Fixed #1959 by postponing ALL script evaluations till the html insertion is done. Before the code would immediately execute any scripts that weren't in subelements if no subelements prior had scripts in them (i.e. once any script was postponed they all were). This could cause inconsistent behavior. Since, at this time, we have to postpone some scripts it makes more sense to postpone them all.

Added a change that triples the speed of all uses of $(...). For example $(DOMElement) was 38ms, is now 13ms.

Fixed clone so that it now properly copies changes to the  innerHTML in IE. Unfortunately, IE stores some modifications to some attributes only as a property and they are still not copied properly. This is documented in ticket #1836.

Fixed #1039 and #1733 by going through the core API and making them text node and comment node safe.

Added ajax error messages in the unit test suite because Safari was sometimes tossing errors that were being ignored and that caused a freeze in the test suite.

Safari sometimes stops in the test suite on the ajax section when I run the whole suite and its running on my local machine.  I went ahead and put better error detection in for the test that is the culprit so it will keep going.  It seems to do fine if I run it from a server.  It gives xml.status as 0.  Maybe this is the reason http://www.pearweb.com/javascript/XMLHttpRequest.html

A partial rollback of the refactoring done in [4032]. It was causing a 4-6% speed decrease on certain hierarchy selectors ( > + and ~ ) since it did an extra .toUpperCase() which wasn't required.

The part left in was moving one of the .toUpperCase() calls to the var nodeName instead of having it in the loop.  This appears to be giving a speed boost of a couple percent for those same hierarchy selectors.

Fixed the build/speed testing application to work with current versions of jQuery. The main changes were implementing :has() where necessary to replace functionality that no longer existed and to use noConflict() so the libraries wouldn't have to be modified for the speed testing.

So now the file jquery-basis.js will be tested against the jquery.js in the /dist directory. Version 1.2.1 is currently in jquery-basis.js.

Refactored jQuery.find() such that jQuery.nodeName() is used to test the node names like .filter() and other functions do.  This puts in place possible fixes for tickets like #1991 where XML and HTML differ in case-sensitivity.

Fixed #1419 where IE failed with .text() on an XML node.  This is part of a series of tickets including #1264 where the context of the DOM manipulation was a problem in xml and iframe documents.

Fixed #1264.  If you read the bug there were many proposed changes.  As it turned out most of them had already been implemented.  The last ones necessary were in .domManip() with when a <table> was 'this' and for .text().  Adding these last changes seems to make dom and text manipulation in IE frames possible.  Unit test cases were added as well.

In addition "submit.gif" was removed from the test suite index.html since it didn't exist.

Fixed #1763 by checking to see if .createElement() is available on the context and if isn't default to other contexts.

Fixed #1438 where a filter could be set in IE but not have opacity in it. The JS error was fixed by checking to make sure 'opacity=' is in the filter before seeing what its value is.

There was a disabled test in the ajax test suite which said there were too many simultainous requests.  I re-enabled it when I found a bug that might have been the cause of the failure instead and it seems to work fine.  We can disable it again if that ends up not being the case.

Fixed [1993] although it actually wasn't a bug in the core but rather a misunderstanding of how the extra function was supposed to work in jQuery.event.trigger().  That said, it seems more useful and robust for the code to work the way the ticket author thought it should work so this change was made.  

Now, if anything is returned from the extra function it will overwrite the return value of the event handlers.  This should only effect custom events unless someone had an extra function that returned a value other than false which would have been ignored before.

Added enchancement for #1994 by adding two parameters to .stop() which give additional functionality.  The first parameter clearQueue will clear the queue on the necessary DOM elements so all animation will stop.  The second parameter will cause the currently playing animation to immediately complete including reseting original styles on show and hide and calling the callback function.  If no parameters are passed it will work as it always did.

While adding unit testing I noticed the stop() unit test wasn't working correctly because the element was hidden so I fixed it and added more unit tests around the new functionality.  I also added a cursor:pointer to the css (because for a long time I didn't know they were clickable).

Removed comments and code related to Safari being unable to do a sync global eval.  With the new globalEval() code that should no longer be a problem.

Fixed #1557, although it doesn't appear to be just an FF3 problem. In this case, $.getJSON() wasn't working from a remote host. I went ahead and added a unit test then added the s.dataType == "json" test for a remote <script> load.  The said that json was allowed but the dataType check was missing. This appears to have fixed the bug across all browsers.

Fixed #1450 by returning 1223 as success.  Couldn't normalize the status since jquery uses the actual xmlhttprequest object.

Back out one of the changes from the previous commit that wasn't necessary to fix the bug and might not be desired.

Fixed #1070 by converting all setAttribute() values to a string which is what all browsers but IE did.  This will bring IE in line with the others and fix the bug.

Fixed #1599 as Brandon suggested to ignore negative values to width and height css.  The fix itself is slightly different as it was moved to .css() instead of staying in .attr() like in his patch.  I decided there was less chance of incorrect behavior (like if someone had an XML file with a width attribute that could be negative).  Also took out some unneeded white space while I was in there.

Fixed #1942 but running jQuery.css() before jQuery.curCSS().  This way when the property is width or height it gets the values through calculation instead of just css first. This appears to fix the problem in Opera without hurting any of the other browsers.

Fixed #1970 by returning true instead of false when the mouse moves over a sub-element.  The side effect is the event will not stop default behavior and will propagate which it didn't used to.  I could find no compelling reason to stop those things from happening.