--- Begin Message ---
Hi again,
> The statement of better choice is here maybe a bit out of context. gij,
> kaffe or sablevm are in the case of using the vm to BUILD a package for
> DEBIAN the (normally) only choice. But that has nothing to do with
> the quality or fitness for a given task of jamvm at all.
>
> That is just because of the debian policy which states that a package
> must be buildable from source on all debian arches. As jamvm is
> currently not available on all arches the consequence is that if
> I use jamvm for BUILDING my package on i386 and a user / other developer
> tries to build it from source on arches without jamvm it will fail and
> therefore we get a release critical bug for it.
>
Yes, sorry, I understand this, and I know I'm going to get flamed for
my reply now. I've seen rumours that Debian may drop support for the
more "obscure" architectures. If this goes ahead I'll only have to do
a version for AMD64 and IA64 :)
The difficulty (once 64-bit support is done) with porting JamVM to
these architectures is the calling convention. Other VMs (e.g.
SableVM) rely on libffi to do this portably. I prefer to instead use
hand-coded routines for each architecture/platform (using assembler),
to make calling JNI methods as efficient as possible -- JNI is
notoriously inefficient as it is.
Would you mind forwarding this to the list? I replied via gmane last
time, as I'm not subscribed.
Thanks,
Rob.
> I myself are really impressed of jamvm ! Keep up the good work !
>
> Regards,
>
> Wolfgang
>
>
--- End Message ---