwq/KWQColorData.gperf: Added. Source file with list of colors and color values.
kwq/Makefile.am: Build KWQColorData.c from KWQColorData.gperf.
WebCore.pbproj/project.pbxproj: Added KWQColorData.gperf.

khtml/misc/helper.cpp: (khtml::setNamedColor): Don't call setNamedColor again with the lowercased color name, since our QColor::setNamedColor already has to lowercase the name (because it has to check all names before checking any hex values).

khtml/rendering/font.h: Don't initialize the QFontMetrics since in every case we end up changing the font later anyway. Also remove unused field.
khtml/rendering/font.cpp: (Font::update): Use the new QFontMetrics::setFont for speed.

khtml/rendering/render_object.cpp: (RenderObject::enclosingLayer): Change it so it doesn't get the layer twice. This function showed up on the sample so we know it's hot.

khtml/xml/dom_stringimpl.cpp: (DOMStringImpl::containsOnlyWhitespace): Did a faster implementation that doesn't call QChar::direction() except for non-ASCII characters.

kwq/KWQFontMetrics.h: Add empty constructor for use in font.cpp. Add accessor and setter for the font too. Made baselineOffset() inline since it's trivial.
kwq/KWQFontMetrics.mm: (QFontMetricsPrivate::QFontMetricsPrivate): Don't make the renderer until we use it. (QFontMetricsPrivate::getRenderer): Make it here. (QFontMetricsPrivate::font): Added. (QFontMetricsPrivate::setFont): Added. Dumps the renderer if the font is different. (QFontMetrics::QFontMetrics): Added empty constructor. (QFontMetrics::setFont): Added.

kwq/KWQKURL.mm: (KURL::parse): Avoid function call overhead calling strncasecmp and strncmp. Also made the check for localhost case insensitive.

kwq/KWQString.h: Remove _isUnicodeInternal and _isAsciiInternal, since they are trivially computed, and it saves time to not have that extra field to manipulate. Removed private data() function since it's the same as *dataHandle and not really more clear. Made ascii() and unicode() simpler and inline. Also made latin1() and getCFString() inline. Renamed QStringData to KWQStringData.
kwq/KWQString.mm: (QStringData::QStringData): Remove initializers for the xxxInternal. (QStringData::initialize): Removed some unnecessary if statements, and made a slight improvement in the case of a 0-length string that comes with a pointer for Unicode (can't do it in the char * case because it's a feature that's used). (QString::makeSharedNull): No need to set _isUnicodeInternal. (QStringData::~QStringData): Use the new functions instead of the old bits. (QStringData::increaseAsciiSize): Grow faster when we get to larger sizes where the "good size" call doesn't have as much of an effect. Simplified by removing unneeded code. (QStringData::increaseUnicodeSize): Ditto. (QStringData::makeAscii): Update for removal of _isAsciiInternal. (QStringData::makeUnicode): Update for removal of _isUnicodeInternal. (QString::detachIfInternal): Add this new inline to speed things up for the common case where there's no detaching needed. (QString::at): Simplify, there were excess if branches here. (QString::toDouble): Use kjs_strtod and don't copy the string using QCString for additional speed. (QString::lower): Don't detach if the string is already all lowercase. Added a FIXME about the code that assumes all "ASCII" bytes are truly ASCII. This code blurs the distinction between Latin-1 and ASCII in a way that will not work right for Latin-1 characters in an 8-byte character QString. (QString::detachInternal): Update for removal of _isUnicodeInternal. (QString::detach): Removed a special case for shared_null that was dead code since shared_null has the _isUnicodeValid flag set. (operator==): Rewrote the comparison with char * to avoid the costly call to strlen.

Makefile.am: Updated the rules here so the force clean timestamp works right.
force-clean-timestamp: Touched this since the header dependencies don't seem right for the DOMString constructor change, at least.
WebCore-tests.exp: Exported a symbol now needed by tests since QString does more inlining.
WebCore-combined.exp: Re-generated.

WebBrowser:

Reviewed by Maciej.

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

1:04 AM Changeset in webkit [3057] by mjs

3 edits in trunk/WebCore

Reviewed by Darin.

slightly improved situation with 3124933 (abcnews.com leads to empty window with sheet complaining about javascript: URL); the popup window still comes up empty but it doesn't put up a sheet.

kwq/KWQKHTMLPartBrowserExtension.mm: (KHTMLPartBrowserExtension::openURLRequest): Handled "javascript:" URLs here.

12:24 AM Changeset in webkit [3056] by gramps

4 edits in trunk/WebKit

Reviewed by darin

Debug/DebugUtilities.m:

(-[DebugUtilities userAgentSubmenuItem]):

Changed "PPC" in one user agent string to "PPC Mac OS X" used for
debugging.

Fixed 3127173 -- REGRESSION: fboweb.com renders incorrectly

Reviewed by darin

English.lproj/StringsNotToBeLocalized.txt:
WebView.subproj/WebController.m:

(-[WebController userAgentForURL:]):

Changed "PPC" in our user agent string to "PPC Mac OS X" in order to
match Mozilla and make the silly server-side user agent string checking
work at fboweb.com. Actually, just "PPC " would work fine at
fboweb.com. Go figure.

Dec 14, 2002:

11:39 PM Changeset in webkit [3055] by hyatt

3 edits in trunk/WebCore

Fix insidemacgames.com. Reinstate rules I removed
and fix other rules to be !important so that the
td[align=right] rules won't affect child tables
with align=center.

Reviewed by gramps

khtml/css/html4.css:

8:43 PM Changeset in webkit [3054] by hyatt

12 edits in trunk/WebCore

Fix for 3016385, menus on webreference.com don't show up.

Fixes to make webreference.com menus show up. This involved
fixing layers to not clip positioned objects when overflow:hidden
is set, fixing the stupid clip/sync layout hack for livepage
so that it doesn't break webreference, and implementing
navigator.productSub (right now the date is set to
Christmas 2002).

Reviewed by gramps/maciej.

khtml/ecma/kjs_navigator.cpp: (Navigator::getValueProperty):
khtml/ecma/kjs_navigator.h:
khtml/ecma/kjs_navigator.lut.h:
khtml/rendering/render_box.cpp: (RenderBox::getOverflowClipRect):
khtml/rendering/render_box.h:
khtml/rendering/render_flow.cpp: (RenderFlow::layout):
khtml/rendering/render_layer.cpp: (RenderLayer::paint): (RenderLayer::nodeAtPoint): (RenderLayer::constructZTree):
khtml/rendering/render_layer.h:
khtml/rendering/render_object.cpp: (RenderObject::setLayouted):
khtml/rendering/render_object.h:

7:33 PM Changeset in webkit [3053] by mjs

3 edits in trunk/JavaScriptCore

Reviewed by Ken.

further corrections to number printing.

kjs/ustring.cpp: (UString::from): Make number printing match the ECMA standard algorithm.

5:48 PM Changeset in webkit [3052] by kocienda

7 edits in trunk/WebCore

Reviewed by Maciej

Fix for this bug:

Radar 3128159 (Netscape non-standard DOM event attribute "which" unsupported)

khtml/dom/dom2_events.cpp: (UIEvent::which): Added which() definition.
khtml/dom/dom2_events.h: Added declarations for which() functions in UIEvent.
khtml/ecma/kjs_events.cpp: (DOMUIEvent::getValueProperty): Added which property.
khtml/ecma/kjs_events.h: Ditto.
khtml/ecma/kjs_events.lut.h: File regenerated.

