Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!mly@prep.ai.mit.edu
From: mly@prep.ai.mit.edu
Newsgroups: net.emacs
Subject: autoload confusion
Message-ID: <994@mit-eddie.UUCP>
Date: Fri, 17-Jan-86 18:49:54 EST
Article-I.D.: mit-eddi.994
Posted: Fri Jan 17 18:49:54 1986
Date-Received: Sun, 19-Jan-86 03:44:46 EST
Sender: daemon@mit-eddi.UUCP
Organization: MIT, Cambridge, MA
Lines: 42

From: mly@prep.ai.mit.edu (Richard Mlynarik)
   From: uoregon!meyer@mit-eddie (meyer)
   Newsgroups: net.emacs
   Date: Fri, 10-Jan-86 11:07:00 EST


   This may not be a *real* bug, but I'm confused about
   the philosphy of "autoload". Specifically, I would like
   to load my own scheme-mode (lisp-mode, prolog-mode,...),
   but since scheme-mode (for example) is already defined, (i.e., 
   !EQ (XSYMBOL (function)->function, Qunbound), where "function"
   is scheme-mode), one cannot define a new scheme-mode. That is,
   autoload returns nil and does nothing if the symbol (function name,
   e.g., 'scheme-mode) is already bound.  Apparently, this will be 
   true for any function name loaded out of loaddefs.el.

For starters, the definition of "autoload" in Emacs17 (the current
version) icontains the following fragment:

  /* If function is defined and not as an autoload, don't override */
  if (!EQ (XSYMBOL (function)->function, Qunbound)
      && !(XTYPE (XSYMBOL (function)->function) == Lisp_Cons
    && EQ (XCONS (XSYMBOL (function)->function)->car, Qautoload)))
    return Qnil;

so what you are trying to do would work if you were running a more
up-to-date Emacs.
Thanks for the [belated] bug report in any case...

   [..]
   Is there a way to get around this (e.g., unbind a symbol;
   I noticed there isn't much in the way of oblist handling
   functions), or should autoload always set up the autoload
   (like Gosling's)?

There is a useful subset of the relevant common-lisp functions:
makunbound, fmakunbound, boundp, fboundp, intern, intern-soft (ie find-symbol)

The behaviour of -always- doing a redefinition when `autoload' is
called is flawed in my opinion for a number of reasons, which I could
go into if you're interested.  I'd prefer to avoid making a religious
discussion out of this.