Thanks for your message. I am running a current kernel now, and gdb works
better, but still has problems single-stepping - it often seems to die with
a sigtrap, or to continue rather than breaking at the next instruction,
etc..

On Fri, 13 Nov 1998, Michael L. Hitch wrote:

>   The problem isn't gdb - it's the kernel.  The R4x00 cpu caches are
> the problem.  The way a program gets started and run under gdb results
> in some cache-coherency problems that haven't been worked out yet.
> I was finally able to get some brute-force cache flushing to work in
> -current, but I haven't had the time to figure out a cleaner way to
> do it.