On Fri, 21 Jan 2022 at 19:53, Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote: > > "On error dst buffer is zeroed out."? This is an explicit guarantee. > I appreciate that existing helpers already do this and it's good to follow suit for consistency, but what is the rationale behind zeroing memory on failure? Can't this step be skipped for performance and leave it up to the callee to perform if they need it?