Hi all,
got a misterious problem here.
Made a fresh build of 2.0.1 with optimizations [ -mcpu=v8 
-mtune=supersparc -pipe ].
Fresh  means Fri 18. Feb 2005.

Now the curious:
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.
Moreover those 'Bad pAUX_base' are back which were so painful in the 
pre-releases.
I even saw some bad opcodes.

Hmmm, same system worked nearly stable with a stock 2.0 sparc release 
build I sucked from netbsd.org for two month now. This was the only 
non-sparcV8-build I used on this system in the last year and the most 
stable.

An expert question:
I know that both cpu's differ in the number of contexts and the size of 
caches.
But to the best of my knowledge they should run the same set of opcodes. 
--- V8.
OK there are some extensions on the turbo sparc side for IO handling 
etc. which
I don't think gcc supports.
Does gcc make some 'smart Alec' optimizations for the turbo sparc,
when it finds itself compiling on one?

Anyone else observing problems on SS4 with V8 code?

Don't know which way to go now.