On May 19, 2010, at 1:41 PM, Kevan Miller wrote: > > On May 19, 2010, at 4:28 PM, David Jencks wrote: > >> >> On May 19, 2010, at 1:16 PM, Donald Woods wrote: >> >>> Good summary Lin. >>> >>> In the OpenJPA community, we've been telling users for awhile now not to >>> use the runtime enhancement, as it has known deficiencies, and should >>> only be used for quick prototyping and never in production. >> >> What are these deficiencies? I thought you got the same bytecode whether >> you ran the enhancer using maven or as the classes are loaded using a class >> transformer. I know the subclassing approach doesn't work well but that is >> not being discussed here AFAIK. > > Right. I hope Donald is referring to the subclassing approach. Which I > agree/understand is "deficient" (we've seen instances of this). > >> >> Is this stuff documented clearly somewhere in the openjpa docs? I've never >> found a comprehensible explanation when I've looked in the past. > > Try this -- > http://openjpa.apache.org/builds/2.0.0/apache-openjpa-2.0.0/docs/manual/main.html > > I don't follow section 2.2.
I think you mean 3.5.2.2, but this is the explanation I've found that to me is difficult to understand. I guess the ee5 runtime hooks they refer to are the method on PersistenceUnitInfo (provided by the container) where the provider can try to install a transformer. david jencks > > --kevan
