How does this circuit handle signed bits. I understand how the two's compliment subtraction is implemented as long as A=>B but I am supposed to allow B to be greater...
What efforts have you made to answer your own question? For instance, have you tried picking some values that represent the case you are concerned about? If not, do so and see if that answers your question.
Another hint is that you may understand how it is implemented, but not why that implementation is valid.
Aha! Okay I see now that it gives me the correct number and I just have to take two's compliment again. So to implement this into a circuit which is expected to do both subtraction and addition and give both positive and negative answers...I need to check for b>a somehow, yes?