https://github.com/boomanaiden154 commented: The ml-opt bots do not run with LLVM_ENABLE_WERROR because we don't control the host toolchain too closely. We can just drop the warning flag changes. https://github.com/llvm/llvm-project/pull/194593