🚀 go-pugleaf

In article <9106252056.AA09412@cs14b.cs.aukuni.ac.nz> mike-w@cs.aukuni.ac.nz writes:

   >>> "Kyle" == kyle@uunet.uu.net writes:

   Kyle> I'll make the next release of VM check to see if the value is nil
   Kyle> before rebinding.  If the value of inhibit-local-variables is
   Kyle> non-nil, then VM will not change it.

     I've encountered an interesting bug in conjunction with this.  It goes
     something like this ...

     * In my ~/.emacs, I 
	 (set-default 'inhibit-local-variables 'query-other)

     * In my vm-mode-hook, I
	 (make-local-variable 'inhibit-local-variables)
	 (setq inhibit-local-variables 'always)

       Now, whenever I call vm, I discover that the *global* value of
     inhibit-local-variables is set to 't !!!  I don''t understand how this
     could be happening -- the only place where it is set to 't is the
     let-binding in vm and vm-gobble-crash-box!!

This is the infamous specbind bug.  If you exit the scope of a binding of
a buffer-local variable with a different current buffer from the current
buffer at the time of entering the scope of the binding, kaboom!

When a variable is bound (either as the parameter of a function or with
"let") the old value is stored on a stack.  When the scope of the binding
is exited, the equivalent of (set 'variable old-value) is done.
Obviously, this can cause a problem with buffer-local variables if the
current buffer was changed within the scope of the binding.

This isn't a complete explanation.  Feel free to inspect the code of
"specbind" and friends for the gory details.

-- 
Joe Wells <jbw@uswest.com>