4:37 PM Changeset in webkit [3051] by mjs

3 edits in trunk/JavaScriptCore

Reviewed by Dave.

fix toString() conversion for numbers less than 1. Negative exponents are still wrong though (things like 1E-34).

kjs/ustring.cpp: (UString::from): Don't print empty string for numbers less than 1, and remember to add extra 0s after the decimal for negative decimal positions.

Dec 13, 2002:

7:49 PM Changeset in webkit [3050] by rjw

8 edits in trunk/WebCore

Fixed 3127310 and 3127920.

Implemented findWordBoundary. We now correctly detect word boundaries.
This function makes use of carbon unicode utilities.

With help from hyatt, made selection correctly account for render
continuations. This fixes many of the 'unable to select' issues.

Reviewed by trey and hyatt.

khtml/khtml_events.cpp: (khtml::MouseEvent::offset):
khtml/khtml_part.cpp: (findWordBoundary): (KHTMLPart::khtmlMousePressEvent): (KHTMLPart::khtmlMouseMoveEvent):
khtml/rendering/render_object.cpp: (RenderObject::checkSelectionPoint):
khtml/rendering/render_object.h:
khtml/rendering/render_text.cpp: (RenderText::checkSelectionPoint):
khtml/rendering/render_text.h:

7:06 PM Changeset in webkit [3049] by mjs

2 edits
1 add in trunk/WebCore

Reviewed by Don.

fixed build.

ForwardingHeaders/rendering/render_layer.h: Added.

6:42 PM Changeset in webkit [3048] by mjs

9 edits in trunk/WebCore

Reviewed by Don.

khtml/dom/dom2_events.cpp: (UIEvent::layerX): Added to support nonstandard Mozilla event.layerX property. (UIEvent::layerY): Added to support nonstandard Mozilla event.layerY property.
khtml/dom/dom2_events.h:
khtml/ecma/kjs_events.cpp: (DOMUIEvent::getValueProperty): Implemented layerX and layerY
khtml/ecma/kjs_events.h:
khtml/ecma/kjs_events.lut.h:
khtml/xml/dom2_eventsimpl.cpp: (MouseEventImpl::MouseEventImpl): Added call to computeLayerPos. (MouseEventImpl::computeLayerPos): Compute layer position. (MouseEventImpl::layerX): Added. (MouseEventImpl::layerY): Added. (MouseEventImpl::initMouseEvent): Added call to computeLayerPos.
khtml/xml/dom2_eventsimpl.h:

6:14 PM Changeset in webkit [3047] by hyatt

3 edits in trunk/WebCore

Fix for 314799, image at wired.com doesn't show up. Block
level replaced floating elements were being treated as though
they didn't float.

Reviewed by gramps

khtml/rendering/render_flow.cpp: (RenderFlow::layoutBlockChildren):

5:28 PM Changeset in webkit [3046] by darin

5 edits in trunk/WebKit

Reviewed by Don.

WebView.subproj/WebController.m: (-[WebController userAgentForURL:]): Add "(like Gecko)" string and change "WebKit" to "AppleWebKit".

English.lproj/StringsNotToBeLocalized.txt: Update for these changes.
Misc.subproj/WebUnicode.h: No need for & 0xFF since we cast to unsigned char.

5:02 PM Changeset in webkit [3045] by hyatt

5 edits in trunk/WebCore

Fix for 3098045.

Do not treat XHTML-style tags e.g., <br/> or <table/> as
self-closing. This is absolutely wrong, and it does not
match the behavior of other browsers. According to the HTML4
spec for parsing, / is an erroneous character and should simply
be dropped, so <table/> should be treated like <table>.

Fixes for crashes on glazman's blog and on tantek's blog. Both
still mis-render, but at least they don't crash.

Reviewed by gramps (and darin too)

khtml/html/htmltokenizer.cpp:
khtml/rendering/render_container.cpp:
khtml/rendering/render_style.cpp:

4:29 PM Changeset in webkit [3044] by sullivan

3 edits in trunk/WebKit

WebKit:

WebKit part of fix for 3028061 -- visiting a bookmarked site that now has a site icon will not update bookmark's icon

No longer store the icon in the WebHistoryItem, since there's no
mechanism for keeping it fresh and telling interested clients
when it changes. Instead, the latest icon is always returned from
the icon database, and it's up to clients to get a fresh one when
they notice that the icon for a URL has been updated.

Reviewed by Darin

History.subproj/WebHistoryItem.h: remove _icon and _loadedIcon ivars

History.subproj/WebHistoryItem.m: (-[WebHistoryItem dealloc]): don't release _icon (-[WebHistoryItem icon]): just return fresh icon from database, don't store (-[WebHistoryItem setURL:]): don't set _loadedIcon to NO

WebBrowser:

WebBrowser part of fix for 3028061 -- visiting a bookmarked site that now has a site icon will not update bookmark's icon

This first cut, which fixes the bug completely, just tosses out
the cached bookmark & history menus & tells the outline view to
redisplay everything. We might improve this later, but in practice
the only difference in behavior that this will cause is to make
the Bookmarks menu rebuild itself from scratch when clicked on
more often. (We can address that multiple ways, probably the best
one being to make rebuilding much faster by using IconRefs instead
of NSImages, see 3120678.)

Reviewed by Darin

BookmarksController.m: (-[BookmarksController awakeFromNib]): register for icon-changed notifications (-[BookmarksController _receivedIconChangedNotification:]): mark menu dirty when any icon has changed

BookmarksViewController.m: (-[BookmarksViewController awakeFromNib]): register for icon-changed notifications (-[BookmarksViewController iconChanged:]): reload outlineview data when any icon has changed. Note that this has no effect when the outlineview is not showing.

GlobalHistory.m: (-[GlobalHistory _registerForHistoryNotifications]): register for icon-changed notifications (-[GlobalHistory receivedIconChangedNotification:]): mark menu dirty when any icon has changed (-[GlobalHistory receivedEntriesChangedNotification:]): removed wayward space character

2:59 PM Changeset in webkit [3043]

3 copies in tags/Alexander-37u1

This commit was manufactured by cvs2svn to create tag
'Alexander-37u1'.

2:59 PM Changeset in webkit [3042] by sheridan

5 edits in trunk

changelog markers for B&I submission (I hope)

2:09 PM Changeset in webkit [3041] by darin

2 edits in trunk/WebKit

WebKit:

Reviewed by Maciej.

English.lproj/StringsNotToBeLocalized.txt: Update for recent changes.

WebBrowser:

Reviewed by Maciej.

fix some relatively new messages to use curly apostrophes

AppController.m: (-[AppController confirmEmptyCache:]): One in here.
LocationChangeError.m: (-[LocationChangeHandler specializedMessageForError:]): Two in here.
English.lproj/Localizable.strings: Generated this.

2:08 PM Changeset in webkit [3040] by trey

2 edits in trunk/WebKit

3108976 - assert _private provisionalItem in -[WebFrame(WebPrivate) _transitionToCommitted
3108865 - frames not maintained going back at directory.apple.com

Both bugs are fixed by the same small change. We no longer try to inherit loadType
across redirects. Instead we just make sure the right thing happens in _transitionToCommitted:
for redirects in the loadType=Standard case.

3122721 - History stores both original and redirected sites

Easy fix while in the neighborhood. We just don't add to History when doing a redirect.

Reviewed by rjw

WebView.subproj/WebFramePrivate.m: (-[WebFrame _transitionToCommitted:]): On redirect, update the URL of the frame's current item, not the current item in the b/f list. Also do not add to history on redirect. (-[WebFrame _loadURL:loadType:triggeringEvent:isFormSubmission:]): Cut out funny business that tried to guess when to inherit loadtype across redirects.

