Re: crash on Alpha box

[Roger M. Burkhart]

> However, we think we have a good handle on what the problem is. There is an
> implicit assumption in the code that an 'unsigned' (ie. unsigned int) is
> sufficient to hold the bits of a pointer. This is broken on the alpha, and
> it is a severe problem since the code makes this assumption all over the
> source tree.

Yes, it's broken, but should be fixable.  It's probably not as widespread
as you indicate, since the main assumption we make (in the docs under
"global portability assumptions") is that an unsigned can fit into an id
(or any other pointer), not the other way around.  But the new code for
zones in defobj violates this assumption, and needs to be fixed.

To support alphas we will need to incorporate testing there into our release
process.  I'm not aware of Swarm having ever run there so far, but running
on new 64-bit architectures like alpha is important for Swarm and has been
intended from the beginning.

> I am just a bit wary of the time all this porting may be taking away
> from actual research --- but then again, the alphas here are the main
> staff machines. Thinking what should we do next ..... 

We need to do a comprehensive review of core Swarm libraries to make sure
we don't have any more of these portability bugs lurking, and figure out
how best to fix them.  Glen has indicated that we've made arrangements for
the needed machine access, so it's just a matter of fitting this in
with everything else.  I'd like to see us start making a few experiments
as our first step, and then we can let the support list know when we think
alphas can be a supported platform.

-Roger