I've been having a lot of trouble with I/O to SBCL through slime recently. This is on PPC, using either :sigio or :fd-handler communication style, and dedicated-output-stream either t or nil.
It's easy for me to reproduce (Slime as of fifteen minutes ago, SBCL 0.8.11.14):
CL-USER> (inspect nil)
The object is a SYMBOL. 0. Name: "NIL" 1. Package: #<PACKAGE "COMMON-LISP"> 2. Value: NIL 3. Function: "unbound" 4. Plist: NIL
0
0 0 00 0 0 90
<C-c C-c, select the second to bottom-most reset> ; Evaluation aborted
The second to bottom-most one (usually an abort, but not TOPLEVEL) seems the most reliable to clear it up.
What's worse is when (as happens on SBCL/PPC sometimes) you get a recursive debug when the backtrace fails, it drops you into a TTY debugger in the Slime REPL. Since you can't interact with it, and it's difficult to get another SLDB debugger to come up, this is kind of bad.
I'm not sure when this started.
Also, is there any way to get the slime version anymore? The date no longer appears at the top of a repl, which is kind of annoying since it was easy to identify what era of slime you were working with.
-bcd