1:24 PM Changeset in webkit [3039] by hyatt

7 edits in trunk/WebCore

Two fixes. Bugs are 3121024, which had to do with handling
escaped characters in the CSS content property, and
bug 3127375 (44854 in bugs.kde.org), which had to do with
making sure table cells relayout properly when images load
async.

Reviewed by darin and john

khtml/css/cssparser.cpp: (StyleBaseImpl::parseContent): (StyleBaseImpl::isHexadecimal):
khtml/css/cssparser.h:
khtml/rendering/bidi.cpp: (RenderFlow::layoutInlineChildren):
khtml/rendering/render_flow.cpp: (RenderFlow::layout):
khtml/rendering/render_flow.h:

1:13 PM Changeset in webkit [3038] by darin

6 edits
4 adds in trunk/WebCore

Reviewed by Trey.

moved KWQFontFamily and KWQPageState to new homes

kwq/KWQFontFamily.h: Added.
kwq/KWQFontFamily.mm: Added.
kwq/KWQPageState.h: Added.
kwq/KWQPageState.mm: Added.
WebCore.pbproj/project.pbxproj: Added new source files.

kwq/KWQFont.h: Removed KWQFontFamily.
kwq/KWQFont.mm: Removed KWQFontFamily.
kwq/WebCoreBridge.mm: Removed KWQPageState.

11:24 AM Changeset in webkit [3037] by mjs

7 edits in trunk/WebCore

Reviewed by Ken.

fixed 3083814 - dhtml at www.bayarea.com doesn't hide on some mouse movements

khtml/ecma/kjs_events.h:
khtml/ecma/kjs_events.cpp: (DOMUIEvent::getValueProperty): Implemented pageX and pageY properties.
khtml/ecma/kjs_events.lut.h: Regenerated.
khtml/dom/dom2_events.h:
khtml/dom/dom2_events.cpp: (UIEvent::pageX): Get x coordinate if a mouse event. (UIEvent::pageY): Get y coordinate if a mouse event.

10:49 AM Changeset in webkit [3036] by rjw

12 edits in trunk

WebFoundation:

Fixed 3127225. Provide new SPI to get available system memory.
(Also currently unused SPI to get number of processors.)

Reviewed by gramps.

Misc.subproj/WebSystemBits.h:
Misc.subproj/WebSystemBits.m: (initCapabilities): (WebSystemMainMemory): (WebSystemProcessors):
WebFoundation.exp:
WebFoundation.pbproj/project.pbxproj:

WebCore:

Fixed 3127225. Set WebCore object cache size when first bridge is created.

Reviewed by gramps.

kwq/WebCoreBridge.h:
kwq/WebCoreBridge.mm: (-[WebCoreBridge init]):

WebKit:

Fixed 3127225. Scale page cache based on available memory.
Also added support for setting WebCore object cache size via
a preference.
Fixed 3126267. Increase CG glyph cache size if font smoothing
is turned on.

Reviewed by gramps.

History.subproj/WebBackForwardList.m: (+[WebBackForwardList pageCacheSize]):
Misc.subproj/WebKitLogging.h:
Misc.subproj/WebKitLogging.m:
WebCoreSupport.subproj/WebBridge.m: (-[WebBridge getObjectCacheSize]):
WebCoreSupport.subproj/WebTextRendererFactory.m: (getAppDefaultValue): (getUserDefaultValue): (getLCDScaleParameters):
WebView.subproj/WebPreferences.m: (+[WebPreferences load]): (-[WebPreferences _pageCacheSize]): (-[WebPreferences _objectCacheSize]):
WebView.subproj/WebPreferencesPrivate.h:

10:27 AM Changeset in webkit [3035] by cblu

2 edits in trunk/WebKit

Fixed: 3105486 - c|net news.com site seems to have lost its favicon again

Reviewed by darin.

Misc.subproj/WebIconDatabase.m: (-[WebIconDatabase _updateFileDatabase]): Only save icons with a size of 16 x 16 since that's the only size we use.

10:04 AM Changeset in webkit [3034] by cblu

2 edits in trunk/WebKit

Catch NULL status strings passed to NPN_Status.

Reviewed by darin.

Plugins.subproj/WebBaseNetscapePluginView.m: (-[WebBaseNetscapePluginView status:]): check the incoming string before calling [NSString stringWithCString]

10:03 AM Changeset in webkit [3033] by darin

18 edits in trunk/WebCore

Reviewed by John.

Overall speedup on cvs-base is about 5%.

reduce QString use in the tokenizer and tighten up QString implementation for a speedup
fix up the QFontFamily implementation, using new QString features too, for a speedup

khtml/html/htmltokenizer.h: Take const QString & instead of QString to avoid an extra copy.
khtml/html/htmltokenizer.cpp: (HTMLTokenizer::begin): Use QString::null instead of "" for QString objects; faster. (HTMLTokenizer::scriptHandler): In APPLE_CHANGES, use new prepend that avoids the need to create a temporary QString. Use QString::null instead of QString() for a null string, and also made other small fixes to avoid temporary QStrings. (HTMLTokenizer::setSrc): Take const QString & instead of QString to avoid an extra copy.

kwq/KWQFont.h: Rename QFontFamily to KWQFontFamily. Make family() inline, add inline familyIsEmpty(). Change KWQFontFamily to store both the QString for the family and the NSString instead of only the NSString to avoid constant conversions back and forth. Remove type cases in CREATE_FAMILY_ARRAY that could result in copying the font object.
kwq/KWQFont.mm: (KWQFontFamily::KWQFontFamily): Fix bug where a copied family object would have a random ref count instead of a ref count of 0. Also update for NSString change. (KWQFontFamily::operator=): Use the ref before deref algorithm instead of the check for self-assignment. (KWQFontFamily::getNSFamily): Compute the NSString for the family name here. To avoid making an NSString every time, use a CFDictionary with QString as a key and NSString as a value instead of using an NSSet. (KWQFontFamily::setFamily): Invalidate the NSString family when the QString family is changed. Moved the code to make things unique to getNSFamily. (KWQFontFamily::operator==): Call getNSFamily rather than

kwq/KWQString.h: Added a prepend and insert overload that take a QChar pointer and length, for use in the tokenizer, and a hash function, for use in CFDictionary. Also added a global set of CFDictionaryKeyCallBacks.
kwq/KWQString.mm: (QString::prepend): New overload calls new insert overload. (QString::insert): Old insert slightly tightened by calling detach() only after the early exit and removing an extra detach from a function that already does forceUnicode. Added the new insert that takes a QChar * and a length. (operator==): Rewrite this so that it is efficient and doesn't involve creating a QString temporary or doing a function call. (QString::hash): Wrote this. Used the same hash algorithm I did in my recent KJS work. (retainQString): Added. For use in CFDictionaryKey callback. (releaseQString): Added. For use in CFDictionaryKey callback. (describeQString): Added. For use in CFDictionaryKey callback. (equalQString): Added. For use in CFDictionaryKey callback. (hashQString): Added. For use in CFDictionaryKey callback.

khtml/rendering/font.h: Change name to KWQFontFamily.
khtml/rendering/font.cpp: (Font::update): Call familyIsEmpty() instead of family().isEmpty() to avoid making and destroying a QString every time. Also put this batch of KWQFontFamily-specific changes inside #if APPLE_CHANGES. Much to my dismay, we have others elsewhere that aren't ifdef'd.

