Try to remove debugger module and keep applet module. The exception in comment is printed.

[jglick] As far as I can tell, the ModuleInstaller just does not make the right calls into ModuleDescription at all. Attached is output of startup after removing debugger as Ian suggested, and adding
-Dorg.openide.modules.ModuleDescription.VERBOSE=true. As you can see, it is not right at all:


- as far as I know, no code in impl ever checks to see if a module can be uninstalled via the ModuleNode.Item property, so illegal uninstalls are permitted (installs OK); *all* module installs or unin
stalls runnable from these nodes should run the ModuleDescription checks first (e.g. rOFRI). There is another bug to this effect elsewhere.

- ModuleDescription.resolveOrdering is being called on everything, even modules which are disabled (according to property sheet and XML and so on); this call assumes that all dependencies are satisfie
d, which in this case they are but only because even uninstalled modules are being included in the list

- ModuleDescription.resolveOrderingForRealInstall, which ought to be the main call used by impl code when starting up, is not called until later, and then with an empty list of modules

- all ModuleDescription objects are created twice, which should not be harmful but is still suspicious

- modules loaded from manifest do not appear to be checked for legality (dependencies) before being installed, though I could be wrong; also not treated as installed for later purposes (deps when inst
alling from node)

Summary: impl should create 1 instance of ModuleDescription for those modules, and *only* those modules, which are listed in XML as enabled; call rOFRI (empty-set, xml-enabled) and restore the result;
 look for modules to install (e.g. JAR in modules/ but no mention in XML), and call rOFRI (last-result-of-rOFRI, newly-installable) and install the result. Property toggles should call dependsOnModule
 as needed.

- see solution...->