Buildable only?

Fri May 6 21:37:00 GMT 2005

>>>>> "Norman" == Norman Hendrich <hendrich@informatik.uni-hamburg.de> writes:

Norman> From my point of view, it is very hard to understand that you keep
Norman> adding fancy compiler options (like gcj-dbtool or gcjx), while even
Norman> pretty simple applications don't work at all - as soon as they use
Norman> AWT or Swing. I admit that most of the java.lang, java.math, java.io,
Norman> java.net, and java.util stuff is pretty good already, and from what
Norman> I've heard some of the javax.* things are good, too.
Norman> Just out of curiousity: during the Cygnus times, gcj seemed
Norman> just like a small and fun project to me. But now that Redhat
Norman> is quite a big player, and gcj the biggest partof gcc (using
Norman> either LOC or build-time as a metric), isn't there any way
Norman> that Redhat can spend some of the IPO money on a few
Norman> man-months of full-time AWT/Swing development?
A few notes on this:
First, Red Hat did fund a substantial amount of AWT and Swing work in
this last year. Look through the ChangeLog for changes by Tom
Fitzsimmons, Graydon Hoare, Kim Ho, Olga Rodimina, and David Jee
(apologies if I missed somebody... but I don't think I did).
Most of those people have left the gcj project for one reason or
another. That means that development by Red Hat on this has slowed
down. On the other hand, we've seen more work from non-RH community
members on Swing.
The biggest priority around here has been getting certain applications
to run and get them into FC4, in particular eclipse, OO.o, ant,
tomcat, and their many dependencies. We've also done a lot of work to
get jonas running, but that doesn't yet build against the free
libraries and so isn't suitable for inclusion in FC.
Also, gcjx is just a personal project of mine. Red Hat has given me
resources to do it (the computer), but I work on it in my spare time.
I hope to change this, but I just wanted it to be clear that at the
moment gcjx doesn't reflect RH's development priorities.
Focusing attention on these applications seems pretty useful because
it translates into real results pretty easily. E.g., Eclipse is cool
and useful, and it was a doable project to make it work. And, the
technology needed for Eclipse (class loaders and BC, basically)
applies directly to jonas and other things.
One thing that would help to make a case for AWT/Swing work would be a
set of target applications. We've been thinking that a Mozilla plugin
looks like a cool target, and that would seem to force the issue for
AWT (and security). But for Swing, I really don't know the landscape.
Personally, I think if we can get the mozilla plugin and gcjx working
in this next year, we'll be doing pretty good.
Tom