khtml/css/cssstyleselector.cpp: (CSSStyleSelector::applyRule): Change name to KWQFontFamily. Clearly this code needs some #if APPLE_CHANGES, and changing the class name helps make that clear.

change KHTMLSettings so using it doesn't require copying QStrings for a speedup

kwq/KWQKHTMLSettings.h: Made all the functions static members to avoid the unnecessary overhead of passing a this pointer. Added lots of setXXX functions for use by the WebCoreSettings code that gets called by WebKit when settings are changed. Return objects by reference because we can and it avoids copying QString objects.
kwq/KWQKHTMLSettings.mm: Changed all the functions in here to return the same global object each time, and made the set functions update those globals.
kwq/WebCoreSettings.mm: All the setters now call through to KHTMLSettings set functions.

fixed 3126513 -- crash in KWQKHTMLPart::clearTimers() deallocating KWQPageState
did some cleanup in the bridge and the part; part of the process of moving real work out of the bridge

kwq/KWQKHTMLPart.h: Rearrange things to be a little more logical. Mark isFrameSet() const. added an attributedString function here so it's not inside WebCoreBridge. Added an overload of clearTimers that takes a view parameter.
kwq/KWQKHTMLPart.cpp: (KWQKHTMLPart::isFrameSet): Mark this const.
kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::clearTimers): Add the overload that takes the view parameter. (KWQKHTMLPart::attributedString): Moved this code in here from the bridge.

kwq/WebCoreBridge.mm: (-[WebCoreBridge selectedAttributedString]): Call the attributedString function in its new home in KWQKHTMLPart. (-[WebCoreBridge attributedStringFrom:startOffset:to:endOffset:]): Ditto. (-[KWQPageState initWithDocument:URL:windowProperties:locationProperties:]): Take a const KURL & instead of a KURL to avoid unnecessary copying. (-[KWQPageState dealloc]): Call the new clearTimers that takes a view parameter. This should fix the crash in autorelease pool deallocation mentioned above.

1:44 AM Changeset in webkit [3032] by hyatt

9 edits in trunk/WebCore

Fix for 3121865, content missing at disney.com. Make
sure blocks track width overflow as well as height
overflow and ensure that a layer's dimensions expand
to encompass both width and height overflow.

Also fixes a bug with line breaking that allowed unbreakable
words to actually be broken up when they were the only word
on a line. This is wrong and differs from all other
browsers, so the patch in bidi.cpp fixes us to match
other browsers. (This was required to fix the above bug.)

Reviewed by mjs.

khtml/rendering/bidi.cpp: (RenderFlow::findNextLineBreak):
khtml/rendering/render_flow.cpp: (RenderFlow::RenderFlow): (RenderFlow::layout): (RenderFlow::layoutBlockChildren):
khtml/rendering/render_flow.h:
khtml/rendering/render_layer.cpp: (RenderLayer::updateLayerPosition):
khtml/rendering/render_object.cpp: (RenderObject::nodeAtPoint):
khtml/rendering/render_object.h:
khtml/rendering/render_table.h:

Dec 12, 2002:

6:29 PM Changeset in webkit [3031] by trey

2 edits in trunk/WebKit

3117101 - PLT slows down as history fills up

I saw a 3% sloth effect from a huge (100k) history. This change cut that cost in half.
I think we still have a marginally measurable cost for our worst practical case.

Reviewed by rjw

History.subproj/WebHistoryPrivate.m: (-[WebHistoryPrivate insertEntry:atDateIndex:]): Comment. (-[WebHistoryPrivate removeEntryForURLString:]): Use removeIdentical instead of remove, since we don't need to do all the equals comparisons.

6:19 PM Changeset in webkit [3030] by rjw

17 edits in trunk

WebKit:

Reviewed by trey. This change was actually made by trey. It
make the "Empty Cache" menu release the page cache.

AppController.m: (-[AppController confirmEmptyCache:]):

WebCore:

Fixed 3116584. Made letter-spacing, word-spacing, and text-decoration play
well together.
Reviewed by hyatt.

Changes made with trey to enable to release of page cache from the
"Empty Cache" menu item in Alex.
Reviewed by trey.

khtml/rendering/font.cpp: (Font::drawLineForText): (Font::drawText):
khtml/rendering/font.h:
khtml/rendering/render_text.cpp: (TextSlave::paintDecoration): (RenderText::paintObject):
khtml/rendering/render_text.h:
kwq/KWQPainter.h:
kwq/KWQPainter.mm: (QPainter::drawLineForText):
kwq/WebCoreBridge.h:
kwq/WebCoreBridge.mm: (-[WebCoreBridge dealloc]): (-[WebCoreBridge openURL:reload:headers:lastModified:pageCache:]): (-[KWQPageState invalidate]): (-[KWQPageState dealloc]):
kwq/WebCoreTextRenderer.h:

WebKit:

Changes to fix 3116584.
Reviewed by hyatt.

Changes to support emptying the page cache from the
"Empty Cache" menu.
Reviewed by trey.

ChangeLog:
History.subproj/WebBackForwardList.h:
History.subproj/WebBackForwardList.m: (-[WebBackForwardList clearPageCache]):
History.subproj/WebHistoryItem.m: (+[WebHistoryItem _scheduleReleaseTimer]): (-[WebHistoryItem _scheduleRelease]): (+[WebHistoryItem _releasePageCache:]): (-[WebHistoryItem setHasPageCache:]):
WebCoreSupport.subproj/WebTextRenderer.m:
WebView.subproj/WebFramePrivate.m: (-[WebFrame _opened]):

Tests:

html/letterspacing.html:

4:12 PM Changeset in webkit [3029] by kocienda

2 edits in trunk/WebCore

Fixed up the reviewer line from my last commit.

4:09 PM Changeset in webkit [3028] by kocienda

8 edits in trunk/WebCore

Reviewed by NOBODY (OOPS!).

Final fix for this bug:

Radar 3062858 (REGRESSION: failed login at www.usaa.com doesn't return error text)

We were down to a case where submitting the login form with the return key
failed since we did not implement either non-standard keyCode (IE-style) or
which (Netscape-style) UI event extension for getting the value of a key
press. The form submission code depended on sensing the value of the
key press by checking one of these properties.

Since the "which" property also can be used for mouse events, and fixing
this bug could be done by implementing either keyCode or which, I chose
to implement keyCode only.

khtml/dom/dom2_events.cpp: Added keyCode function definition.
khtml/dom/dom2_events.h: Added keyCode function declaration.
khtml/ecma/kjs_events.cpp: (DOMUIEvent::getValueProperty): Added new KeyCode attribute.
khtml/ecma/kjs_events.h: New generated file.
khtml/ecma/kjs_events.lut.h: Ditto.
khtml/xml/dom2_eventsimpl.cpp: (KeyEventImpl::KeyEventImpl): Fixed a bug in khtml that prevented the m_keyVal field on KeyEventImpl from ever being set correctly. I guess I am the first person ever to care about that value. :-)

3:03 PM Changeset in webkit [3027] by sheridan

3 edits in trunk

version 37u ... tree is open

3:01 PM Changeset in webkit [3026]

3 copies in tags/Alexander-36

This commit was manufactured by cvs2svn to create tag 'Alexander-36'.

