I've just coded a little test to demonstrate one of the problems we had
with JTree. Unzip, type "ant test" and you should get an assertGC failure. You can also open the project in NetBeans IDE
if you want to work with the sources.

###@###.### adviced me to report this as a bug in our email conversation:

>      [junit] Testsuite: jtree.leak.JTreeDoesNotClearDrawCacheTest
>      [junit] private static Object jtree.leak.JTreeDoesNotClearDrawCacheTest.holder->
> javax.swing.JTree@a45536-expandedState->
>      [junit] java.util.Hashtable@129b922-table->
>      [junit] [Ljava.util.Hashtable$Entry;@e2e687-[10]->
>      [junit] java.util.Hashtable$Entry@135dd72-key->
>      [junit] javax.swing.tree.TreePath@3acf34-lastPathComponent->
>      [junit] jtree.leak.VisualizerNode@2b7db1)

> The log says that jtree.leak.JTreeDoesNotClearDrawCacheTest.holder should
> be GC'ed. Looking in the source code I'm seeing that this is just a private
> static field declared in the test and assigned to "tree" on setup. The
> question is what was the purpose of the field and why it should have been
> GC'ed on hiding the frame? Did I miss something?...

This field has a reference to JTree. Indeed we do not want our JTree to
disappear. However there is a node in the model, which has been properly
notified to the infrastructure as removed, yet it is still kept in memory by
some JTree's expanded state. As far as I can tell it is quite strange to have
expanded state to an object that is no longer in the model.

I'd like your team to fix this, as otherwise we cannot really write garbage
collect tests in NetBeans.