Re: patch for 0.12 (was: Re: java.lang.System changes)

[Steven Augart]

Mark Wielaard wrote:
> Hi,
>
> On Fri, 2004-11-12 at 17:45, Jeroen Frijters wrote:
>
> > Steven Augart wrote:
> >
> > > And, to that end, I have the following small patch that makes
> > > the current Classpath CVS head work with Jikes RVM.  If someone
> > > else approves it, I'll commit it.
> >
> > It works fine for me.
>
> Thanks to you both for working on this issue.
> This works for now, but I hope you won't mind if we rip it out again for
> 0.13. One Thing I don't like about this is that it creates a new compile
> time constant which makes it harder to share an GNU Classpath glibj.zip
> class byte code between runtimes installed on the same system. I would
> like to see something like the following for breaking up the
> initialization of java.lang.System into two phases in the future:
>
>  boolean explicitly_initialize = VMSystem.explicitlyInitializeStatics();
>  if (explicitly_initialize)
>    {
>      initLoadLibrary();
>      initProperties();
>    }
>
> That way a runtime could introduce its own VMSystem, but still share the
> system installed glibj.zip.

Your suggestion is far superior to the configuration constant.  Let's
rip it out as soon as 0.12 comes out.  I'll even prep the patch.

> Both David and Jeroen made a couple of good suggestions for improving
> the runtime/platform interface. We should work on that more in the
> future. It will also help with the libgcj and kaffe integration since
> the bits that haven't been shared completely yet are precisely these
> kind of system/bootstrap classes.

This sounds great to me.  I will be glad to have better control over
some of the fragility of the Jikes RVM boot process, and I'm sure the
other runtime integrators will feel the same way.