3:01 PM Changeset in webkit [3025] by sheridan

8 edits in trunk

Alex-36 versions

12:48 PM Changeset in webkit [3024] by darin

4 edits in trunk/WebCore

Reviewed by Ken.

fixed 3126042 -- Failure to send onkeypress event on key down causes forms to submit incorrectly

kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::keyEvent): Fixed the code to send the second event with repeat set equal to true. Before it was constructing the event but never sending it because result was true. Also fixed a problem where it was considering sending a second mouse up event. What we're doing here is a bit of a hack and probably should be fixed -- it seems that the behavior in dispatchKeyEvent where it sends only a down and not a press for the a non-autorepeat event is a KHTML bug that we should fix instead of work around eventually.

fix to something that made this harder to debug

khtml/xml/dom2_eventsimpl.cpp: (KeyEventImpl::KeyEventImpl): Logging the type was showing the wrong type, and I narrowed it down to this code here. Changing m_id after the fact doesn't change the type string, so you need to compute the correct ID and pass it to the constructor.

12:42 PM Changeset in webkit [3023] by rjw

2 edits in trunk/WebKit

i Change relating to 3083287. This doesn't fix the problem but

flips the geometry calcs to get most incremental images to draw
correctly. .mac now slideshows draw with the incorrect sliding
behavior while loading.

3083287 has been moved to 0.9/1.

Reviewed by darin.

WebCoreSupport.subproj/WebImageRenderer.m: (-[WebImageRenderer incrementalLoadWithBytes:length:complete:]): (-[WebImageRenderer drawClippedToValidInRect:fromRect:]):

11:46 AM Changeset in webkit [3022] by darin

3 edits in trunk/WebCore

Reviewed by Ken.

fixed 3125886 -- Failure to set document onkeypress handler causes return key to submit forms incorrectly

The handler was getting ignored because the property setting code never got to the DOMNode level.

khtml/ecma/kjs_dom.cpp: (DOMDocument::tryPut): Pass correct parent class, DOMNode, instead of DOMObject. By passing DOMObject, we skipped all the properties inherited from DOMNode, including onkeypress.

10:54 AM Changeset in webkit [3021] by darin

5 edits in trunk/WebCore

Reviewed by Trey and Ken.

fixed 3125613 -- REGRESSION: resizing frames after clicking on scroll bar doesn't work right

The problem was that the mouse dragged events were going to the scroll bar because the new click
on the frame resizer didn't fall into the code path that sets _mouseDownView.

kwq/KWQKHTMLPart.h: Add override of khtmlMouseMoveEvent.
kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::passWidgetMouseDownEventToWidget): Don't set _mouseDownView. This is now handled at the higher level when the mouseDown comes in. (KWQKHTMLPart::khtmlMouseMoveEvent): Added. Pass along mouseDragged: events. I didn't have this before because I don't want or need to pass mouseMoved: events along, but the KHTML mouse move event corresponds to both AppKit mouseDragged and mouseMoved. (KWQKHTMLPart::khtmlMouseReleaseEvent): Don't clear _mouseDownView. This is now handled at the higher level when the mouseUp comes in. (KWQKHTMLPart::passSubframeEventToSubframe): Ditto. (KWQKHTMLPart::mouseDown): Set _mouseDownView to nil; it will be set by passWidgetMouseDownEventToWidget if we end up passing the event down. This is the change that actually fixes the bug mentioned above. (KWQKHTMLPart::mouseUp): Set _mouseDownView to nil after handling the mouseUp. It's OK if we don't clear this out in cases where we don't get a mouseUp event -- we'll get a mouseDown later -- but it's still probably a good idea to clear it when we know the mouse is no longer down.

used update-kwq-assert to update the code that's a shared copy of WebFoundation code

kwq/KWQAssertions.m: (vprintf_stderr_objc): Got the UTF-8 fix I did a while back.

9:15 AM Changeset in webkit [3020] by trey

2 edits in trunk/WebKit

Fixed 3094525 - Need to use SPI to fix flipped drag image problem

We call the new SPI. I also had to rework the dissolve steps to get it
to non flip the image in Panther in millions (but leave the old code for
the Jaguar case).

Reviewed by cblu

Misc.subproj/WebNSImageExtras.m: (+[NSImage load]): Call the SPI. (-[NSImage _web_dissolveToFraction:]): Add new way of building the image for Panther.

2:56 AM Changeset in webkit [3019] by darin

7 edits in trunk/WebKit

WebFoundation:

Reviewed by Maciej.

fixed most likely cause of 3125565 -- 2% regression running the PLT for uncached loads

Misc.subproj/WebNSUserDefaultsExtras.m: (+[NSUserDefaults _web_defaultsDidChange]): Added. Clears out cache when defaults change. (+[NSUserDefaults _web_addDefaultsChangeObserver]): Added. Sets up the above to run at the right time. (+[NSUserDefaults _web_acceptLanguageHeaderForPreferredLanguages]): Changed to cache the result in a global and regenerate it when the global is cleared. (+[NSUserDefaults _web_preferredLanguageCode]): Ditto.

WebKit:

Reviewed by Maciej.

fixed most likely cause of 3125565 -- 2% regression running the PLT for uncached loads

Don't recompute the user agent when it doesn't change. It almost never changes.

WebView.subproj/WebController.m: (-[WebController initWithView:controllerSetName:]): Add observer so we know when defaults change. (-[WebController dealloc]): Remove observer. (-[WebController setApplicationNameForUserAgent:]): Clear out computed user agent to force it to be recomputed later. (-[WebController applicationNameForUserAgent]): Just retain since we copied when we stored it so we know it's not mutable. (-[WebController customUserAgent]): Ditto. (-[WebController userAgentForURL:]): Use the cached user agent if it's good. Otherwise compute and cache the user agent string. This means that we will almost never recompute it.

WebView.subproj/WebControllerPrivate.h: Add userAgent field to cache in.
WebView.subproj/WebControllerPrivate.m: (-[WebControllerPrivate dealloc]): Release userAgent. (-[WebController _defaultsDidChange]): Release and nil userAgent.

2:38 AM Changeset in webkit [3018] by mjs

8 edits in trunk

JavaScriptCore:

Reviewed by Darin.

fixed 3056449 - can't select state at tucows.com

kjs/array_instance.h:
kjs/array_object.cpp: (ArrayInstanceImp::propList): Add numeric proprties that are in special storage.
kjs/array_object.h:
kjs/object.h: Make propList a virtual method.

WebCore:

Reviewed by Darin.

force-js-clean-timestamp: Update for JavaScriptCore changes.

2:31 AM Changeset in webkit [3017] by darin

7 edits in trunk

WebCore:

Reviewed by Maciej.

fixed 3125504 -- REGRESSION: Selection not working correctly for text area on http://glish.com/css/7.asp

kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::khtmlMousePressEvent): Call makeFirstResponder here explicitly, since we are disabling the makeFirstResponder that normally happens during window event dispatching in WebKit.

kwq/KWQScrollView.h: Define getDocumentView().
kwq/KWQScrollView.mm: Use getDocumentView() everywhere. (QScrollView::getDocumentView): Define it here.

WebKit:

Reviewed by Maciej.

fixed 3125504 -- REGRESSION: Selection not working correctly for text area on http://glish.com/css/7.asp

The problem is that the text area lost its first responder status because WebHTMLView took the click,
and NSWindow wanted WebHTMLView to become first responder too.

WebView.subproj/WebHTMLView.m: (-[WebHTMLView acceptsFirstResponder]): Don't allow the NSWindow to make this the first responder during the early part of mouseDown event handling. But do allow anyone else to make this the first responder, for example from keyboard events, or from calls back from WebCore once we begin mouse-down event handling.

