Re: server.el test failures

[Eli Zaretskii <eliz@gnu.org>]

> Date: Thu, 2 Mar 2023 14:15:49 -0800
> Cc: Eli Zaretskii <eliz@gnu.org>, eggert@cs.ucla.edu, emacs-devel@gnu.org
> From: Jim Porter <jporterbugs@gmail.com>
> 
> On 3/1/2023 12:22 AM, Robert Pluim wrote:
> >>>>>> On Tue, 28 Feb 2023 11:41:41 -0800, Jim Porter <jporterbugs@gmail.com> said:
> >      Jim> +  ;; When running in a batch session, we can't create a GUI frame.
> > 
> > Itʼs not that we canʼt, itʼs that we shouldnʼt. If we install a SIGIO
> > handler and run the test with DISPLAY set, a GUI frame appears
> > briefly.
> 
> Well, since we don't have a SIGIO handler (and merely adding one 
> wouldn't be a useful change, it seems), I'd say "can't" applies for now 
> at least. :)
> 
> But it would probably make sense to expand this comment to explain 
> *why*, since it's not as informative as it could be...
> 
> >      Jim> +  (when noninteractive
> >      Jim> +    (push '(display . nil) parameters))
> >      Jim> +
> > 
> > I think that might be better off using `alist-get' + `setf' to avoid
> > having two `display' elements in the parameters.
> 
> Would that be safe? Since 'parameters' is an argument, we don't actually 
> know how it was created or if modifying in-place would be ok. If it were 
> just a regular quoted list, using 'setf' could cause problems.

If this aspect could be an issue, perhaps it is better to implement
the change without touching 'parameters' at all?  It isn't hard, it
just might make the change less elegant and slightly larger.