Return-Path: mao
Received: by postgres.Berkeley.EDU (5.61/1.29)
	id AA20899; Mon, 16 Nov 92 09:12:09 -0800
Message-Id: <9211161712.AA20899@postgres.Berkeley.EDU>
From: <mao@postgres.berkeley.edu>
Subject: Re: alignment errors on decstation5000 running ultrix 4.2
To: postgres@postgres.berkeley.edu
Sender: pg_adm@postgres.berkeley.edu
In-Reply-To: Your message of Mon, 16 Nov 92 07:15:01 PST.
             <9211161515.AA20384@postgres.Berkeley.EDU> 
Date: Mon, 16 Nov 92 09:18:30 PST

In message <9211161515.AA20384@postgres.Berkeley.EDU> you write:

> postgres 3.1 on a decstation5000/200 running ultrix 4.2 is puking
> CONSTANTLY with unaligned access errors. Is there some -D option I can
> recompile with to get rid of the problem? Here is a sample error
> message; judging from the pc's logged, we are in a tight loop somewhere.
> Needless to say, the performance of the backend on this platform is less
> than abysmal.

3.1 had a number of known bugs that could cause this behavior; that
release is no longer supported.  you should install 4.0.1.

4.0.1 also disables the address fixups done by the ultrix kernel, forcing
a coredump if an unaligned access happens.  we don't know of any examples
of this in a 4.0.1 system, but if an unaligned access happens, dumping
core is certainly preferable to having the kernel scramble pointers.

					mike