Craig will correct me if I am wrong, but the idea is that in a server
environment, the server container *may* set the conxtext class loader
for you (at the webapp level, presumably). I am under the assumption
that this is done so that you can completly seal one webapp off from
another at a class access level.
We changed the code to check for the context class loader first,
assuming that would be better from a security perspective.
Scott Sanders
Jeff Turner wrote:
> Hi,
>
> I'm curious as to why the Digester code always tries to use the thread's
> classloader:
>
> ClassLoader ctxLoader = Thread.currentThread().getContextClassLoader();
> if (ctxLoader == null) {
> paramTypes[0] =
> Class.forName(paramType);
> } else {
> paramTypes[0]
> =
> ctxLoader.loadClass(paramType);
> }
>
> I'm pretty ignorant of classloaders in general. If anyone has pointers on this
> subject, perhaps suggesting why the above is needed, it would be appreciated.
>
>
> The reason I'm asking is because when I use Digester from within a
> servlet, it fails silently. I have (laboriously) traced the problem down to the
> fact that Digester uses the thread's classloader. If I remove the above checks,
> and just have Class.forName(), my code (as follows) works:
>
> Digester digester = new Digester();
> digester.push(this); // note: "this" is the servlet
> digester.setDebug(9);
> digester.setValidating(false);
> // set rules
>
> digester.addObjectCreate("services/service",
>"net.socialchange.bob.AbstractService"); // this fails silently
> digester.addSetProperties("services/service");
> digester.addSetNext("services/service", "addService",
>"net.socialchange.bob.Service"); // as does this
> // do parse
>
> My XML is:
>
> <services>
> <service name="Service 1" URL="http://localhost:8090/bob3/serv1.jsp"/>
> <service name="Service 2" URL="/serv2.jsp"/>
> </services>
>
> I'm using tc3.2.2b5, jdk1.3, no securitymanager, latest Digester from CVS.
>
> If this sounds like a bug (in tomcat or Digester), rather than my ignorance,
> I'll make a test case and file a proper bug report.
>
> thanks,
>
> --Jeff
