"Yar Hwee Boon" <hboon at motionobj.com> wrote:

> I posted a similar error before. Not sure if it is related, but I had been  
> using the 3.6 VM with a 3.7 image. When I switched to the 3.7 VM  
> (windows), the problem seem to go away.
It is absolutely related. It's nothing to do with particular VMs as
such, nor cpu architecture (well....) but all to do with the OS and
it's approach to allocating memory.

If the chunk allocated for Object Space end above 2Gb then plausible
OOPS will appear to be negative numbers when treated as ints. The VM
code uses ints all over the place - as it has for eight plus years now
- and unless you have some bizarre cpu/compiler combination the
comparison of two OOPS may well go wrong. It is a strange coincidence
that the only place where this appears to have an effect is in a
routine involved in loading and image segment.

We _will_ fix it sometime. Just not today. If I were being paid to work
on such issues it would have been fixed already but I'm currently
occupied with other issues and so it has to sit on my todo list.

tim
--
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
The steady state of disks is full.  - Ken Thompson