On 6/8/06, Andrus Adamchik <andru..bjectstyle.org> wrote:
> Can you run it in Eclipse (or some other) debugger and step through
> 'getToLevel3Object' method? Then you can actually check the state of
> the objects and see if there is anything weird.
>
> If you still can't find anything, please open a Jira issue with a
> DataMap XML file attached with at least these three entities:
> Level1Object, Level2Object and Level3Object, with relationships
> between the entities preserved the way you have them now. I'll see if
> I can reproduce this.
I fell back to the scientific method: I made a list of suspects and
tested them one by one, changing only one thing at a time.
Here's the culprit (from cayenne.xml):
<property name="cayenne.DataRowStore.snapshot.size" value="10"/>
Now, when I change the value to, say, 10000, I don't see the missing
values problem.
The good news is obviously (or apparently?) that I don't have a
showstopper bug anymore. There's still a worm of doubt in my mind that
I simply masked the problem, made it appear later (when a lot more
objects are loaded into memory) rather than sooner. Is this so?
The bad news is that if I make the DataRowStoresnapshot.size large,
I'm back to square one: I'll have a number of rich clients in a LAN,
each running inside his own jvm, with his own cache - in this case,
the cache gets in the way of data consistency.
So:
1.) why did I get null objects with value="10"?
2.) did I solve the problem (don't think so, the more I think about
it) or simply hidd it?
2.) how do I solve the
multiple-rich-clients-in-individual-jvms-with-their-own-cache problem?
t.n.a.
This archive was generated by hypermail 2.0.0 : Sat Jun 10 2006 - 03:24:58 EDT