Re: Runtime fixlet

[Mark Wielaard]

Hi,

On Wed, 2002-08-14 at 01:57, Tom Tromey wrote:
> This changes the reference VM's Runtime so that it doesn't create a
> new environment when ENV==null.  Is this something that ought to go
> in?  I'm putting it into libgcj (along with a patch to pass DIR to
> execInternal), but I don't really know the rules for changing this
> part of Classpath.
> 
> The rationale for this patch is that you want to preserve the case
> where ENV==null so that the native code can decide whether to preserve
> the current process' environment.

I think this change should go in. But be sure to update the execInternal
documentation and add an item to the NEWS file to alert VM implementors
that the contract has changed. (Classpath currently does not contain any
native code for java.lang.Runtime.) As long as we are still in alpha we
don't have to be to strict about the VM interface as long as we document
it clearly.

In the future we can split this in a common Runtime.java and a
VMRuntime.java part that contains only the VM interface. That would make
it more clear what impacts VM writers and what is the responsibility of
the Classpath hackers. I have a couple of ideas about making the VM
interface more clear which I would like to test out when I try
interfacing with Kaffe later this year.

Cheers,

Mark