replicate() does not handle cascade correctly --------------------------------------------- Key: HHH-1047 URL: http://opensource.atlassian.com/projects/hibernate/browse/HHH-1047 Project: Hibernate Core Issue Type: Bug Components: core Affects Versions: 3.1 rc 1 Environment: 3.1b3, postgres 8 Reporter: snorbii Hello, I'm working with sequence-based identifiers, connected to a long property. My workflow: - I have an already saved object A. - A is disconnected for editing by the UI layer. - After editing is ready I call replicate() (with replication mode "latest") to synchronize with the database. My problem is that during editing a new, unsaved object B is added to a many-to-one association of A. This many-to-one association has CascadeType.ALL. - When replicate() is executed, it correctly cascades the replication. - But in DefaultReplicateEventListener.onReplicate() I found the following code: // get the id from the object /*if ( persister.isUnsaved(entity, source) ) { throw new TransientObjectException("transient instance passed to replicate()"); }*/ Serializable id = persister.getIdentifier( entity, source.getEntityMode() ); if (id==null) { throw new TransientObjectException("instance with null id passed to replicate()"); } persister.getIdentifier() returns 0 because object B is unsaved. This means that Hibernate does not realize that it is an unsaved object. So it generates an SQL INSERT for inserting it as a new object, but without generating an ID for it. The result is that object B is inserted to the database with ID=0 instead of an ID got from the sequence. If object A has two such associations then two rows are inserted with ID=0 which causes constraint violation. I think that it would be more correct if replication would save the associated unsaved object B with a normally generated ID if it is unsaved. (Am I right that CASCADED replication should save transient instances instead of throwing an exception?) BR, Norbi