This is not a big concern, because right now, all (non-test) inputs to the parser are `'\0'`-terminated, but we should be future-proof here and not perform these OOB reads.
