https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88433
--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> --- Then again, the "problem" would disappear if the middle-end could be made to understand that memcmp(&p, &q, sizeof p) is the same thing as p == q for any integer or pointer types p and q. So maybe it is a middle-end issue because the back-end doesn't know and can't tell that y is the same as *p.
