Re: added optional --with-jvm=

[Mark Wielaard]

Hi,

On Sun, 2004-03-28 at 21:08, Etienne Gagnon wrote:
> > I don't object to a --with-vm configure option in principle, it could be
> > useful to build classpath for a specific environment. But I do object to
> > encoding specific knowledge about particular VMs in our tree. We did
> > this in the past and it wasn't a success. (See the kaffe, orp and japhar
> > bits and pieces that have been rotting in the tree because we don't
> > maintain them anymore and/or the vm makers lost interest.)
>
> For now, the option simply picks the right default for zipping or not
> the output.  This is not *specific VM code*.  I do object to your objection.

Noted. I still see hardcoded values for one specific runtime (sablevm)
in the patch so I qualify it as VM specific.

> Otherwise, Classpath should simply provide default configuration options
> that work with *all* VMs.  So, we should change it to not zipping so that
> it also work with SableVM.

That might be an option worth discussing, although I don't think that is
a good default. Another one is for SableVM to finally include the patch
Archie made for zip file support posted here:
http://mail.gnu.org/archive/html/classpath/2003-12/msg00066.html

> > Please revert this and discuss the general idea and how it should work
> > on the list first. The argument for the --with-vm option should probably
> > be a directory location or file that comes with the vm
> > sources/installation describing specific optimizations for that
> > environment.
> 
> Please look at cvs diff before argumenting on a non-issue.

I looked at the cvs diff of course. Otherwise I wouldn't have said that
I don't agree with this patch going in and that it should be reverted
till it has been properly discussed on-list.

Cheers,

Mark

signature.asc
Description: This is a digitally signed message part