Re: Misleading `window-system' docstring

[Eli Zaretskii]

> From: Štěpán Němec <address@hidden>
> Cc: Julien Danjou <address@hidden>,  address@hidden
> Date: Wed, 20 Oct 2010 15:05:02 +0200
> 
> Eli Zaretskii <address@hidden> writes:
> 
> >> From: Julien Danjou <address@hidden>
> >> Cc: address@hidden
> >> Date: Tue, 19 Oct 2010 12:32:33 +0200
> >> 
> >> +         `(:eval (unless window-system
> >> +                   ,(propertize "-%-" 'help-echo help-echo)))))
> >
> > Using window-system is not right here, because the DOS port defines a
> > non-nil value, but uses a text terminal.
> 
> Could you please fix the docstring of `window-system' to reflect that?
> It seems to be widely used that (incorrect, as you point out) way

Actually, there are only a handful of places that use it incorrectly.
Most of the uses compare its value with `x', `w32', etc., which is the
right usage.

> its docstring is pretty clear (and thus incorrect) on the matter:
> 
>   The value is nil if the selected frame is on a text-only-terminal.

I don't mind spelling out the exception, but I actually would like to
add there that using it as a predicate is deprecated, and point to
display-*-p predicates as TRT.  Stefan and Chong, any objections?