[ https://issues.apache.org/jira/brow...action_12474011 ]

Craig Russell commented on JDO-460:
-----------------------------------

>OK. I have tried to fix this but come across the issue of the object state being pa
ssed in to this operation being incorrect. In the StateTransitions.java when creatin
g objects HOLLOW and P_NONTRANS are treated the same in general. This leads to a P_N
ONT
RANS being passed in to part 1 above, and HOLLOW being passed in to part 2 a
bove. Going backwards getHollowInstance() will sometimes return a P_NONTRANS
, and getPersistentNontransactionalInstance() will sometimes return a HOLLOW
 ... which upsets things a
little when there is different expected behaviour between these two.

Right. The spec was updated so that detachCopy with either HOLLOW or P_NONTR
ANS should transition to P_CLEAN. The rationale is that detaching instances 
causes their fields to be read, which should cause a transition to P_CLEAN w
ith a datastore transaction

[ Post a follow-up to this message ]