Michael Kay wrote:

> One also has to be careful not to be too aggressive in
> type inferencing, particularly with this situation where
> the expression is "void" (guaranteed to return nothing)
> rather than actually being erroneous.

  Yes.  It's why, I guess, Saxon emits warnings instead of
treating them as errors.  Actually, I was surprised when I
looked for the option to pass to Saxon to activate the
static type-checks, because it needs no option.

  IMHO, it could be intresting to have a set of warning
options, for example to toggle warnings on these "void"
expressions, or to treat warnings as errors, or to toggle
other kind of warning.  Like the C++ front-end of GCC (and
others) has a plethora of warning-related options.

  PS: Maybe this discussion has to go on the Saxon ML.  I
don't know.

  Regards,

--drkm



















	

	
		
___________________________________________________________________________ 
Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger 
Tilichargez cette version sur http://fr.messenger.yahoo.com