Looks like we are transforming:


<script type="text/javascript" language="javascript">
// <![CDATA[
SI.Mint.collapse    = true;
window.onload   = function() { SI.Mint.staggerPaneLoading(true); 
SI.Mint.sizePanes(); SI.Mint.onloadScrolls(); };
window.onresize = function() { SI.Mint.sizePanes(); };
// ]]>
</script>

with

<script type="text/javascript" 
language="javascript">SI.Mint.collapse=true;window.onload=function(){SI.Mint.sta
ggerPaneLoading(true);SI.Mint.sizePanes();SI.Mint.onloadScrolls();};window.onres
ize=function(){SI.Mint.sizePanes();};</script>

We have lost the 'CDATA' syntax, which I thought we were pretty careful to put 
in when we ineline scripts.  But evidently we are not retaining it when we 
rewrite scripts.

The breakage appears to be due to the use of && in a script block in XHTML.  
Good StackOverflow summary here:

http://stackoverflow.com/questions/66837/when-is-a-cdata-section-necessary-withi
n-a-script-tag

Basically, the script parser worries about different characters (< and & rather 
than </) when parsing a script in the absence of CDATA.  The CDATA puts the 
X[HT]ML parser into "just the characters you see, please" mode.

This is also an issue with scripts we choose to inline into html vs xhtml.

This is now fixed in trunk.