rsmith added a comment. Herald added a subscriber: jfb. There's still something wrong here: in the cases where we produce these warnings, we still produce calls to `__atomic_*_<n>` library functions. Those functions are specified to only work on properly-aligned inputs, so that's wrong. And there seems to be a problem with alignment in compiler-rt too: the "unoptimized" `__atomic_load_n` implementation only looks at the size and not the alignment of the pointer when determining whether to use a lock.
See also PR38846. Repository: rC Clang https://reviews.llvm.org/D45319 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits