>Actually I don't see how I can do that. A TransformerException is pretty
>much immutable, so I guess I must throw the one that was passed in without
>altering it.
Or generate a new one based on the one that was passed in ... though that's
how we're winding up with the hard-to-read multiplyh nested exceptions.
>Xalan should probably not interpret a halt-processing-now
>exception coming back from an ErrorListener as a reason to call
>ErrorListener#fatalError
ErrorListener's normal path is to return void. Since it could have caught
and handled any exceptions that occurred during its processing, having it
actually _throw_ an exception should be considered... well, an exceptional
case, and either a bug or a deliberate attempt to signal an unrecoverable
condition.
Whether error-handler bugs and panics should be reported as fatalErrors, or
whether only fatal errors arising in main processing stream should be
passed to that method, is certainly debatable. We should probably get the
consensus of the SAX community rather than reinventing that decision
ourselves.
