[Openmcl-devel] Re: Backtrace from SLIME when loading openmcl with an image from SLIME

[Marco Baringer]

Gary Byers <gb at clozure.com> writes:

> About all that the backtrace says is that FORCE-OUTPUT is called (I
> assume that it's called on *ERROR-OUTPUT*, but that's mostly a guess),
> gets an error (somewhere in a STREAM-FORCE-OUTPUT method), enters a
> break loop, calls FORCE-OUTPUT, gets an error ... after about 3 or 4
> recursive errors, it gives up and drops into the kernel debugger.
>
> I don't use SLIME and don't really how or when it sets up
> *ERROR-OUTPUT*; in a non-SLIME environment, *ERROR-OUTPUT* is usually
> a SYNONYM-STREAM to *TERMINAL-IO* (which is -usually- created very
> soon after a saved image starts up.)

slime, in order to get the lisp's output to go to an emacs buffer,
changes the followings bindings when starting up: *standard-output*,
*error-output*, *trace-output*, *standard-input*, *debug-io*,
*query-io*, *terminal-io*. so i think (this is a best a wild guess)
that the image is being saved with *error-output* bound to a broadcast
stream to an emacs buffer which doesn't exist when the image is
reloaded (or somethnig weird along those lines). try undoing the
stream redicetion before creating the image:

(progn
  (swank::revert-global-io-redirection)
  (ccl::save-application ...))

-- 
-Marco
Ring the bells that still can ring.
Forget the perfect offering.
There is a crack in everything.
That's how the light gets in.
	-Leonard Cohen