fold_builtin_memcpy cannot optimize <retval>?

[Gabriel Dos Reis]

Andrew Pinski <pinskia@physics.uc.edu> writes:

| On Jan 21, 2005, at 2:37 PM, Mark Mitchell wrote:
| 
| > I'm not sure.  I seemed to believe in that thread that they could
| > never be NULL, but I'm not sure if everyone agreed, and I see that
| > we suggested a non-NULL tag on objects somewhere.  So, in that case,
| > there may be no difference at all to the middle-end.  In any case, I
| > think that it would make sense for us to allow a REFERENCE_TYPE
| > anywhere we allow a POINTER_TYPE.
| 
| 
| Isn't this valid C++ then (or is it just undefined):

It is undefined behaviour.

-- Gaby