In one of the cases I've seen, it's actually in a read-only scenario where
the objects are not explicitly cached and where no custom query is used.
However, the Expression that is used to build this query is undeniably ugly
and involves numerous relationships, flattened or otherwise.
I will continue to poke around to see if I can't isolate it. We've had 3
separate occurences of this, each in a unique location. It's possible that
it is Tomcat munging up the sessions somehow - it wouldn't be the first
time.
Thanks,
Cris
> -----Original Message-----
> From: Andrus Adamchik [mailto:andru..bjectstyle.org]
> Sent: Friday, March 04, 2005 12:33 PM
> To: cayenne-use..bjectstyle.org
> Subject: Re: Corrupt object graphs
>
> Cris,
>
> I am not aware of any general problems in Cayene that corrupt object
> graph, but I can imagine a scenario when you can get objects that you
> didn't expect in a to-many relationship. This may happen if you use a
> custom query to resolve a relationship, applying permissons based
> filtering or if you do in-memory filtering and then stick the
> result back
> to the relationship list.
>
> Andrus
>
>
>
>
> > We've recently deployed one of our Cayenne applications to
> production,
> > where it is taking approximately 20k hits per day in a
> JDK1.4/Tomcat 5.0
> > environment. This is fairly light load, but it is a highly
> data driven
> > application.
> >
> > We've had two or three occasions now where users have had
> objects that
> > are cached in their session scope "change" or receive
> connections to new
> > objects that they did not previously have access to. This is led to
> > harmless escalation of privileges (seeing other people's
> information in
> > a list view, but not being able to open it for manipulation).
> >
> > Our object graphs are quite large due to the scope of data,
> and so its
> > not unreasonable that at any time we have several hundred thousand
> > objects floating across all user sessions. While I have not linked
> > Cayenne to the problem, I'm wondering if its ever been
> tested with this
> > many objects. I noticed that the default object limit was
> fairly low,
> > and in fact we had to increase it because certain queries
> were having
> > their prefetches lost, triggering 1-at-a-time reloads of
> massive amounts
> > of data (it is common for us to have a query that yields
> 30k results,
> > not including prefetches).
> >
> > Anyone have any ideas on what might be causing this?
> >
> > Thanks,
> >
> > Cris
>
>
>
>
>
This archive was generated by hypermail 2.0.0 : Fri Mar 04 2005 - 12:38:51 EST