⚠ Archived content — this site is no longer maintained.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

Timeline



Jun 9, 2003:

6:54 PM Changeset in webkit [4510] by rjw
  • 6 edits in trunk/WebCore

Fixed 3281761.
This was a real doozie. It boiled down to a frame
attempting to access a siblings window object when that
frame wasn't loaded and didn't have a document.

Reviewed by Dave.

  • khtml/ecma/kjs_window.cpp: (Window::get): (Window::isSafeScript): (FrameArray::get): Made the isSafeScript more tolerant. Always allow JS to execute if the active domain is nil (local). Also allow JS to execute if the target part has no document yet.
  • khtml/khtml_part.cpp: (KHTMLPart::init): (KHTMLPart::openURL): (KHTMLPart::jScript): (KHTMLPart::scheduleRedirection): (KHTMLPart::findFrame): Flag redirects during load so we can later avoid cancelling the redirect when the document would otherwise normally load.
  • khtml/khtmlpart_p.h:
  • kwq/WebCoreBridge.mm: (-[WebCoreBridge addData:]): Document may now be nil. Replace assert with conditional check.
4:42 PM Changeset in webkit [4509] by sullivan
  • 3 edits in trunk/WebCore
  • finished fixing crash in 3275675 -- REGRESSION: crash when replacing <div> which contains iframe (at www.kbs.co.kr)

Reviewed by Darin

  • khtml/rendering/render_replaced.cpp: (RenderWidget::resizeWidget): check for nil element() before ref/deref'ing
4:32 PM Changeset in webkit [4508] by cblu
  • 2 edits in trunk/WebKit
  • WebView.subproj/WebDataSource.h: Removed FIXME related to pageTitle.
3:40 PM Changeset in webkit [4507] by cblu
  • 3 edits in trunk/WebKit

<rdar://problem/3283359>: don't load Cocoa Java plug-in if in Carbon app

Reviewed by darin.

  • Plugins.subproj/WebPluginDatabase.h: removed pluginForFilename, wasn't being used
  • Plugins.subproj/WebPluginDatabase.m: (-[WebPluginDatabase pluginForKey:withEnumeratorSelector:]): don't use the plug-in if ![self isCocoa] && [[[webPlugin bundle] bundleIdentifier] isEqualToString:JavaCocoaPluginIdentifier]
3:39 PM Changeset in webkit [4506] by darin
  • 5 edits in trunk/WebCore

Reviewed by John.

  • first step in fixing 3275675 -- REGRESSION: crash when replacing <div> which contains iframe (at www.kbs.co.kr)
  • khtml/rendering/render_object.h: Made m_node protected, so RenderWidget can zero it out.
  • khtml/rendering/render_replaced.cpp: (RenderWidget::detach): Set m_node to 0 here. RenderWidget does this strange thing of staying around after the DOM node is gone, but we need to get rid of the dangling pointer to the DOM node.
  • khtml/rendering/render_frames.cpp: (RenderFrame::slotViewCleared): Do nothing if element() is 0, since we are on our way out anyway. (RenderPartObject::slotViewCleared): Ditto.
2:29 PM Changeset in webkit [4505] by sullivan
  • 3 edits in trunk/WebCore
  • fixed 3284312 -- repro nil-deref in khtml::RenderTable::cellAbove at pi.nersc.gov

Reviewed by Darin

  • khtml/rendering/render_table.cpp: (RenderTable::cellAbove): When the section had 0 rows, we were trying to index the -1st row. Fixed by falling into the "didn't find anything" case when the row we're going to look in is < 0.
2:16 PM Changeset in webkit [4504] by cblu
  • 2 edits in trunk/WebKit

Fixed: <rdar://problem/3284848>: REGRESSION: repro crash in Flash handling null event, going back to Japanese Disney page

When restarting plug-ins from the BF cache, we were not calling NPP_SetWindow.

Reviewed by darin.

  • Plugins.subproj/WebBaseNetscapePluginView.m: (-[WebBaseNetscapePluginView stop]): set the window type to 0 to force the calling of NPP_SetWindow
12:40 PM Changeset in webkit [4503] by darin
  • 3 edits in trunk

Removed misleading Safari-74 markers.

12:36 PM Changeset in webkit [4502] by darin
  • 20 edits in trunk/WebCore

Reviewed by John.

