[slime-devel] Re: SLIME, OpenMCL and aborting a process

[Gary Byers]

On Wed, 16 Jun 2004, Luke Gorrie wrote:

> Gary King <gwking at cs.umass.edu> writes:
>
> > Suppose I type
> >
> > (loop repeat 10000 do (sleep 0.5) (princ "."))
> >
>
> > Also, no periods are printing. Where are they going? Are they
> > filling up my computer like those little holes from a three hold
> > punch? What if it runs out of space? (Sorry, long night).
>
> There is some buffering and I'm not sure of the exact semantics on
> OpenMCL. I think it'll flush before filling the code with . characters
> though :-)
>
> -Luke

A few times a second, a thread wakes up and tries to call FORCE-OUTPUT
on every stream on the list CCL::*AUTO-FLUSH-STREAMS*.  FORCE-OUTPUT
isn't otherwise called automatically until the buffer's full (i.e.,
streams aren't line-buffered). The default buffer size is 2K bytes,
so the code above should produce a bufferful of output every 1K seconds
or so.

It's more awkward than it could be to register/deregister a stream for
this handling (OTOH, it's not supposed to be the sort of thing that
happens all that often, either.)  At one point, I think that the
swank backend for OpenMCL did this, but it doesn't seem to do so
currently.