Timeline

Oct 4, 2004:

4:20 PM Changeset in webkit [7758] by cblu

• 2 edits in trunk/WebKit

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

3:54 PM Changeset in webkit [7757] by darin

• 3 edits in trunk/WebCore

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.

3:45 PM Changeset in webkit [7756] by darin

• 2 edits in trunk/WebCore

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.

3:14 PM Changeset in webkit [7755] by cblu

• 13 edits in trunk/WebKit

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:

2:32 PM Changeset in webkit [7754] by darin

• 2 edits in trunk/WebCore

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.

11:52 AM Changeset in webkit [7753] by kocienda

• 2 edits in trunk/WebCore

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.

11:30 AM Changeset in webkit [7752]

• 25 copies
  3 deletes in tags/WebKit-125~5~3

This commit was manufactured by cvs2svn to create tag
'WebKit-125~5~3'.

11:30 AM Changeset in webkit [7751] by vicki

• 2 edits in branches/Safari-1-2-branch/WebKit

• versioning for SUPanNavy, WebKit-125.5.3

11:01 AM Changeset in webkit [7750] by cblu

• 2 edits in trunk/WebKit

Fixed: <rdar://problem/3758113> REGRESSION: Macromedia ColdFusion page doesn't show main content

After bumping up our plug-in version, Flash now sends 2 CRLF's between the headers and body of their POST request. Our code was not prepared for this.

Reviewed by darin.

• Plugins.subproj/WebBaseNetscapePluginView.m: (-[NSData _web_locationAfterFirstBlankLine]): looks for 2 CRLF's, not for 2 LF's

10:43 AM Changeset in webkit [7749] by darin

• 3 edits in trunk/WebKit

Reviewed by Ken.

• fixed a potential storage leak when we turn on CGImageRef image rendering

• WebCoreSupport.subproj/WebImageRenderer.m: (-[WebImageRenderer dealloc]): Fix potential storage leak by adding [super dealloc], but leak was not real yet because the code is commented out.

• make paste style work with color as part of fix to <rdar://problem/3814237> REGRESSION (Mail): Copy/paste style does not set color in Mail compose window

• WebView.subproj/WebHTMLView.m: (-[WebHTMLView _selectionFontAttributes]): Change structure so it's easy to add more attributes. For now I haven't added any yet. (-[WebHTMLView _colorAsString:]): Moved this earlier in the file. (-[WebHTMLView _shadowAsString:]): Ditto. (-[WebHTMLView _styleFromFontAttributes:]): Add background color, foreground color, and text shadow.

10:40 AM Changeset in webkit [7748] by darin

• 5 edits
  1 delete in trunk/WebCore

Reviewed by Ken.

• fixed <rdar://problem/3800667> REGRESSION (Mail): double-clicking multiple spaces only selects two spaces

• kwq/KWQTextUtilities.mm: (KWQFindWordBoundary): Moved here from the .cpp file. Changed to use the doubleClickAtIndex: method from NSAttributedString rather than using Unicode Utilities.
• kwq/KWQTextUtilities.cpp: Removed.
• WebCore.pbproj/project.pbxproj: Removed KWQTextUtilities.cpp.

• fixed a problem that would show up using HTML editing under garbage collection

• kwq/KWQEditCommand.mm: (-[KWQEditCommand finalize]): Fixed a [super dealloc] that should have been a [super finalize].

• another small change

• khtml/editing/visible_units.cpp: (khtml::nextWordBoundary): Tweaked a comment.

10:34 AM Changeset in webkit [7747] by darin

• 3 edits in trunk/JavaScriptCore

Reviewed by Ken.

• rolled in a fix the KDE folks did for the operations that generate HTML fragments

• kjs/string_object.cpp: (StringProtoFuncImp::call): Added quote marks to generated HTML.

• rolled out an old workaround we don't need any more

• JavaScriptCore.pbproj/project.pbxproj: Remove -Wno-long-double because the <math.h> issue that required it is no longer there.

Oct 1, 2004:

10:25 AM Changeset in webkit [7746] by darin

• 3 edits in trunk/WebCore

Reviewed by Ken (or arguably done by Ken, reviewed by Darin).

