CL> I've reverted this check-in because with the change applied, the CL> error messages are incorrect:

Ok, so the problem is rooted in the way the args list is manipulated in the subcommand handlers. As it stands now (without the patch), some commands give a good error message, while others dump a stack trace if the domain doesn't exist:

This is because destroy adds "bogus" into the args list at index 0, but domid does not. I've attached a new patch that checks for this condition, and removes the "bogus" entry from the list if the subcommand added it. This makes the call to handle_xend_error() work in both situations:

# xm destroy foo Error: Domain 'foo' not found when running 'xm destroy' # xm domid foo Error: Domain 'foo' not found when running 'xm domid'

Is that an acceptable solution for now? Perhaps a cleanup of the inconsistent subcommand handler behavior is in order. I can do that when I start work on the remaining interface changes.