Parent ClassLoaders are not being properly passed into the groovy scripts.

The following test case shows the problem:

public void testWithParentClassPathPropagatesWithNoParentContainer()throws IOException {
File testCompJar = new File(System.getProperty("testcomp.jar"));
URLClassLoader classLoader = new URLClassLoader(new URL[]

{testCompJar.toURL()}

, this.getClass().getClassLoader());
Class testComp = null;

try

{ testComp = classLoader.loadClass("TestComp"); }

catch (ClassNotFoundException ex)

{ fail("Unable to load test component from the jar using a url classloader"); }

Reader script = new StringReader(
""
+ " pico = builder.container(parent:parent)

{\n" + " component(class:\"TestComp\")\n" + " }

"
+ "");

PicoContainer pico = buildContainer(new GroovyContainerBuilder(script, classLoader), null, null);
Object testCompInstance = pico.getComponentInstance(testComp);
assertEquals(testCompInstance.getClass().getName(), testComp.getName());

}

It fails claiming that the TestComp class is not found and the classloader stack trace does not show the URL classloader.

I was under the impression that the groovy container builder's classpath SHOULD get passed into the script somehow, otherwise, groovy scripts are always loaded under the basic system classpath.

I am working on a solution, but I wanted to make sure that I understood that the GroovyContainerBuilder classpath passed in SHOULD be the classpath used for building containers. (Aside from dynamic classpaths, of course!)

Thanks for taking a look at it. If people don't think I'm misunderstanding the use of the API, I'll check in a fix shortly.