  • fixed 3277733 -- REGRESSION: \ in JavaScript mishandled when encoding is Japanese (istweb.apple.com)

This is the third time I've fixed handling of backslash and yen sign. Each time I fixed part
of the problem but either didn't fix the whole thing or caused a regression. This time I did
more experiments with other browsers, and I came to the conclusion that all other browsers keep
the backslash character internally unchanged, and only change to the yen sign when displaying
rather than decoding to the Unicode "yen sign" character. This makes the backslash character
different from any other, and requires special code, but it's the only way to match the other
browsers' behavior. This seems to work great; I retested all the backslash/yen bugs from the
past to make sure I didn't cause new regressions. The only loose end is the DOM API. It's not
clear whether the DOM API should return the strings with backslashes or with yen signs, but
we can probably ship 1.0 without getting that 100% right.

  • kwq/KWQTextCodec.h: Added backslashAsCurrencySymbol() member function.
  • kwq/KWQTextCodec.mm: (QTextCodec::fromUnicode): Removed the old workaround for tilde problems, which is not needed now that we are back to using "force ASCII" mode, so tildes are always tildes. Updated the comment about the backslash hack to better reflect the new situation and change it to use the new backslashAsCurrencySymbol() member function. (QTextCodec::backslashAsCurrencySymbol): Added. Returns the yen sign for the encodings where it's appropriate. Later might do the same thing with the won sign for Korean, depending on what we discover as we research bug 3221037 (Korean currency symbol should replace backslash). (KWQTextDecoder::convertUsingTEC): Add back the call to TECSetBasicOptions since we once-again want to use force-ASCII mode; we turn backslashes into yen signs at a higher level now; internally they are kept as backslashes.
  • khtml/xml/dom_docimpl.h: Made setDecoder no longer be inline, and added a decoder member function so callers could get the decoder (they need it to figure out what to do with backslashes).
  • khtml/xml/dom_docimpl.cpp: (DocumentImpl::setDecoder): Made this no longer be an inline. Also made sure it derefs the existing decoder. This probably does not happen in practice, but the old version would leak if it was ever called twice on the same document.
  • kwq/KWQString.mm: (QString::replace): Made sure that the case of replacing a character with the same character has an early out, since that's now the common case with this "map backslash to currency character if needed" feature.
  • khtml/xml/dom_stringimpl.h: Made capitalize() const and added a new replace() member function.
  • khtml/xml/dom_stringimpl.cpp: (DOMStringImpl::replace): Implemented this new function that replaces one character with another. Returns the existing string if there is no replacing to do, otherwise makes a new one.
  • khtml/rendering/render_object.h: Added backslashAsCurrencySymbol() member function.
  • khtml/rendering/render_object.cpp: (RenderObject::backslashAsCurrencySymbol): Added. Helper function so render objects have an easier time digging for the "do I need to turn the backslashes into currency symbols" flag.
  • kwq/KWQKHTMLPart.h: Added backslashAsCurrencySymbol() member function.
  • kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::setTitle): Convert title to use yen sign. (KWQKHTMLPart::setStatusBarText): Convert status bar text to use yen sign. (KWQKHTMLPart::runJavaScriptAlert): Convert message text to use yen sign. (KWQKHTMLPart::runJavaScriptConfirm): Convert message text to use yen sign. (KWQKHTMLPart::runJavaScriptPrompt): Convert message text and default value to use yen sign, convert result back from yen sign to backslash. (KWQKHTMLPart::attributedString): Convert pieces of text to use yen sign when building up the attributed text. (KWQKHTMLPart::backslashAsCurrencySymbol): Added. Helper function so render objects have an easier time digging for the "do I need to turn the backslashes into currency symbols" flag.
  • kwq/WebCoreBridge.mm: (-[WebCoreBridge selectedString]): Convert text to use yen sign before returning to WebKit. (-[WebCoreBridge elementAtPoint:]): Convert title and alt strings to use yen sign before returning them to WebKit.
  • khtml/rendering/render_form.cpp: (RenderSubmitButton::rawText): Convert button label to use yen sign. (RenderLineEdit::updateFromElement): Convert text for line edit to use yen sign. (RenderLineEdit::slotTextChanged): Convert yen sign to backslash on the way in from line edit. (RenderSelect::updateFromElement): Convert text for select items to use yen sign. (RenderTextArea::updateFromElement): Convert text for text areas to use yen sign. (RenderTextArea::text): Convert yen sign to backslash on the way in from text area.
  • khtml/rendering/render_image.cpp: (RenderImage::paintObject): Convert alt text to use yen sign before drawing it.
  • khtml/rendering/render_text.cpp: (RenderText::RenderText): Convert text to use yen sign before drawing it. (RenderText::setText): Ditto.
  • fix a problem that was bugging the shit out of me where we kept rebuilding config.h
  • Makefile.am: Use the config-timestamp file to decide when to remake config.h so we don't end up making it over and over again. Also don't bother to grep out NO_LICENSE any more since that's long gone.
  • .cvsignore: Ignore the config-timestamp file.
9:40 AM Changeset in webkit [4501] by sheridan
  • 3 edits in trunk

Safari-84u

9:34 AM Changeset in webkit [4500]
  • 3 copies in tags/Safari-83

This commit was manufactured by cvs2svn to create tag 'Safari-83'.

