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.
See e.g. following root path to a project that I don't have opened in my IDE: NbModuleProject[MasterFileObject@d43b6f[space/nenik/work/nb/core/swing/plaf]]: private static org.netbeans.modules.javadoc.search.JavadocRegistry org.netbeans.modules.javadoc.search.JavadocRegistry.INSTANCE-> org.netbeans.modules.javadoc.search.JavadocRegistry@d66dfb-results-> java.util.HashSet@1ea4bd8-map-> java.util.HashMap@191382f-table-> [Ljava.util.HashMap$Entry;@18f3497-[156]-> java.util.HashMap$Entry@17555fa-next-> java.util.HashMap$Entry@d1920d-key-> org.netbeans.modules.apisupport.project.queries.JavadocForBinaryImpl$1@1cbbf42-this$0-> org.netbeans.modules.apisupport.project.queries.JavadocForBinaryImpl@230529-project-> org.netbeans.modules.apisupport.project.NbModuleProject@1ccb4f6
JavadocRegistry keeps what JavadocForBinaryQuery returns and it listens to changes of aquired results. It is a valid use case IMO. As I can read in your comment the project is referenced by JFBQ.Result of apisupport/project and in addition to that there is no impl of ChangeEvent support. Reassigning to apisupport/project for evaluation.
Could easily use a static nested class; the project field is unused anyway. No plans to implement change listening for this case. Too much work.
This looks like a trivial and rally safe fix with no performance implications (i.e. releasing the project won't slow down javadoc access once the result is computed, right?). If it proves it really is enough to free the projects from memory, I would vote for fixing this even for 6.0, as the memory usage impact can be huge (megabytes of heap in many use cases).
Checking in JavadocForBinaryImpl.java; /shared/data/ccvs/repository/apisupport/project/src/org/netbeans/modules/apisupport/project/queries/JavadocForBinaryImpl.java,v <-- JavadocForBinaryImpl.java new revision: 1.15; previous revision: 1.14 done