1:45 AM Changeset in webkit [3016] by darin

11 edits
1 delete in trunk

WebCore:

Reviewed by Maciej and Richard.

fixed 3125497 -- REGRESSION: Infinite recursion with mouseUp: events

The problem was that events could end up back at the same part due to the responder chain.

kwq/KWQKHTMLPart.h: Added mouseDown, mouseDragged, mouseUp, mouseMoved, buttonForCurrentEvent, and stateForCurrentEvent. Remove setCurrentEvent and currentEvent. Renamed handleMouseDownEventForWidget to passWidgetMouseDownEventToWidget. John and I had agreed to this name change before the last commit, but I forgot.

kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::KWQKHTMLPart): Initialize _sendingEventToSubview to false. No need to initialize _mouseDownWasInSubframe at all since it's only looked at when _mouseDownView is non-nil. (KWQKHTMLPart::keyEvent): Save and restore _currentEvent instead of setting to nil. Use stateForCurrentEvent instead of stateForEvent:. (KWQKHTMLPart::khtmlMousePressEvent): Name change. (KWQKHTMLPart::khtmlMouseDoubleClickEvent): Name change. (KWQKHTMLPart::passWidgetMouseDownEventToWidget): Name change. Set _sendingEventToSubview when sending event to subview. (KWQKHTMLPart::khtmlMouseReleaseEvent): Name change. Set _sendingEventToSubview when sending event to subview. (KWQKHTMLPart::passSubframeEventToSubframe): Name change. Set _sendingEventToSubview when sending event to subview. (KWQKHTMLPart::buttonForCurrentEvent): Added. Similar to code I moved here from WebCoreBridge. (KWQKHTMLPart::stateForCurrentEvent): Added. Similar to code I moved here from WebCoreBridge. (KWQKHTMLPart::mouseDown): Added. Similar to code I moved here from WebCoreBridge, except returns and does nothing if _sendingEventToSubview is set. (KWQKHTMLPart::mouseDragged): Ditto. (KWQKHTMLPart::mouseUp): Ditto. (KWQKHTMLPart::mouseMoved): Added. Similar to code I moved here from WebCoreBridge.

kwq/WebCoreBridge.mm: (-[WebCoreBridge mouseDown:]): Just call _part->mouseDown(). (-[WebCoreBridge mouseDragged:]): Just call _part->mouseDragged(). (-[WebCoreBridge mouseUp:]): Just call _part->mouseUp(). (-[WebCoreBridge mouseMoved:]): Just call _part->mouseMoved().

kwq/WebCoreBridgePrivate.h: Removed.
WebCore.pbproj/project.pbxproj: Removed WebCoreBridgePrivate.h.

WebKit:

Reviewed by Maciej and Richard.

fixed reentrancy crash I ran into while debugging infinite recursion bugs

WebView.subproj/WebHTMLViewPrivate.m: (-[WebHTMLView _updateMouseoverWithFakeEvent]): Added. Does part of what _frameOrBoundsChanged did. (-[WebHTMLView _frameOrBoundsChanged]): Schedule the mouseover update to happen soon, rather than doing it right away. If we do it right away, we might reenter because sending a mouse moved event can result in another layout since mouse moved events are the same as mouse dragged events in KHTML.
WebView.subproj/WebHTMLView.m: (-[WebHTMLView viewWillMoveToWindow:]): Cancel the scheduled mouseover update.

WebKit.pbproj/project.pbxproj: Electron uber alles.

1:20 AM Changeset in webkit [3015] by mjs

4 edits in trunk/JavaScriptCore/kjs

Revert stuff I committed by mistake.

1:16 AM Changeset in webkit [3014] by rjw

3 edits in trunk/WebCore

Fixed 3125583. Don't redraw view under resize bar
on mouseup.

Reviewed by gramps.

khtml/rendering/render_frames.cpp: (RenderFrameSet::userResize):

1:14 AM Changeset in webkit [3013] by mjs

6 edits in trunk/JavaScriptCore

Added missing reviewer.

12:48 AM Changeset in webkit [3012] by rjw

3 edits in trunk/WebKit

Fixed 3125585. One click crasher option clicking on
any link. Added additional check to curtail overly
zealous ASSERT.

Reviewed by mjs.

WebView.subproj/WebControllerPrivate.m: (-[WebController _mainReceivedError:fromDataSource:complete:]):

Dec 11, 2002:

10:20 PM Changeset in webkit [3011] by hyatt

4 edits in trunk/WebCore

Fix hang on monkeyvoodoo.net and on skramkoob.com. The bugs
are 3105714 and 3045965. Also fix margin collapsing churn
on thestar by not laying out a second time when collapsing
margins make you move unless you contain floats.

Reviewed by: gramps

khtml/rendering/bidi.cpp: (RenderFlow::findNextLineBreak):
khtml/rendering/render_flow.cpp: (RenderFlow::layoutBlockChildren);

8:02 PM Changeset in webkit [3010] by rjw

2 edits in trunk/WebKit

Fixed 3125425. Just call super if view isn't in view heirarchy,
rather than asserting. The assert was firing because NSText
was trying to perform a background layout on an item view that
was moved to the page cache.

Reviewed by gramps.

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

7:59 PM Changeset in webkit [3009] by mjs

7 edits in trunk/JavaScriptCore

Reviewed by Don.

Add kjsprint global function in Development build for ease of debugging.
Print uncaught JavaScript exceptions to the console in Development.
Improve wording of exception error messages.

kjs/function.cpp: (GlobalFuncImp::call):
kjs/function.h:
kjs/internal.cpp: (InterpreterImp::initGlobalObject):
kjs/interpreter.cpp: (Interpreter::evaluate):
kjs/nodes.cpp: (NewExprNode::evaluate): (FunctionCallNode::evaluate): (RelationalNode::evaluate):

7:06 PM Changeset in webkit [3008] by rjw

10 edits in trunk

WebKit:

Fixed 3124121, 3124716 (and other dupes). Regressions related to b/f crash.

Reviewed by hyatt.

ChangeLog:
History.subproj/WebHistoryItem.m: (-[WebHistoryItem _scheduleRelease]):
WebCoreSupport.subproj/WebBridge.m: (-[WebBridge saveDocumentToPageCache:]):
WebCoreSupport.subproj/WebImageRenderer.m: (-[WebImageRenderer incrementalLoadWithBytes:length:complete:]): (-[WebImageRenderer drawClippedToValidInRect:fromRect:]):
WebView.subproj/WebFramePrivate.m: (-[WebFrame _setState:]):

WebCore:

Fixed 3124121, 3124716 (and other dupes). Regressions related to b/f crash.

Reviewed by hyatt.

kwq/KWQKHTMLPart.h:
kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::openURLFromPageCache):
kwq/WebCoreBridge.h:
kwq/WebCoreBridge.mm: (-[WebCoreBridge dealloc]): (-[WebCoreBridge openURL:reload:headers:lastModified:pageCache:]):

5:15 PM Changeset in webkit [3007] by rjw

6 edits
1 add in trunk/WebKit

Fixed 3123375. Provide SPI to release all pending page caches

Reviewed by kocienda.

History.subproj/WebBackForwardList.m:
History.subproj/WebHistoryItem.h:
History.subproj/WebHistoryItem.m:
History.subproj/WebHistoryItemPrivate.h: Added.
WebCoreSupport.subproj/WebBridge.m:
WebKit.pbproj/project.pbxproj:
WebView.subproj/WebFramePrivate.m:

Fixed 3123375. Call SPI to release all pending page caches

Reviewed by kocienda.

