Not sure if this is the recommended way, but I set autoflush to never
and manually call both flush and commit (though I have my base
repository commit call flush automatically) that way I know exactly
when these things happen. After reading ayende's article I realized
that is how I have been dodging this bullet as I have never seen the
behavior you describe. I tend to setup transactions in test and call
flush to make sure it has gone to DB but don't want it committed, then
at the end of the test I rollback and no changes are persisted which
works quite well.
> 2009/10/18 acl123 <
andrewclawre...@gmail.com>
> Fabio Maulo- Hide quoted text -
>
> - Show quoted text -