• fixed <rdar://problem/3823828> REGRESSION (Mail): Clicking past end of any line puts insertion point at beginning of next line

• khtml/editing/visible_position.cpp: (khtml::VisiblePosition::VisiblePosition): Check for an offset of 0 and a <br> and use UPSTREAM affinity in that case. This is a short term fix for something that needs a better longer-term fix.

• fixed <rdar://problem/3823816> REGRESSION (Mail): double-clicking first word on line also selects previous empty line

• khtml/editing/visible_units.cpp: (khtml::previousWordBoundary): Added a special case for <br>. While I'm not sure why this regressed, I'm sure this fix is good.

8:18 AM Changeset in webkit [7745] by darin

• 2 edits in trunk/WebCore

Reviewed by John.

• fixed <rdar://problem/3782117> CrashTracer: ..722 crashes at com.apple.WebCore: DOM::HTMLBodyElementImpl::insertedIntoDocument + 0x2c (AOL website)

• khtml/html/html_baseimpl.cpp: (HTMLBodyElementImpl::insertedIntoDocument): Check for nil document case. (HTMLFrameElementImpl::isURLAllowed): Ditto. (HTMLFrameElementImpl::openURL): Ditto.

1:02 AM Changeset in webkit [7744] by darin

• 3 edits in trunk/WebCore

Reviewed by Maciej.

• fixed <rdar://problem/3822218> REGRESSION (164-165): images don't update on rollover on directv.com page

I introduced a major regression where various JavaScript window properties would not be found when I
fixed bug 3809600.

• khtml/ecma/kjs_window.h: Added hasProperty.
• khtml/ecma/kjs_window.cpp: (Window::hasProperty): Return true in all the cases where get returns something.

Sep 30, 2004:

11:43 PM Changeset in webkit [7743] by darin

• 5 edits in trunk/WebCore

Reviewed by Maciej.

• fixed <rdar://problem/3461499> JavaScript function document.open() is buggy with 2nd argument "replace"

Experiments with Firefox indicate that document.open() should not be treated as window.open() unless there
are more than two parameters. Also, Firefox does not implement the "replace" behavior, so we don't have
to worry about it either.

• khtml/ecma/kjs_html.cpp: (KJS::HTMLDocFunction::tryCall): Only forward to window if there are more than two parameters, rather than if there are more than one.

• fixed <rdar://problem/3672933> oninput is firing at page load time for <input type=range>

• kwq/KWQSlider.mm: (QSlider::setValue): Don't emit a signal here. This follows the usual pattern, where a signal is only emitted for changes that are not explicitly requested by the caller.

• fixed <rdar://problem/3821167> leaks of something allocated by recalcStyle after loading altavista page from cvs-base

• khtml/xml/dom_docimpl.cpp: (DocumentImpl::recalcStyle): Ref and deref the style we allocate so it's not left floating if setStyle decides not to ref it.

• fixed <rdar://problem/3821172> leak of HTML attribute string after loading the ESPN page from cvs-base

• khtml/html/html_elementimpl.cpp: (HTMLElementImpl::parseHTMLAttribute): Use the DOMString version of the lower() operation so we don't end up leaving a DOMStringImpl * floating if AtomicString decides not to ref it.

5:24 PM Changeset in webkit [7742] by rjw

• 15 edits in trunk

WebKit:

Fixed <rdar://problem/3821215> NPN hasMethod and hasProperty functions should take NPObjects, not NPClass

Also changed dashboard regions dictionary to use "control"
for scroller region label, instead of "scroller, per
request from ouch.

Reviewed by Chris.

• Plugins.subproj/npruntime.h:
• WebView.subproj/WebView.m: (-[WebView _addScrollerDashboardRegions:from:]):

JavaScriptCore:

Fixed <rdar://problem/3821215> NPN hasMethod and hasProperty functions should take NPObjects, not NPClass

Reviewed by Chris.

