https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65752
--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Robbert from comment #9)
> (In reply to Richard Biener from comment #6)
> > which fixes the testcase (but is incomplete - equivalences built on
> > gimple assignment RHS need to be considered as well).
> Can you give an example?
just do
_Bool cmp = (i == (uintptr_t)&x);
if (cmp) { .... }
in your original testcase (ok, optimization will probably make them equivalent,
but you can't rely on that one).
Similar for function calls
if (equal (i, (uintptr_t)&x))) { ... }
with the obvious implementation of equal. Or
global = (uintptr_t)&x;
if (equal (i)) { .... }
and see how this will make PTA useless (all pointers passed to a function
whose result might be used in a way to take advantage of an equality relation
need to be considered pointing to anything). [and then thorougly specify
"take advantage of an equality relation"]