9:34 AM Changeset in webkit [4499] by sheridan
  • 8 edits in trunk

Safari-83

8:42 AM Changeset in webkit [4498] by darin
  • 7 edits in trunk/WebCore

Reviewed by Dave.

  • fixed 3280538 -- JavaScript infinite loop loading frame b/c scrollHeight is always zero (easons.ie)
  • khtml/xml/dom_docimpl.h: Added updateLayout method.
  • khtml/xml/dom_docimpl.cpp: (DocumentImpl::updateLayout): Added. Code moved here from KJS::Window::updateLayout so it can be shared. Named this updateLayout() to match the name from KJS::Window.
  • khtml/ecma/kjs_window.cpp: (Window::updateLayout): Call updateLayout() instead of updateRendering(); the extra logic here was moved into updateLayout().
  • khtml/ecma/kjs_dom.cpp: (DOMNode::getValueProperty): Ditto.
  • khtml/ecma/kjs_html.cpp: (KJS::HTMLElement::getValueProperty): Call updateLayout() on the document before computing the BodyScrollHeight or BodyScrollWidth, much as we already do in kjs_dom.cpp for ScrollHeight and ScrollWidth and other properties.

Jun 7, 2003:

5:59 PM Changeset in webkit [4497] by darin
  • 2 edits in trunk/WebKit
  • English.lproj/StringsNotToBeLocalized.txt: Updated for recent changes.
5:51 PM Changeset in webkit [4496] by darin
  • 3 edits in trunk/WebKit

Rolled out Chris's fix for 3156169 because it was causing a lot of crashes
and problems with basic behavior. We can try again later.

One of the problems was that Back wasn't working. Another was that you could
not follow the link at the top of the page at kbb.com. More testing seemed to
reveal still more problems.

  • WebView.subproj/WebFramePrivate.h: Rolled back to previous version.
  • WebView.subproj/WebFramePrivate.m: Ditto.

Jun 6, 2003:

4:48 PM Changeset in webkit [4495] by rjw
  • 2 edits in trunk/WebKit

Fixed 3283236. Remove use of forward declarations in public header because
CodeWarrior pukes on 'em.

Reviewed by Chris.

  • WebView.subproj/WebDataSource.h:
4:22 PM Changeset in webkit [4494] by cblu
  • 3 edits in trunk/WebKit

Fixed: <rdar://problem/3156169>: cmd-click opens new win but stops loading in prev win

Reviewed by darin.

  • WebView.subproj/WebFramePrivate.h: added policyDataSource and policyLoadType as ivars
  • WebView.subproj/WebFramePrivate.m: (-[WebFramePrivate dealloc]): added ASSERT (-[WebFramePrivate setProvisionalDataSource:]): added ASSERT (-[WebFrame _isLoadComplete]): formatting tweak (-[WebFrame _invalidatePendingPolicyDecisionCallingDefaultAction:]): clear policyDataSource (-[WebFrame _checkNewWindowPolicyForRequest:action:frameName:formState:andCall:withSelector:]): formatting tweak (-[WebFrame _checkNavigationPolicyForRequest:dataSource:formState:andCall:withSelector:]): retain policyDataSource (-[WebFrame _continueAfterNavigationPolicy:]): stop the load, set the load type, set the provisional data source in the "use" case (-[WebFrame _loadURL:referrer:loadType:target:triggeringEvent:form:formValues:]): formatting tweak (-[WebFrame _continueLoadRequestAfterNavigationPolicy:formState:]): don't reset things if something other than "use" has been chosen (-[WebFrame _loadDataSource:withLoadType:formState:]): DON'T stop the load, set the load type, set the provisional data source
3:53 PM Changeset in webkit [4493] by cblu
  • 2 edits in trunk/WebKit

Fixed: <rdar://problem/3282881>: Java plugin fails in carbon WebKit apps

Fixed by Mike Hay, reviewed by me.

  • Plugins.subproj/WebPluginDatabase.m: (-[WebPluginDatabase initIsCocoa]): (-[WebPluginDatabase isCocoa]): (-[WebPluginDatabase pluginForKey:withEnumeratorSelector:]): don't load cocoa plug-ins if isCocoa.
10:47 AM Changeset in webkit [4492] by sheridan
  • 3 edits in trunk

v83u, forgot to do this after the build on Tuesday

3:31 AM Changeset in webkit [4491] by hyatt
  • 3 edits in trunk/WebCore

Fix for crash on an XHTML frameset page. It was assuming
that all docs were HTML. The bug # is 3277973.

