https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100810
--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #6) > GCC does warn about an unitialized variable i even in "i && d" > For: > if (!b) > i &&d; > > b is even set to 1. Note setting b to be 0 still get the wrong code. The bug is IV-OPTS which selects a variable which is based on an unintialized value. CCP is not the problem, it was just happen to find that h_lsm.23_29 is based on an unintialized value and props the undefinedness forward.
