Re: gcj and jikes

[ebb9 at email dot byu dot edu]

> >>>>> "Per" == Per Bothner <per at bothner dot com> writes:
> 
> Per> I think we need to look ahead to JDK 1.5, with its semi-announced
> Per> support for generics and more.  Iff the Jikes maintainers are
> Per> committed to supporting future Java languages changes, at least the
> Per> ones we can see coming doing down the pike, then it may be
> Per> a good thing to do.  Though it might still be reasonable to wait
> Per> until the first JDK 1.5 alpha is out.
> 
> That makes sense.  Here's what Eric had to say about jikes and
> generics last time we talked about this:
> 
> http://gcc.gnu.org/ml/java/2003-02/msg00391.html

I have been one of the main developers on the jikes projects lately, but have 
been pretty inactive on it during the past four months thanks to school.  (But 
on the bright side, I finally graduated today!)  As I have followed both 
projects, I do think that it would be beneficial to join the jikes front end 
with gcj.  But as noticed, the biggest problem would be licensing issues.  And 
I don't know if I'm the person to get the ball rolling on resolving those sort 
of issues.

I'm interested in seeing how this conversation develops.  And I hope that I 
will still have time to piddle around with compilers, even though I will not 
have the free schedule of a student any more.

> 
> 
> I'm bringing this up since I'm dissatisfied with gcj's progress.  We
> aren't making great progress fixing bugs -- the gcj .java front end
> still has many known problems, and my impression is that most large
> packages require a tweak or two to build with gcj.  I really don't see
> how something as big as generics is going to be implemented, given our
> current apparent manpower.
> 
> Basically, I think we're behind the curve on the front end.  Even if
> jikes never implements generics, using it as the front end would be a
> noticeable improvement for users.  On top of that, jikes is simply a
> better base to build on.  I find the code cleaner and easier to
> follow.

Even jikes is behind on implementing generics - I have been the only person 
working on it, and my status was the same as stated before.  I have generic 
classes working, but not generic methods, and jikes still does not 
generate/parse the added Signature attribute in .class files.  It doesn't help 
that the public draft of JSR 14 (the original proposal to implement generics is 
not type-safe, and Sun engineers Neal Gafter and Gilad Bracha, in charge of 
javac and the 3rd edition JLS, respectively, have not let me know what changes 
are in the pipe to rectify the problems).

> 
> Not that jikes couldn't use some improvement.  E.g., there are a lot
> of down-casts that should be changed to use dynamic_cast<>, for
> safety.

I'm sure jikes would welcome patches.  I've personally been trying to fix some 
of these bugs as I come across them.

> 
> Tom

Eric Blake