Re: What is the EndOfLife of GCJ-4.2.x or GCJX-4.2.x when appears Harmony?

[Dalibor Topic <robilad at kaffe dot org>]

On Tue, Jun 20, 2006 at 04:45:58PM +0200, P.O. Gaillard wrote:
> Anny Blackyew wrote:
> >Apache's Harmony's DRLVM's Jitrino can kill GCJ-4.2.x.
> 
> I hope not, GCJ does AOT (ahead of time) compiling that is more predictable 
> (and hence more testable) than JIT. That makes is more useful for soft 
> real-time applications.
> 
> GCJ is a good piece of software and while other implementations might take 
> a bigger share of the Java "market share" in the future, GCJ can be 
> expected to stay around for good reasons.
> 
> I have also noted the GCJ team's good track record regarding interaction 
> with other similar projects :
>   - they are using GNU classpath and contributing to it instead of keeping 
> libgcj separare.
>   - they are using implementation techniques that make Java play along very 
> well  with C, C++ etc. This allows me to use GDB, Oprofile etc.
>   - they are going to reuse the Eclipse compiler front-end for GCJ thus 
> reducing the amount of work they do on a part of the compiler that has 
> little added value (i.e. a working front-end is all we want)
> 
> I therefore perceive GCJ as a project with a convincing roadmap. This does 
> not mean that it will have more users than Kaffe, Jitrino or the Sun JDK. 
> But GCJ really looks like it is here to stay and quite a few people will be 
> happy to use it.
> 

These days gcj has about as many regular users as Kaffe, and quote a bit
more than the Sun JDK on Debian, given the numbers on popcon[1]. Since
there is no Apache Harmony release yet, it's hard to guess how much
impact it will have. Given that there is no lack of packaged good
runtimes (JamVM, Cacao, IKVM, etc.) in Debian, and none of them has ever
'killed another off', it's very, very unlikely that Harmony will have
the effect Anny writes about.

I recall people predicting Mono to make gcj/Kaffe/* obsolete a few years
ago, during the GNOME managed runtime flamefest. That didn't happen,
either. gcj is very well maintained, has a good foundation to
build upon in gcc, has a stronly committed FSF behind it, so it's unlikely 
to go away any time soon. If useful new code appears in some other
project, it's more likely to be merged in/reused (see ecj) to improve gcj, 
rather than fought with. 

As far as JITs go, I'd expect LLVM to be a more interesting choice than 
JVM-specific solutions. Kaffe has had its own JITs for almost 10 years. :) 
Writing JITs (and JVMs) is sorta 'easy' ... maintaining them is the really 
hard part. In that respect, insular solutions tend to lose out to
reusable & reused components in the long term.

cheers,
dalibor topic

[1] 
http://people.debian.org/~igloo/popcon-graphs/index.php?packages=sun-java5-jre%2Ckaffe%2Cjava-gcj-compat&show_installed=on&show_vote=on&want_percent=on&want_legend=on&beenhere=1

> 
>    these were my 2 cents about a really nice and useful piece of software,
> 
>   P.O. Gaillard
> 
> 
> 
>