1 Posted by sryo on 24 November 2005 | Permalink

It's been fixed in gecko 1.8.

2 Posted by Tom on 21 December 2005 | Permalink

Has this been fixed in Gecko 1.8? I believe I'm seeing the described behavior in Firefox 1.5.

3 Posted by Alex Bocast on 24 May 2007 | Permalink

This behavior persists in Firefox 2.0.0.3, Gecko 1.8.3. Indeed, it is even weirder than Wilson reports. If you swipe horizontally over a text box that is completely within a containing element, four events are always generated, but the mouseOver or mouseOut events for which the text box should be the relatedTarget generate instead a faulty relatedTarget, which throws an exception if any property is requested. If you swipe over the same text box vertically, the number of reported events may may be 0, 4, 6, or 8. The spurious events are of two types, as noted by Wilson: one type generates a faulty relatedTarget, which throws an exception if any property is requested; the other type reports that the relatedTarget is the same as the target. If one or more of the latter type of event is present, they seem to be always between a pair of mouseOver & mouseOut events with faulty relatedTargets.

4 Posted by Peter Hart on 11 October 2007 | Permalink

I've been looking at the same issue, I found this reference through google. I did notice, additionally, that in firebug if you log the event.relatedTarget node as a an object (e.g. using console.log("%s : %o","related target",event.relatedTarget) ), it appears that there's a hidden 'div' that is a child of the input node. Cut and Paste the HTML (below, minus angle-brackets to get through comment restrictions) and you see something like:

div class="anonymous-div" style="white-space: pre;" _moz_editor_bogus_node="TRUE" _moz_dirty=""

Presumably this is some artifact of the implementation of the input node? If you click on the node hyperlink in the logs, you can see that this is a child of the input node itself.