This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
I tried: $ ANT_OPTS="-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/home/jarda/dump.hprof" ant clean build commit-validation on revision 23cea3a6189a and the build fails on output memory exception due to lack of permgen space. The dump shows at least two com.sun.istack.tools.ParallelWorldClassLoaders holding huge amount of classes. If I am right, this is some utility need to pre-process web services classes. The build was executing some target in in ide.ergonomics. It is clear that the com.sun.istack.tools.ParallelWorldClassLoader and their classes were no longer necessary. Please make sure the build task that needs com.sun.istack.tools.ParallelWorldClassLoaders properly cleans after itself (or it is executed in a forked VM). The dump is now available at http://beetle.czech.sun.com/~jt97948/dump-with-clean-build-commit-validation.hprof
Result of access to URL : You don't have permission to access /~jt97948/dump-with-clean-build-commit-validation.hprof on this server.
It is easy to reproduce the problem locally and the issue contains steps to do it. Anyway I also fix the access rights.
Today morning the server was up, now it is down. Probably because of migrating to new Oracle new infrastructure. If you have a place where I shall upload the file, let me know. Otherwise please generate new one yourself.
Please send this file to me via e-mail. I have no time to rebuild NB .
I don't think I can send a file of this size like this thru email: 165M dump-with-clean-build-commit-validation.hprof Even after gzip it is still too big, I think: 39M dump-with-clean-build-commit-validation.hprof.gz I'll try to attach it.
Too large. But I managed to upload it to http://netbeans.org/projects/performance/downloads/download/dump-with-clean-build-commit-validation.hprof.gz
>It is easy to reproduce the problem locally and the issue contains steps to do it. I'm not able to reproduce it according your suggestion. Trying to follow your scenario I get the OutOfMemoryError after downloadbinaries task. The build procedure starts to building nb.cluster.ruby modules: build-one-cluster: [echo] Building nb.cluster.ruby modules OutOfMemoryError arises right after this. So it has no relation to WS. You are right that com.sun.istack.tools.ParallelWorldClassLoader relates to WS. I have found usage of JAXB xjc task in the module websvc.saas.api. The ParallelWorldClassLoader class is included in the jar with xjc task. There are no modifications in this module for a long time. So it cannot be a reason for suddenly appeared OOME. This is just one module. XJC task is external task and I cannot change it. The problem is deeper than just one module issue. I believe this is a common issue with build procedure. Default setting for heap size is not enough to build NB. There is a recommendation for ant options: http://wiki.netbeans.org/WorkingWithNetBeansSources These settings works fine for me. I'm able to do ant clean build commit-validation with success. So I can't reproduce it with recommended settings for Java Heap size and there is another problem with default Java Heap size not related to WS.