Re: bulk with i386 chroot on amd64? xen?

[David Holland <dholland-pkgtech%netbsd.org@localhost>]

On Wed, Aug 10, 2022 at 11:32:08AM +0200, is%bnhb484.de@localhost wrote:
 > > >   2) building packages under xen and running them on not xen, or the
 > > >   other way around.  In particular, i think this might be emacs and PAE
 > > >   vs non-PAE.
 > > 
 > > Yes. Our XEN3PAE kernels offer a bigger address space to the userspace
 > [...]
 > 
 > Probably disregard my message, as David answered that the problem has been
 > fixed.

Well, I thought it was. The issue with the location of the stack
(which definitely changes if the virtual address space size changes,
because the stack lives at the top of whatever it is) got fixed.

Not entirely sure what the state is regarding the dumped data segment,
though. If that comes out in what is sometimes kernel space, it'll
still fail. I have a vague recollection that was something was done
about this too, but I'm not sure what or when. That or it wasn't ever
actually a problem and just a conjecture about what turned out to be
the other problem. I would expect it to appear starting at the
traditional heap location, that is, the _end of the pre-dump temacs,
which should not cause trouble.

It's easy to check if you have a binary, anyway; just run nm -n |
tail, or failing that, readelf -l.

I have an emacs21 binary here that was built on an amd64 kernel with
i386 userland, and while nm claims "no symbols", readelf says the
largest load address is 0x081bce8c, which should not pose any
problems.

-- 
David A. Holland
dholland%netbsd.org@localhost