Re: [cp-patches] RFC: implementation of getPackage for the vmreference

[Nicolas Geoffray]

Jeroen Frijters wrote:

> Nicolas Geoffray wrote:
>  
>
> > That's what I thought too in the first place, but the
> > ClassLoader.getPackage calls the parent loader's getPackage
> > method. If the parent is null, then it calls 
> > VMClassLoader.getPackage().
> >
> > If VMClassLoader.getPackage had to add the package given as
> > parameter, then all packages (system + applications) would be
> > defined in VMClassLoader.definedPackages.
> >
> > I hope I'm clear enough.
> >
> > The suggestion you gave for getPackages (having all bootstrap packages
> > in VMClassLoader.definedPackages at startup) could remove the 
> > additional VMClassLoader.getVMPackage. Maybe that's a better solution?
> >    
>
>  
>
> > (Sun does it)
> >    
>
> Ah, I see what you mean now. In my VM I also have a full list of the
> packages defined by the boot class loader, so there my approach does
> work (I just tested it). I can't think of any other way to solve this,
> there should be a native method that returns a String[] with the native
> package names.
>
>  
Why should it be native? It might be the same work for all VMs.
Maybe we could add a static initializer to VMClassLoader
that would look for java.boot.class.path property and fill the packages
(eg the directories) it finds in the definedPackages hashmap.

> Regards,
> Jeroen
>  

Nicolas