Re: [PATCH]: Handle CONJ_EXPR in negate_expr_p and fold_negate_expr

["Richard Guenther" <richard dot guenther at gmail dot com>]

On 2/5/07, Paolo Bonzini <paolo.bonzini@lu.unisi.ch> wrote:
> >> For example, (a,b) + -~(x,5) would change to (a,b) + (-x,5) which is
> >> wrong; but (a,b) + -~(5,x) would change to (a,b) + (-5,x) which is ok.
> >
> > The patch does fold -~z to ~-z which is ok, can you elaborate why
> > -~(x,5) -> (-x,5) is wrong?  (it wouldn't do that, beause negate_expr_p
> > will return false for (x,5))
>
> For the same reason which Kaveh explained in his previous patch.

Now, if we start with -~(x, 5) then first we fold that to -(x, -5) and leave it
this way.  If we start with -~(5,x) we fold this to -(5,-x) (maybe questionable,
see fold_unary of CONJ_EXPR) and then to (-5,x).  I think both are ok.
(The wrong folding won't occur with the patch posted)

Richard.