On Tue, Nov 12, 2002 at 12:22:19PM -0800, Frank Cusack wrote:

> This consistently occurs after < 1 minute of running (same error).  Is
> this likely to be a hardware problem, or a tcpdump 64-bitness problem?
> I would guess the latter.

Tcpdump every now and then shows this behaviour on all strict-alignement archs
when seeing certain traffic.

We've fixed it for several cases already, but you seem to have found a new 
one ;-)

For debugging this, it would be very conveninent to get the unbuffered output
of tcpdump right before dying on the sparc64 machine and the same output
(not dying) from another (i.e. i386) machine.

Martin