Oh I see! You (redundantly) include the length of the plaintext, and require extra code on the decryption side to remove the padding. If you used Crypt::CBC, you wouldn't need to include the length of the plaintext in the ciphertext.

Caveat: I'm assuming FlushFinalBlock uses a padding method compatible with Crypt::CBC's. Well, Crypt::CBC since Crypt::CBC can handle any padding, so I guess I meant "compatible with the default Crypt::CBC padding method".