Reviewed by john

  • kwq/KWQKHTMLPart.mm: (KWQKHTMLPart::updatePolicyBaseURL): (KWQKHTMLPart::setPolicyBaseURL):

Jun 5, 2003:

12:24 PM Changeset in webkit [4490] by hyatt
  • 1 edit in trunk/WebCore/khtml/html/html_headimpl.h

* empty log message *

12:05 PM Changeset in webkit [4489]
  • 1 copy
    79 deletes in branches/unlabeled-1.35.18

This commit was manufactured by cvs2svn to create branch
'unlabeled-1.35.18'.

12:05 PM Changeset in webkit [4488] by hyatt
  • 4 edits in trunk/WebCore

Fix for 3281030, stylesheets can't be enabled properly via
script. Test case from Dirk. The fix is to make the setting
of the state happen independently of the attribute setting.

Reviewed by john

  • khtml/dom/html_head.cpp: (HTMLLinkElement::setDisabled):
  • khtml/html/html_headimpl.cpp: (HTMLLinkElementImpl::setDisabledState): (HTMLLinkElementImpl::parseAttribute):
  • khtml/html/html_headimpl.h:
11:36 AM Changeset in webkit [4487] by cblu
  • 3 edits in trunk/WebCore

Fixed: <rdar://problem/3268751>: REGRESSION: crash in KWQValueListImpl selecting connection speed at news.com

Reviewed by darin.

  • kwq/KWQButton.mm: (QButton::clicked): Don't call clicked if the button was destroyed inside of sendConsumedMouseUpIfNeeded.
11:23 AM Changeset in webkit [4486] by sullivan
  • 8 edits in trunk/WebKit
  • fixed 3266216 -- repro crash in -[WebBaseResourceHandleDelegate connection:didReceiveData:] in GIA Application

The problem was that an NSURLConnection delegate object (in this
case a WebMainResourceClient) was being dealloc'ed during one of
its connection delegate methods. To prevent this kind of problem,
I added [self retain]/[self release] guards around the meat of
all of the connection delegate methods in which arbitrary code
could be run. Another approach would be to do this retain/release
pair in NSURLConnection, but Darin deemed it wiser not to muck with
Foundation at this point for this issue.

Reviewed by Darin

  • Plugins.subproj/WebNetscapePluginStream.m: (-[WebNetscapePluginConnectionDelegate connection:didReceiveResponse:]): guard with [self retain]/[self release] (-[WebNetscapePluginConnectionDelegate connection:didReceiveData:]): ditto (-[WebNetscapePluginConnectionDelegate connection:didFailWithError:]): ditto
  • WebCoreSupport.subproj/WebSubresourceClient.m: (-[WebSubresourceClient connection:didReceiveResponse:]): ditto (-[WebSubresourceClient connection:didReceiveData:]): ditto
  • WebView.subproj/WebBaseResourceHandleDelegate.m: (-[WebBaseResourceHandleDelegate connection:willSendRequest:redirectResponse:]): ditto (-[WebBaseResourceHandleDelegate connection:didReceiveAuthenticationChallenge:]): ditto (-[WebBaseResourceHandleDelegate connection:didCancelAuthenticationChallenge:]): ditto (-[WebBaseResourceHandleDelegate connection:didReceiveResponse:]): ditto (-[WebBaseResourceHandleDelegate connection:didReceiveData:]): ditto. Also, commented out two assertions that fire illegitimately in the steps in this bug report. (-[WebBaseResourceHandleDelegate connection:didFailWithError:]): ditto
  • WebView.subproj/WebMainResourceClient.m: (-[WebMainResourceClient connection:willSendRequest:redirectResponse:]): ditto (-[WebMainResourceClient connection:didReceiveResponse:]): ditto (-[WebMainResourceClient connection:didReceiveData:]): ditto
9:47 AM Changeset in webkit [4485] by darin
  • 3 edits in trunk/WebCore

Reviewed by John.

  • fixed 3279206 -- REGRESSION: new page does not load after selecting item from list and hitting 'Go' button
  • khtml/html/html_formimpl.cpp: (HTMLInputElementImpl::encoding): Use value() in a couple of places that were still directly using m_value.
8:41 AM Changeset in webkit [4484] by darin
  • 3 edits in trunk/WebCore

Reviewed by John.

  • fixed 3278079 -- REGRESSION: UTF-8 page claiming to be UTF-16 in XML header gives garbage characters
  • khtml/misc/decoder.cpp: (Decoder::setEncoding): Add EncodingFromXMLHeader as one of the types where we know the encoding is an 8-bit one. The reason this is a regression is that we used to ignore the encoding in the XML header altogether.
Note: See TracTimeline for information about the timeline view.