CancelTarget problem when unmarshalling

[Brent Putman]

On 2/4/14 2:06 PM, Cantor, Scott wrote:
> I stared at it and completely overlooked the bug.

It's easy to miss, I didn't see it at first either.

I looked and there actually is a test for this class.  But all of that
contributed WS-* code uses a very suspect and unreliable mechanism for
the unit tests, with no control XML.  It marshalls to a DOM tree, then
unmarshalls a new XMLObject around it, drops the DOM and then marshalls
again to get a second DOM tree.  And then compares the 2 DOM tress. 
It's not going to catch a bug like this one, where the object impl
itself is horked.  We've always known that, but it's going to be a lot
of work to implement proper tests with control XML for all that
contributed WS-* code.  We probably have some number of other bugs like
this waiting to be found...