As often happens, the effort of producing a test case found the bug on
our side. We have some low-level code that was clearing the dataContext
(rollbackChanges()) whenever a new page is invoked, and it was being
called between the create and the commit. I didn't see this at first,
being led astray by the fact that the commit used to work on the older
cayenne (weird, eh?) and the fact that the object looked correct when
it was printed to the log. Sorry about the false alarm.
Andrus Adamchik wrote:
> Bryan,
>
> nothing comes to mind right away. Could you possibly create a Jira
> issue with a DataMap that contains the entities involved - I'll run
> it in debugger and see what's going on.
>
> Andrus
>
>
This archive was generated by hypermail 2.0.0 : Wed Apr 19 2006 - 00:21:27 EDT