https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79929
--- Comment #10 from Jeffrey A. Law <law at redhat dot com> --- At -O2 the memset is removed very early in the optimization pipeline. Thus there's no warning at O2, but there is a warning at O1. Reality is some warnings are going to give false positives when optimizations are throttled back and I wouldn't call that a regression for the C testcase. What I am concerned about is in the original Fortran based test, we warned at -O2, but not at -O1. Something has been lost in the translation to C.