Test/PageLoadTestController.m: (-[PageLoadTestController anyWindowWillClose:]):
Test/PageLoadTestRunner.m: (-[PageLoadTestRunner closeBrowserWindowsAndFinishCheckingForWorldLeaks]): (-[PageLoadTestRunner checkForWorldLeaksSoon]): (-[PageLoadTestRunner checkForWorldLeaksNow]): (-[PageLoadTestRunner startTest]):

2:37 PM Changeset in webkit [3006]

1 copy
58 deletes in tags/WebFoundation-37u2

This commit was manufactured by cvs2svn to create tag
'WebFoundation-37u2'.

2:37 PM Changeset in webkit [3005] by kocienda

1 edit in trunk/WebKit/Misc.subproj/WebFileDatabase.m

Reviewed by: Gramps

Change name of cache size file from ".size" to "size"

Database.subproj/WebFileDatabase.m:
English.lproj/StringsNotToBeLocalized.txt:

2:28 PM Changeset in webkit [3004] by darin

8 edits in trunk

WebFoundation:

Reviewed by Richard.

fixed 3124837 -- Crash trying to handle weird javascript URL in page address field

Misc.subproj/WebNSURLExtras.h: Added _web_scriptIfJavaScriptURL for both NSURL and NSString, and _web_stringByReplacingValidPercentEscapes.
Misc.subproj/WebNSURLExtras.m: (-[NSURL _web_scriptIfJavaScriptURL]): Added. Call the NSString version on our absoluteString. (isHexDigit): Added. (hexDigitValue): Added. (-[NSString _web_stringByReplacingValidPercentEscapes]): Added. Converts the sequences into character, and then re-encodes as UTF-8. Added a FIXME about doing even better. (-[NSString _web_scriptIfJavaScriptURL]): Check for a javascript: prefix, then turn the suffix into a script by replacing percent escapes.

English.lproj/StringsNotToBeLocalized.txt: Updated.

WebCore:

Reviewed by Richard.

fixed 3124263 -- REGRESSION: Crash running development build of PLT

kwq/KWQWidget.mm: (QWidget::~QWidget): Don't worry about the outer view here, since we can never hit one of those. Just pass the view to KWQKHTMLPart::widgetWillReleaseView. (QWidget::getOuterView): Re-add the assertion I had removed.

fixed 3124837 -- Crash trying to handle weird javascript URL in page address field

kwq/KWQKURL.mm: (KURL::decode_string): Made this function return the URL still encoded rather than a null string (actually, crashing) when CFURLCreateStringByReplacingPercentEscapes rejects the passed-in string. Also updated the FIXME. (appendEscapingBadChars): Fixed problem with high-bit-set characters.
kwq/KWQString.mm: (QString::setBufferFromCFString): Made this tolerant of NULL.

WebKit:

Reviewed by Richard.

fixed 3124837 -- Crash trying to handle weird javascript URL in page address field

Plugins.subproj/WebBaseNetscapePluginView.m: (-[WebBaseNetscapePluginView loadRequest:inTarget:withNotifyData:]): Use the new _web_scriptIfJavaScriptURL to simplify the check for JavaScript URLs, and to use a more tolerant unescaper.

English.lproj/StringsNotToBeLocalized.txt: Updated.

WebBrowser:

Reviewed by Richard.

fixed 3124837 -- Crash trying to handle weird javascript URL in page address field

BrowserWebBookmarkExtras.m: (-[WebBookmark _goToWithWindowPolicy:]):
BrowserWindowController.m: (-[BrowserWindowController goToToolbarLocation:]): Use the new _web_scriptIfJavaScriptURL to simplify the check for JavaScript URLs, and to use a more tolerant unescaper.

BrowserWebController.m: (-[BrowserWebController mouseoverTextForElement:modifierFlags:]): Use _web_stringByReplacingValidPercentEscapes, since it is more tolerant than CFURLCreateStringByReplacingPercentEscapes.

English.lproj/StringsNotToBeLocalized.txt: Updated.

12:26 PM Changeset in webkit [3003] by darin

15 edits in trunk

WebCore:

Reviewed by John.

fixed handling of mouse events so I could turn the mechanism on all the way
fixed anomalies in the red-painting-for-null-view-detection
fixed 3123105 -- REGRESSION: Crash in KHTMLView::unscheduleRelayout on exit
fixed 3124047 -- REGRESSION: crash in KWQKHTMLPart::handleMouseDownEventForWidget at nwa.com

kwq/KWQKHTMLPart.h: Add setCurrentEvent()/currentEvent()/_currentEvent back, because we can end up doing a lot of work and processing a lot of events before we get around to handling the current event, so [NSApp currentEvent] won't cut it. Also add clearTimers, passSubframeEventToSubframe, an overload for handleMouseDownEventForWidget, and _mouseDownWasInSubframe boolean.

kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::submitForm): Use _currentEvent. (KWQKHTMLPart::urlSelected): Use _currentEvent. (KWQKHTMLPart::paint): Fixed red-drawing-for-debug code by using KWQPainter instead of going directly to the AppKit level. (KWQKHTMLPart::keyEvent): Moved the check for null node up a bit, and set _currentEvent. (KWQKHTMLPart::handleMouseDownEventForWidget): Break into two pieces, so we can share the code with the new passSubframeEventToSubframe. Set _mouseDownWasInSubframe to false when setting _mouseDownView. Use getView() instead of getOuterView() since we really don't care about the WebView after all. Use _currentEvent. Add a check for nodes that lack a renderer to fix the crash at nwa.com above. (KWQKHTMLPart::khtmlMouseReleaseEvent): Use _currentEvent. (KWQKHTMLPart::clearTimers): Added. Checks for null view, which fixes the crash on exit above. (KWQKHTMLPart::passSubframeEventToSubframe): Added. Checks for the categories of events that we can pass to subframes, which are mouse down, up, and dragged events, and passes them along. This is the main fix that makes subframes work properly.

khtml/khtmlview.cpp: (KHTMLView::viewportMousePressEvent): Add call to the new passSubframeEventToSubframe. (KHTMLView::viewportMouseDoubleClickEvent): Ditto. (KHTMLView::viewportMouseMoveEvent): Ditto. (KHTMLView::viewportMouseReleaseEvent): Ditto.

kwq/WebCoreBridge.mm: (-[WebCoreBridge saveDocumentToPageCache]): Call clearTimers on the part now, instead of the document. (-[WebCoreBridge mouseUp:]): Call setCurrentEvent. (-[WebCoreBridge mouseDown:]): Ditto. (-[WebCoreBridge mouseMoved:]): Ditto. (-[WebCoreBridge mouseDragged:]): Ditto. (-[KWQPageState initWithDocument:URL:windowProperties:locationProperties:]): Call clearTimers on the part now, instead of the document.

other changes

khtml/ecma/kjs_window.lut.h: Let this be regenerated. I think I accidentally checked in an obsolete version of this earlier.

khtml/html/html_documentimpl.h: Remove clearTimers(). We don't need it here in APPLE_CHANGES.
khtml/html/html_documentimpl.cpp: Ditto.

khtml/khtmlpart_p.h: Removed unneeded APPLE_CHANGES. We don't use m_frameNameId any more, nor do we need m_savingPageState or m_restoringPageState.
khtml/khtml_part.cpp: (KHTMLPart::clear): Removed unneeded APPLE_CHANGES, now that we don't use m_frameNameId any more.

kwq/KWQWidget.mm: (KWQWidget::getOuterView): Tolerate nil for view here, so we can use this from the QWidget destructor.

