Re: [Ikvm-developers] Javassist

[Jeroen F. <je...@su...>]

Mark Wielaard wrote:
> On Tue, 2006-08-01 at 22:54 +0200, Jeroen Frijters wrote:
> > Kaushal Patel wrote:
> > > Caused by: javassist.NotFoundException: java.lang.Object
> > >         at javassist.ClassPool.get (ClassPool.java:295)
> >=20
> > Javassist tries to load the class files for the core classes (in the
> > above case it is looking for java/lang/Object.class). Since=20
> IKVM doesn't
> > have any Java class files for the core classes, this fails.
> >=20
> > I don't know how to work around this, you could try asking the
> > Javassist/Hivemind/Tapestry developers if there is a way to=20
> explicitly
> > point it to an rt.jar. It may be as simple as pointing the CLASSPATH
> > environment variable to an rt.jar.
>=20
> FYI this is very similar to a libgcj bug report by Anthony=20
> Green (CCed):
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D15474
> As he points out there are some applications which seem to=20
> depend on the
> class byte code of the classes from the bootstrapclassloader available
> somehow. Bryce and I argued that this is not a sane thing for=20
> a program to depend on (like ikvm, gcj has a completely different=20
> execution model for its core classes). But if this pops up more often
> maybe some way to provide a byte code version of these classes through

> getResource() might still be needed. Note that all instances of this
> have been some "sanity check" in a testsuite, not real application
code.
> I haven't yet looked at what javassist does.

I've since implemented a workaround for this. getResource will now
synthesize a .class file for the corresponding class (using reflection).

For the code see
http://ikvm.cvs.sourceforge.net/ikvm/ikvm/classpath/ikvm/internal/stubge
n/StubGenerator.java?view=3Dmarkup

I've tested this with Javassist and it appears to work (at least for a
simple demo that they include).

Regards,
Jeroen