This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: Problems with Mohans GCJ 20030522 build
- From: Jeff Sturm <jsturm at one-point dot com>
- To: Andrew Haley <aph at redhat dot com>
- Cc: Øyvind Harboe <oyvind dot harboe at zylin dot com>, <java at gcc dot gnu dot org>, <gnustuff at thisiscool dot com>
- Date: Wed, 4 Jun 2003 11:45:33 -0400 (EDT)
- Subject: Re: Problems with Mohans GCJ 20030522 build
On Wed, 4 Jun 2003, Andrew Haley wrote:
> Øyvind Harboe writes:
> >
> > IMHO, the biggest impediment to GCJ being used more broadly
> > is the amount of reading and tweaking required to get
> > going.
>
> It depends on your perspective. gcj is pre-installed on most free
> operating systems, so there is no reading and tweaking required to get
> it to work. On the other hand, there are still compatibility problems
> due to missing components and bugs, and we must cure those before gcj
> will be the Java-compatible system of choice.
Definitely. But the existence of collections like RHUG points to
the difficulty of native-code compiling in particular.
We already have "gcj -C" as a drop-in for javac, and "gij" for java. Both
work quite well. But retrofitting any substantial java package, its build
system, etc. for AOT compilation is *hard*. And it won't get much easier
until we (users & developers) embrace something like what you've described
in:
http://gcc.gnu.org/ml/java/2003-01/msg00022.html
The two biggest hurdles I see are that a) -fno-assume-compiled isn't
finished yet and b) the duplicate class registration bug that prevents
ever loading two DSO's declaring the same class.
Jeff