>> That is because the BytecodeLoaders happens to search the
>> SyspathJavaLoader before searching the referent class loaders. It's a
>> bug.
>
>No, I would'nt say it is a bug. The problem is that
>the guilty class is loadable
>both by MultiClassLoader and probably from classpath.
>In those cases it is better if the user dis-entangle that, because
>, yes we can with good-will, change the order, but the only safe
>solution is to avoid that at all.
Right.
>Without knowing the specific up to the details, I will not
>change the actual order.
Ok, maybe simply reversing the order is too naive, but I can't imagine a
situation where it make sense to try and load a superclass by the
SyspathJavaLoader when we in BytecodeLoader.makeClass() know that the
superclass was initially loaded from one of the referent loaders.
regards,
finn
|