Newsgroups: comp.parallel.pvm
From: cierniak@cs.rochester.edu (Michal Cierniak)
Subject: FIN|ACK
Organization: Computer Science Department, University of Rochester
Date: 03 Aug 1994 16:40:42 GMT
Message-ID: <CIERNIAK.94Aug3124045@moose.cs.rochester.edu>

I have the following problem:  We have different workstations in our
department.  For my experiments, I wanted to use two models of Sun
workstations: SPARCstation LX and SPARCstation 1.  They all run SunOS
4.1.3.

The problem is that the same program with the same input and the same
number of computers in the virtual machine runs fine on LX's, but
breaks on sparc1's.  I'm getting messages like this one:

[t80040000] netinput() FIN|ACK from tern
[t80040000]  hd_dump() ref 1 t80000 n "tern" a "" ar "SUN4"
[t80040000]            lo "" so "" dx "" ep "" bx "" wd "" sp 1000
[t80040000]            sa 192.5.53.46:1595 mtu 4096 f 0x0 e 0 txq 0
[t80040000]            tx 20 rx 14 rtt 0.021800


I don't really know what this message means, but I found the following
sentence in the User's Guide: "If the pvmd panics, (e.g. on a trapped
segment violation) it tries to send a packet with FIN and ACK bits set
to every peer before it exits."  After reading that I assumed that
this strange behavior is caused by a bug in my program, which causes
something like "segment violation".  But I couldn't find any problem
yet, besides I thought the LX's and sparc1's are binary compatible
(please correct me if I'm wrong).  Has anyone experienced a similar
problem?  Do you know what I'm doing wrong?

Thanks,
Michal

