Re: edebug: Symbol's function definition is void: nil

[Nick Roberts]

 >     I can't reproduce it now.  It think it occurred when I was stepping
 >     through a function past a symbol that evaluated to the empty string.  I
 >     enabled debug-on-error and seemed to get a backtrace at one point with
 >     just an entry for nil (but I didn't find it till later because it
 >     didn't pop up).  '
 > 
 > It sure sounds like a bug, but it is hard to guess the cause from this
 > info.

It seems to be related to using vm (7.19).

1) If I set debug-on-error to t
2) Type M-x vm
3) Turn tool-bar-mode on

I get:

Debugger entered--Lisp error: (void-function nil)
  (nil)

(I think it can also happen without setting debug-on-error to t)

I think this is due to vm-add-undo-boundaries being added to
post-command-hook (I've said before vm does hairy things):

(defun vm-add-undo-boundaries ()
  (save-excursion
    (mapatoms (function
               (lambda (b)
                 (setq b (get-buffer (symbol-name b)))
                 (if b
                     (progn
                       (set-buffer b)
                       (vm-undo-boundary)))))
              vm-buffers-needing-undo-boundaries)
    (fillarray vm-buffers-needing-undo-boundaries 0)))

  I suspect that there is a bug in Emacs and it should report a more normal
backtrace, but this involves Lisp concepts that I don't fully understand.

Nick