> The problem is that if your machine crashes and you don't have another 
> machine attached to it, ddb is your only hope. cgd thinks this isn't
> an issue, but for me, that is the usual case by a big margin.

Run stable software/hardware that only crashes Infrequently on your
server machine.  If you can't do that, you're probably trying to run
the wrong OS on your server machine.

Connect your client machines to its serial ports (you may need to do
something tricky if you need more serial ports).

When your server machine crashes, which should be very infrequent,
you're left with a core.

In general, for most _crashes_ (i.e. not debugging anomalous behaviour
seen in a live system), i find core dumps + gdb to be _far_ better
than ddb as an analysis tool, for a variety of reasons.

On live kernels, KGDB is far better (if it can be used).

Just about the only time i can think of people regularly wanting to
use DDB is if they _have_ to debug something 'live' on a laptop while
travelling (not a situation you really want to be in 8-), or if they
don't have a copmuter they can use as a server.

I think i'd say that the people who do the most serious development
tend to have servers they can attach serial lines to and run gdb on.



cgd