Hi all!

I've finally gotten my system working OK again (I actually
found a friend working at AT&T service organization in Sweden
who had access to a crashed CONNER disk, from which I could have
the controller board which was allright :-) Gee, that's luck!)
I've reexamined the GVP 24-bit access to odd-aligned 16-bit
mem addresses and found that the paging I thought was necessary
to hit the bug indeed wasn't.  It was just a case of a lower
probability when the system didn't page.  With that in mind,
the only SW fix available is to add -mnobitfield to the cc1
flags in the "specs" file and recompile binaries that fail.
For me it's still only "as" and "emacs", so it's really not
much work.  The problem is reported to GVP, which promptly
responded saying that the PAL upgrade wasn't for the combo,
but they will continue investigate my report.  I'll let you
know as soon as I hear anything.

Niklas

------------------------------------------------------------------------------