|
|
Rutger Ovidius writes:
> Friday, April 21, 2006, 2:33:41 AM, you wrote:
>
> RM> -----BEGIN PGP SIGNED MESSAGE-----
> RM> The only saving grace is that GCJ is able to create
> RM> .class files (but not object files, since it uses the
> RM> same driver as the rest of the GCC) and "gij" is able
> RM> to run interpreted programmes. So at least "libgcj.jar"
> RM> is located properly when the toolchain is relocated.
>
> Just to add to the list of win32 problems:
>
> System.loadLibrary() does not work without the hack from here:
>
> http://gcc.gnu.org/ml/java/2006-02/msg00066.html
>
> Class.forName() does not work because java/lang/natClass.cc ::forName
> uses stacktrace.cc ::GetCallingClass which can not work with SJLJ
> exceptions. (this is a regression; Class.forName() worked in previous
> versions).
True. I hae a new version of Class.forName() that will help, but the
real solution is to fix stacktraces on Windows. It can be done, we
just need a keen gcj-on-Windows maintainer.
Andrew.
|
|