It is desirable to compute per-method hashes. This would allow to better pin-point vulnerable code, and reduce the risk somewhat that unrelated modifications obscure the presence of a vulnerable method because the hash does not match any known-vulnerable class anymore.