[slime-devel] swank+sbcl and cpu utilization

[Chris Goellner]

I'm trying to get a remote SWANK+SBCL setup going and I've hit a snag.
Everything works and I can connect to the running SBCL. However, when
I disconnect the SBCL process drives the CPU to 100%.

I've tried this on OS X 10.5, Solaris 2.8, and Linux Ubuntu with SBCL
1.0.20. I've checked that the slime on the remote and local ends are
the same version which is the latest from the CVS. My guess is I'm
doing something obviously wrong since it looks like a lot of people
use this setup without a problem. So, few that I could only find one
Google entry that seemed on point but the recommended change is to a
function that isn't in my version of slime.

my .emacs file looks like:

(add-to-list 'load-path "path/to/slime")
(setq inferior-lisp-program "/usr/local/bin/sbcl")
(require 'slime)
(slime-setup)

and my .sbclrc file looks like:

(require 'asdf)
(require 'asdf-install)

I've got another file called swank.lisp that starts up my remote side
sbcl+swank process:

(asdf:oos 'asdf:load-op 'swank)
(setf swank:*communication-style* :fd-handler)
(setf swank:*use-dedicated-output-stream* nil)
(swank:create-server :dont-close t)

I call this file as 'sbcl --load swank.lisp' and the whole system
works great. Until I disconnect from the sbcl instance. Then it spikes
the CPU and will spin until I stop it.

I've got this working in CLISP but I've been using SBCL and I would
like to have both working. Any suggestions?