• bindings/NP_jsobject.cpp: (_NPN_GetProperty): (_NPN_HasProperty): (_NPN_HasMethod):
• bindings/c/c_class.cpp: (CClass::methodsNamed): (CClass::fieldNamed):
• bindings/c/c_class.h:
• bindings/c/c_instance.cpp: (CInstance::invokeMethod):
• bindings/jni/jni_class.cpp: (JavaClass::methodsNamed):
• bindings/jni/jni_class.h:
• bindings/npruntime.h:
• bindings/objc/objc_class.h:
• bindings/objc/objc_class.mm: (ObjcClass::methodsNamed):
• bindings/runtime.h:
• bindings/runtime_object.cpp: (RuntimeObjectImp::get): (RuntimeObjectImp::hasProperty):

5:18 PM Changeset in webkit [7741] by cblu

• 2 edits in trunk/WebKit

Fixed: <rdar://problem/3498680> switching back and forth between tabs stops calling anything in a plug-in

Reviewed by darin.

• Plugins.subproj/WebBaseNetscapePluginView.m: (-[WebBaseNetscapePluginView sendEvent:]): call setWindowIfNecessary because the window may have changed (-[WebBaseNetscapePluginView updateAndSetWindow]): new (-[WebBaseNetscapePluginView setWindowIfNecessary]): was setWindow, this method now just sets the window (-[WebBaseNetscapePluginView start]): call updateAndSetWindow (-[WebBaseNetscapePluginView viewDidMoveToWindow]): ditto (-[WebBaseNetscapePluginView viewHasMoved:]): ditto

4:14 PM Changeset in webkit [7740] by cblu

• 3 edits in trunk/WebKit

Fixed: <rdar://problem/3498668> switching out of tab doesn't send loseFocusEvent to plug-in

Reviewed by rjw.

• Plugins.subproj/WebBaseNetscapePluginView.h:
• Plugins.subproj/WebBaseNetscapePluginView.m: (-[WebBaseNetscapePluginView setHasFocus:]): new, sends events to plug-in (-[WebBaseNetscapePluginView becomeFirstResponder]): call setHasFocus (-[WebBaseNetscapePluginView resignFirstResponder]): ditto (-[WebBaseNetscapePluginView viewWillMoveToWindow:]): ditto

1:04 PM Changeset in webkit [7739] by rjw

• 2 edits in trunk/WebCore

Fixed <rdar://problem/3822330> REGRESSION: crash on launch when homepage is set to about:blank

Added nil check.

• kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::setDisplaysWithFocusAttributes):

12:48 PM Changeset in webkit [7738] by cblu

• 2 edits in trunk/WebCore

Fixed: <rdar://problem/3792822> Safari is calling the Cocoa QuickTime plugin twice for the OBJECT and EMBED tags

Reviewed by hyatt.

• khtml/html/html_objectimpl.cpp: (HTMLObjectElementImpl::attach): call dispatchHTMLEvent after updateWidget instead of every time this method is called (HTMLObjectElementImpl::recalcStyle): ditto

11:56 AM Changeset in webkit [7737] by darin

• 2 edits in trunk/WebCore

• rolled out bad image change that caused performance regression

• khtml/rendering/render_image.cpp: (RenderImage::setPixmap): Don't reference the new image before doing the assignment. This forced an unwanted.

10:52 AM Changeset in webkit [7736] by cblu

• 6 edits in trunk/WebKit

Fixed:
Assertion failure when loading standalone netscape plug-in content.
Document loads of WebKit plug-in content should be cancelled since the plug-in does its own loading.

Reviewed by john.

• Misc.subproj/WebKitErrors.m: removed deprecated method
• Misc.subproj/WebKitErrorsPrivate.h:
• Plugins.subproj/WebNetscapePluginDocumentView.m: (-[WebNetscapePluginDocumentView setDataSource:]): fixed the assertion statement
• Plugins.subproj/WebPluginDocumentView.h:
• Plugins.subproj/WebPluginDocumentView.m: (-[WebPluginDocumentView dealloc]): remove retained plug-in (-[WebPluginDocumentView setDataSource:]): retain the plug-in, cancel the laod

9:59 AM Changeset in webkit [7735] by kocienda

• 3 edits in trunk/WebCore

Reviewed by me, coded by Darin

• fixed <rdar://problem/3818305> REGRESSION (Mail): Shift + page up has no effect; should modify selection

• khtml/editing/selection.cpp: (khtml::Selection::modify): Fix problem where vertical distance was used as a distance threshold, but was a negative number. Now make it positive at the start of the function (and make a couple related changes).