Additional bug in Range.extractContents

[Lynn Monson <lm...@flipdog.com>]

I've found an additional test case for Range.extractContents where Xerces
produces the incorrect result.

Given the following document:

<a><b/><c/></a>

Create a range:

Start container: <b>
Start offset: 0
End container: <c>
End offset: 0

After invoking range.extractContents, I get a document fragment that looks
like:

<b/><c/>

This is correct.  Unfortunately, the original document has been modified
incorrectly.  The original document now looks like:

<a/>

But should still look like:

<a><b/><c/></a>

Since neither <b> nor <c> were wholly selected.  In addition, the <b> and
<c> nodes in the document fragment are the *original* nodes, not a clone of
those nodes as specified in Dom 2.