ahmedbougacha wrote: > Why do we want a separate builtin, as opposed to just constant-folding calls > to __builtin_ptrauth_sign?
That's a good question. Mechanically, I assumed constant-evaluating `__builtin_ptrauth_sign_unauthenticated` would be a challenge, but looking around it doesn't seem particularly difficult. Conceptually though, the distinction does seem useful, because we treat `__builtin_ptrauth_sign_unauthenticated` as dangerous (because in general we can't guarantee that it won't get lowered to a signing oracle), and only to be used carefully, when absolutely necessary (e.g., in a dynamic loader.) `__builtin_ptrauth_sign_constant` is safe, because we do always have safe ways to materialize signed constants (with the constants in IR, and either relocations or safe pseudos in the backend.) (but cc @rjmccall @ahatanak for the real clang perspective ;) https://github.com/llvm/llvm-project/pull/93904 _______________________________________________ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
