I've done a full build with an empty maven repo and I'm still able to reproduce most of the time the hanging thread problem. I've also managed to reproduce it in Eclipse in debug mode but couldn't find the root cause as all I can see is that the process is hanging on the DestoryJavaVM thread. I've looked into the VM with VisualVM and I can see that along with the DestroyJavaVM thread there are some RMI related threads still running (possibly the RMI registry?). Here are some screenshots after I've left the app to run about 20 mins, maybe you can spot something.
eclipse in debug mode http://img17.imageshack.us/img17/5489/eclipsef.png monitor http://img10.imageshack.us/img10/9939/monitoro.png threads http://img806.imageshack.us/img806/7286/threads.png Same behavior with both Sun JDK and OpenJDK On Wed, Feb 2, 2011 at 2:24 PM, Florian MOGA <moga....@gmail.com> wrote: > The path problem was generated by the fact that I've taken out fork="true" > in order to see a more specific error than "Java returned 130". Without > fork, a different classpath is loaded and the contribution can't be found. > With fork="true", the contribution is loaded successfully but the test hangs > after removing the endpoints for RMI as long as I don't kill the process. > > I've tried debugging through it in Eclipse but everything worked fine. The > following lines appeared in the Eclipse log and didn't appear in the Maven > log: > > Feb 2, 2011 1:55:51 PM org.apache.tuscany.sca.http.jetty.JettyServer > removeServletMapping > INFO: Removed Servlet mapping: > /CalculatorServiceComponent/CalculatorService > > Eventually, I've set the JAVA_HOME variable to OpenJDK 1.6.0_20 and > everything worked fine. I wonder if the problem is reproducible on other > platforms as well with Sun JVM 6u22 and if it has something to do with the > embedded Jetty instance or is it because of the Tuscany runtime... > > > On Wed, Feb 2, 2011 at 1:03 PM, Simon Laws <simonsl...@googlemail.com>wrote: > >> > >> > >> > 2) Samples issue >> > >> > Yes, rather well timed those problems with the samples reported by users >> - >> > these are exactly the same problem that I ran into. We should fix these >> > since the samples are exactly the place that new users are going to >> start >> > and there is nothing worse than a binary runtime being "broken out of >> the >> > box" when they try to run "hello world". >> > >> > >> > Yours, Mike. >> > >> >> +1 I think we should at least discuss what we want to achieve and >> identify those changes we can make now to fix the pressing issues. >> I've opened TUSCANY-3825 to track. >> >> Simon >> >> -- >> Apache Tuscany committer: tuscany.apache.org >> Co-author of a book about Tuscany and SCA: tuscanyinaction.com >> > >
