[Bug middle-end/69526] ivopts candidate strangeness

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69526

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note that VRP figures out that

_15 = n_18 + 4294967295;
Found new range for _15: [0, 4294967294]

(so no underflow)

_2 = (sizetype) _15;
Found new range for _2: [0, 4294967294]
_1 = _2 + 1;
Found new range for _1: [1, 4294967295]

but VRP doesn't have any transform that would take advantage of this
(like computing _2 + 1 in unsigned int and only then casting to sizetype).