This fixes the problem by checking for whether ReadMemory read *any*
bytes, instead of checking whether it returned an error. A greater
question is whether ReadMemory should even return an error if it read at
least one byte, but I'm leaving that for a separate patch.