Thanks for this information. I'm about to rewrite this stuff a bit anyway due to some other bug, and I'll take your findings into account. Thanks again.
What I really wonder about is, that this code used to work actually. This is why this whole logic with filtering an images in there... It surely needs a proper unit test this time! --Klaus On 11 Feb., 02:51, brian <bally.z...@gmail.com> wrote: > On Tue, Feb 10, 2009 at 4:17 PM, Klaus Hartl <klaus.ha...@googlemail.com> > wrote: > > > I've created the ticket aready: > >http://dev.jqueryui.com/ticket/4109 > > Thanks, Klaus. I think I've sorted this out, actually. Although I'm > not certain of a fix. > > I should have been logging this[0].tagName, not > inner(this)[0].tagName. The answer was staring me right in the face, > above that cleanup block: > > var inner = function(parent) { > var $parent = $(parent), $inner = $parent.find('*:last'); > return $inner.length && $inner.is(':not(img)') && $inner || $parent; > > }; > > It's the $parent.find('*:last') that's the culprit. Wrapping the img > in a span does no good because the img, being deepest, will be > selected every time. Because we don't want the img returned, we'll > always get $parent--which won't ever be the parent of the IMG because > that's wrapped in the EM. > > I changed that to: > $inner = $parent.find('*:not(img):last'); > > This works! > > Of course, it fails if I have: > spinner: '<span><img src="/css/images/misc/spinner.gif" /></span>' > > ... because the SPAN will be returned by inner() instead of the desired EM. > > So, it seems that there needs to be some other way to get the inner > element. Maybe the spinner option html fragment needs to be parsed and > the tags saved in data to better access them later? > > Or, what if, instead of having the EM always, you instead made the > default to be: > <em>loading ...</em> > > Don't wrap the user-provided HTML fragment inside of an EM and store > the data in the A instead of the EM. ... and, I'm not sure where I'm > going with this. I'll have to think about it some. > > Anyway, this seems to be the problem.