WebKit:

Reviewed by John.

turned on the mechanism that passes events through WebCore, now that it's working better

WebView.subproj/WebHTMLViewPrivate.m: (-[WebHTMLView hitTest:]): Moved this function in here. Rewrote it. It has a new feature where you can set a global and do a normal hit test, needed for the mouse-moved handling below. Also, for efficiency, it does the actual hit testing of itself, rather than calling super, which recurses, and ignoring the result. The sum total is that it always returns self for mouse-moved events, so we pass them all over the bridge at the topmost frame.

(-[WebHTMLView _updateMouseoverWithEvent:]):

WebView.subproj/WebHTMLView.m: Moved hitTest: out of here.

9:43 AM Changeset in webkit [3002] by sullivan

2 edits in trunk/WebKit

WebKit:

fixed 3124640 -- Crash importing IE Favorites if there are no IE favorites

Reviewed by Darin

Bookmarks.subproj/WebBookmarkImporter.m: (-[WebBookmarkImporter initWithPath:]): use alloc/init instead of autoreleasing constructor for error since it's kept around until dealloc.

WebBrowser:

fixed 3123763 -- lose context when organizing bookmarks
handled empty or non-existent IE Favorites case better

Reviewed by Darin

BookmarksController.m: (-[BookmarksController importInternetExplorerFavorites]): Don't add a folder for IE Favorites if there are none.

BookmarksViewController.m: (-[BookmarksViewController selectBookmarkSource:]): new method to select an item in Collections column (-[BookmarksViewController revealFavorite:expandChildren:]): call selectBookmarkSource (-[BookmarksViewController tableView:draggedImage:endedAt:operation:]): save and restore selected item in Collections column since row indices might change (-[BookmarksViewController tableView:acceptDrop:row:operation:]): save and restore selected item in Collections column since row indices might change

9:11 AM Changeset in webkit [3001] by darin

3 edits in trunk/WebKit

Reviewed by John.

fixed 3117611 -- REGRESSION: exception in mouseoverTextForElement with accented characters in status

WebView.subproj/WebHTMLViewPrivate.m: (-[NSMutableDictionary _web_setObjectIfNotNil:forKey:]): Remove the object from the dictionary rather than just leaving the dictionary alone if it's nil. This is needed since we now are reusing an already-existing dictionary.

English.lproj/StringsNotToBeLocalized.txt: Update for recent changes.

8:42 AM Changeset in webkit [3000] by cblu

5 edits in trunk/WebKit

Fixed: 3118430 - crash / loop trying to copy url of link to clipboard
Fixed: 3122585 - REGRESSION: dragging links to the desktop or a Finder window does not do anything

Reviewed by john.

We were reusing the drag types from the previous drag pasteboard. For example, we would declare image types when dragging URLs. This would confuse the Finder, so location files weren't being created. This would occasional cause us to crash because we weren't providing the declared data.

Misc.subproj/WebNSPasteboardExtras.h:
Misc.subproj/WebNSPasteboardExtras.m: (+[NSPasteboard _web_dragTypesForURL]): Added more types that we handle (-[NSPasteboard _web_writeURL:andTitle:withOwner:declareTypes:]): declares the provided types, writes URL and title (-[NSPasteboard _web_writeURL:andTitle:withOwner:]): call _web_writeURL:andTitle:withOwner:types: with _web_dragTypesForURL
Misc.subproj/WebNSViewExtras.h:
Misc.subproj/WebNSViewExtras.m: (-[NSView _web_dragPromisedImage:origin:URL:fileType:title:event:]): call _web_writeURL:andTitle:withOwner: types: plus images types

8:23 AM Changeset in webkit [2999] by darin

11 edits in trunk/WebCore

Rolled out changes I accidentally checked in last night.

12:52 AM Changeset in webkit [2998] by trey

11 edits in trunk/WebKit

WebKit:

3092966 - going back goes to different page (can't go back to a POST page)

We will rePOST data upon back/forward/refresh if our caches fail us. The policy delegate
gets a crack at confirming this operation. Latent bugs where the policy delegate was double
queried are fixed. A bug in the b/f cursor when a page failed to load is fixed.

Reviewed by Maciej.

English.lproj/StringsNotToBeLocalized.txt: Usual suspects.
History.subproj/WebHistoryItem.h: Add state for reposting forms.
History.subproj/WebHistoryItem.m: Boilerplate changes for new state. (-[WebHistoryItem dealloc]): (-[WebHistoryItem setFormData:]): (-[WebHistoryItem setFormContentType:]): (-[WebHistoryItem formData]): (-[WebHistoryItem formContentType]): (-[WebHistoryItem description]):
WebView.subproj/WebController.m: (-[WebController _goToItem:withLoadType:]): Tighten up an assert as I clarified an assumption as I worked through this task.
WebView.subproj/WebControllerPolicyDelegate.h: New WebNavigationType's for back/forward, refresh, and repost.
WebView.subproj/WebFrame.m: (-[WebFrame loadRequest:]): Eliminate extra copy of a request. (-[WebFrame reload]): Eliminate extra copy of a request. Setup triggeringAction properly if we're about to rePOST.
WebView.subproj/WebFramePrivate.h: Started moving some private methods into the .m file.
WebView.subproj/WebFramePrivate.m: (-[WebFrame _createItem]): Save form state with history items. (-[WebFrame _isLoadComplete]): Fix up b/f cursor on page error before commit succeeds. (-[WebFrame _loadItem:fromItem:withLoadType:]): Set up request to rePOST if that's what the HistoryItem demands. Add call to _addExtraFieldsToRequest so we don't do a double query of the policy delegate. Pre-flight the form post vs. WF cache to setup triggering action properly. (-[WebFrame _actionInformationForLoadType:isFormSubmission:event:originalURL:]): New utility method to help build action dict. (-[WebFrame _continueAfterNavigationPolicy:]): Comment only. (-[WebFrame _loadURL:loadType:triggeringEvent:isFormSubmission:]): Call new utility method instead (just code factoring). (-[WebFrame _postWithURL:data:contentType:triggeringEvent:]): Call new utility method instead (just code factoring). Add call to _addExtraFieldsToRequest so we don't do a double query of the policy delegate. (-[WebFrame _reloadAllowingStaleDataWithOverrideEncoding:]): Whitespace only. (-[WebFrame _resetBackForwardListToCurrent]): Utility routine to fix up b/f cursor on page error. (-[WebFrame _continueLoadRequestAfterNavigationPolicy:request:]): Fix up b/f cursor on page error before commit succeeds.

WebBrowser:

3092966 - going back goes to different page (can't go back to a POST page)

The app now puts up a confirmation sheet when the user tries to return to a
page resulting from a POST, if all of our caches failed.

Reviewed by Maciej.

BrowserWebController.m: (-[BrowserWebController confirmSendSheetDidEnd:returnCode:contextInfo:]): Changed name of existing routine to be more generic, as its used for two sheets. (-[BrowserWebController decideNavigationPolicyForAction:andRequest:inFrame:decisionListener:]): Put up the sheet if we're about to rePOST.
English.lproj/Localizable.strings: The usual.
English.lproj/StringsNotToBeLocalized.txt: The usual.

12:13 AM Changeset in webkit [2997] by darin

2 edits in trunk/WebKit

Reviewed by Maciej.

fixed 3124302 -- REGRESSION: Can't use directory.apple.com because frame resize bar intercepts mouse clicks

WebView.subproj/WebHTMLView.m: (-[WebHTMLView hitTest:]): Don't steal clicks for views that are inside nested WebViews. This isn't the real fix, but it's good enough for now. The real fix will be in WebCore.

Note: See TracTimeline for information about the timeline view.

Download in other formats:

RSS Feed