Jorn, I am puzzled by your problem. At this point I have no idea what could be going wrong. We use scaled-out synch aggregates in uima-as deployements and dont see the problem. We run on fast 8-way machines and I would think any possible race conditions would have shown self by now. Perhaps you have a different configuration that perhaps is somehow different. Is there any way you can reduce your setup to something small that you can package and send to us for debugging? Also, I just checked the stack trace you sent against the current uima-core and cant seem to reconcile the line numbers in the stack trace. It appears that you may not be running with the latest uima core. Check .... org.apache.uima.cas.impl.CASImpl.release(CASImpl.java:3568) ....
In the latest code CASImpl.java line 3568 does not appear to be inside release() method This to me is an evidence that perhaps you may be running with an older version of uima core.. Regards, Jerry On Wed, Oct 7, 2009 at 3:09 PM, Jörn Kottmann <[email protected]> wrote: > Jaroslaw Cwiklik wrote: > >> Jorn, did you update the uima core to the head as well? >> >> > Yes, I think so, updated everything to head in eclipse, > then I did a mvn install in uimaj, afterwards in > uimaj-as and then a mvn assembly:assembly > in uima-as-distr. On the server I copied over all > jar files from the newly created uima as distribution. > > Do you think uima core is not the head version ? > I doubled checked and the project is the head version. > > Jörn >
