Re: VMClassLoader

[Robert Lougher]

On Sun, 15 Aug 2004 12:07:48 +0200, Mark Wielaard <address@hidden> wrote:
> Hi,
> 
> On Thu, 2004-08-12 at 19:18, Tom Tromey wrote:
> > I notice that VMClassLoader still has this:
> >
> >    * <strong>For backward compatibility, this just ignores the protection
> >    * domain; that is the wrong behavior, and you should directly implement
> >    * this method natively if you can.</strong>
> >
> > I suggest we pick a flag day to switch this.  In particular I think we
> > ought to simply remove the old form of defineClass() and only have the
> > form that accepts a ProtectionDomain.
> 
> I agree. And a since the old form of defineClass() isn't even used from
> the rest of GNU Claspath I think we can remove it immediatly. Runtimes
> will probably have provided their own VMClassLoader implementation
> already.
> 

Hi All,

FYI, JamVM implements the version with the ProtectionDomain natively,
as suggested in the comments.  I also implement getPrimitiveClass
natively with a type char, rather than relying on the Java method that
converts it to a string (e.g. D->double).  This method is also marked
as destined to be removed.

Would it be a good idea to tidy up both at the same time?  Of course,
the class would still need modifications by the runtime because
loadClass/getResource, etc. are stubs which must be implemented.

Rob.

P.S.  I'm going by the Classpath-0.10 release snapshot, and not CVS.

> If nobody objects I want to commit the attached patch next week.
> 
> 2004-08-14  Mark Wielaard  <address@hidden>
> 
>         * vm/reference/java/lang/VMClassLoader.java (defineClass): Removed
>         version that didn't take a ProtectionDomain.
> 
> Cheers,
> 
> Mark