I have some problematic behavior with Hibernate 3.2, and want to ascertain whether this is a bug:

class A {
Set<B> bset;
}
(for this class, the bset relationship is cascade="all-delete-orphan")

a = new A();
session.saveOrUpdate(a);
session.close();

session = hsf.openSession();
a.bset.add(b);
a = session.merge(a);
a.bset.remove(b);
session.saveOrUpdate(a);
session.close();


Most of this works as expected, with the exception that at the end of the code sequence, b is persisted.

b should not be saved since the cascade includes delete-orphan, correct?

Hi,
You can apply the following changes to your code,
Replace
>>> a.bset.remove(b);
with
<<< session.flush() ; (1)
<<< a.bset.remove(a.bset.iterator().next()) ; (2)

Reason: (2)
a = session.merge(a) ; would cascade merge to child b ( in a ).
So, the b that we have with us is no longer associated with sesssion ( and with 'a' ).
You can test this. ( a.bset.size() = 1 , even after you remove b, because they are different objects. )

Reason: (1)
I am not clear at this point of time.

I assume that transactions are in place. ( obvious, else you don't expect
data to commit in database )
--------------------------------------------------------
Rate the reply if you find it helpful

thanks for the reply pramodkp --

this certainly is a workaround, but actually an easier workaround is to simply to add: session.delete(b) after the call a.bset.remove(b)
I tested this and it correctly deletes the newly created b object.

However, what i'm really asking is whether the code I presented is buggy behavior, because its misleading and seems contrary to the rest of hibernate's behavior.