Re: gnus shouldn't be making general-purpose variables buffer-local

[Katsumi Yamaoka <yamaoka@jpl.org>]

>>>>> Reiner Steib wrote:
> However, I wonder if the more general patch suggested by David Engster
> is better.  Does anyone see a problem with it?

> Ami, does David's patch solve your problem?

> --- a/lisp/gnus-sum.el
> +++ b/lisp/gnus-sum.el
> @@ -3831,6 +3831,7 @@ This function is intended to be used in
>        (and (consp elem)			; Has to be a cons.
>  	   (consp (cdr elem))		; The cdr has to be a list.
>  	   (symbolp (car elem))		; Has to be a symbol in there.
> +	   (boundp (car elem))		; Has to be already bound
>  	   (not (memq (car elem) vars))
>  	   (ignore-errors		; So we set it.
>  	     (push (car elem) vars)

>>>>> In <b4mfxkms7yl.fsf@jpl.org> Katsumi Yamaoka wrote:
> Cool!  But I agree not to use it. :)

But I found no evil with that patch so far.  Variables like gnus-*
globally bound need to get to be buffer-local but it's harmless.
Moreover, those parameters have been to be set as (VAR VAL), not
(VAR . VAL).  OTOH, parameters used as just parameters, e.g.
`timestamp', should not need to be bound; those are set in the
`gnus-parameters' variable or the newsrc database.  Only one anxiety
is the case that a user or some program binds such a variable, but
the fault will lie with the user or what should be complained will
be the program.  So, I've installed David Engster's patch with a
comment: http://article.gmane.org/gmane.emacs.gnus.commits/6091

Regards,