Subj : Re: Memory leak with Rhino class loader?
To   : Tom <tdwang@yahoo.com>
From : Igor Bukanov <igor@icesoft.no>
Date : Thu Jan 23 2003 10:52 am

Tom wrote:
> Hello,
> 
> I really need some pointer on this so any help is greatly appreciated!
> 
> I have a servlet running in Tomcat:
> 
>     public void doGet(HttpServletRequest request, HttpServletResponse
> response) throws ServletException, IOException {
>         Context cx = Context.enter();
>         try {
>             Scriptable scope = cx.initStandardObjects(null);
>             String src = "function foo(i) {return i;}";
>             int count = 1;
>             while (true) {
>                 Function function = cx.compileFunction(scope, src,
> "foo", 1, null);
>                 Object[] args = new Object[1];
>                 args[0] = String.valueOf(count++);
>                 Object result = function.call(cx, scope, null, args);
>                 System.out.println("result=" + result.toString());
>                 if (count > 100) {
>                     break;
>                 }
>             }
>         } catch (Exception e) {
>             e.printStackTrace();
>         } finally {
>             cx.exit();
>         }
>     }
> 
> I found out that during each compilation Rhino creates a new
> DefiningClassLoader and uses it to load the compiled class.  So after
> 100 loops there will be 100 java classes (org.mozilla.gen.cxx) and 100
> new class loaders.  But these new classes never get garbage collected
> even after servlet exits.  So the memory just keeps piling up.  Is
> this a bug in Rhino or Tomcat, or am I missing something?

This is a bug in JVM that failed to collect loaded classes. Which 
version do you use? For details, see
http://bugzilla.mozilla.org/show_bug.cgi?id=64397

Try to use the interpreter mode which does not generate any Java classes 
that may even be faster if you always need to compile a function before 
calling it:

Context cx = Context.enter();
try {
   cx.setOptimizationLevel(-1);
....

Regards, Igor

.