On Tue, Feb 22, 2005 at 09:34:51AM +0100, Andreas Hallmann wrote:
> Resulting sets are working perfectly on a turbo sparc based system, 
> where as sh is dumping core
> on a micro sparc II based SS4-110.

There was a sh bug recently fixed in:
  $NetBSD: expand.c,v 1.68 2005/02/14 20:46:26 dsl Exp $ 
This was triggered by random stack content, but I haven't seen it core
dump.

I've been running supersparc and v8 optimized builds on my Krups and LX
for years, and never seen any problems like you describe.

Martin