Re: DataInputStream changes...

[Etienne M. Gagnon]

Hi Brian,

Brian Jones wrote:
> 
> Would it be useful for us to accomodate the SableVM as we do Japhar?
> 

Yes it would:-)

----
Opinion about the current Classpath build:

(hoping to start a discussion of the matter; my opinion is NOT final;-)

I do not understand why Classpath needs to know about a specific VM in
order to compile.  Unless I am wrong, my understanding is that Classpath
is a VM independent class library. 

The way I would think Classpath should compile is the following:
1- It should be able to compile all the native code, without any prior
knowledge of the target VM.  The JNI has been specially designed so that
would be possible.  In Sablepath, I provided a "vanilla" jni.h file
which declares function prototypes based on the official JNI
specification (as published in the book "The Java Native Interface" ISBN
0-201-32577-2).
2- Some classes should be implemented by the vm, we agree.  But,
Classpath could simply provide skeleton classes in /vm that provide
definitions for these classes with empty method bodies.
3- The build process would compile all the classes, but would only
"install" non-vm classes.

It should be the responsibility of each VM to add its own classes, then
compile and package them.  It shouldn't be the responsibility of
Classpath.

Now, there's nothing wrong in providing "functional" vm classes, which
would minimize the work of a vm implementor, but I do not think that
this is a sufficient reason to make the Classpath build VM dependent.

Just a thought.

Etienne
-- 
----------------------------------------------------------------------
Etienne M. Gagnon, M.Sc.                     e-mail: address@hidden
Author of SableCC:                 http://www.sable.mcgill.ca/sablecc/
and SableVM:                                   http://www.sablevm.org/
----------------------------------------------------------------------