On Mon, 20 Aug 2018, Richard Biener wrote:
P p+ N is nonnull if P or N is known to be nonnull
(and something similar for &p->field and others)
But we already do that.
Oups... I never noticed, I should have checked.
else if (code == POINTER_PLUS_EXPR)
{
/* For pointer types, we are really only interested in asserting
whether the expression evaluates to non-NULL. */
if (range_is_nonnull (&vr0) || range_is_nonnull (&vr1))
set_value_range_to_nonnull (vr, expr_type);
else if (range_is_null (&vr0) && range_is_null (&vr1))
set_value_range_to_null (vr, expr_type);
else
set_value_range_to_varying (vr);
}
Ah, range_is_nonnull (&vr1) is only matching ~[0,0]. We'd
probably want VR_RANGE && !range_includes_zero_p here. That
range_is_nonnull is probably never true due to canonicalization.
That explains it. Yes please. I am surprised there isn't a helper like
range_includes_zero_p or value_inside_range that takes a value_range* as
argument so we don't have to worry about the type of range (the closest
seems to be value_ranges_intersect_p with a singleton range, but that
function seems dead and broken). When POINTER_PLUS_EXPR is changed to take
a signed argument, your suggested test will need updating :-(
--
Marc Glisse
