Improving Java for Linux

[Per Bothner]

[Resent bounced message from Per Bothner <bothner at cygnus.com>, Daniel]

Paul Michael Reilly <pmr at copa.pajato.com> says:
>I'm anxious to embrace Cygnus' Java/GCC efforts, but so far these have
>been vaporware as near as I can tell.

I guess technically it is vaporware, since nothing has been released ...

Nelson Minar <nelson at media.mit.edu> writes:
> I think they're farther along than that - it looks to me from the
> outside that Cygnus is quite committed to improving the Java situation
> for Unix. Between gcc and kaffe they have quite a base, and I have
> heard rumours they've got some code working. But we haven't heard much
> from them, so it's hard to know how far along they are.

At this point we have an almost fully-functional "jc1".  (That is the
name for the new Gcc front-end.)  A few bytecodes (such as fmod,
mointorenter, etc) are not yet implemented.  The main thing missing
is exception handling.

We are able to take non-trivial Java packages, compile them, link
them with Kaffe (statically or dynamically), run them, and debug
them with gdb.

So far, we only handle javac-compiled .class files, but we have
recently started work on compiling from source-code.

I don't know when a prototype will be available to the Net.
I would like ot make it soon, but there are technical problems,
as well as licensing and political issue to resolve.  At this point,
there is little the Net could help with the compiler side.

The main place where the Net can help is with clean-room re-implementation
of the Java classes.  This is a huge undertaking, and the main place where
Sun keeps moving the target.  We will initially concentrate on the
core classes, and hope the Net will work on awt, jdbc, swing, etc etc.

Paul Michael Reilly <pmr at copa.pajato.com writes:
> Also, Kaffe is less than thrilling in that its quality is lacking and
> Wilkinson more often than not throws up license purity as an obstacle
> to accepting hacking help.
and:
> My purview says that the right thing to do is to fix the problems I'm aware
> of with Kaffe, distribute it GPLd early and often, merging in Tim's
> distributions and accepting patches from all comers.

>From our point of view, the right thing is defintely *not* to be more
casual about about accepting patches and license purity.  We are most
definitely concerned about getting sued, or our customers (with deeper
pockets) being sued.  I think the right thing to do for Net contributions
is to require copyright assignments or disclaimers, like the FSF
requires for any significant contribution to Gcc or Emacs.  In fact it
is even more important:  While I think it is very unlikely that someone
would sue over Gcc or Emacs, there is a very real concern (justified or
not) that Sun might sue, unless we have a strong paper trail of where
the code came from, and can document in court that we did not copy
Sun's code.

	--Per Bothner
Cygnus Solutions     bothner at cygnus.com     http://www.cygnus.com/~bothner