built-in variable `font-use-system-font' not bound

built-in variable `font-use-system-font' not bound

[Eli Zaretskii]

Building Emacs on MS-Windows produces this message at loadup time:

    Loading cus-start...
    Note, built-in variable `font-use-system-font' not bound

Re: built-in variable `font-use-system-font' not bound

[Jan Djärv]

Eli Zaretskii skrev:
> Building Emacs on MS-Windows produces this message at loadup time:
> 
>     Loading cus-start...
>     Note, built-in variable `font-use-system-font' not bound
> 

What does that even mean?  Why does it complain on that variable and not any 
of the other X-specific variables in cus-start?

	Jan D.

Re: built-in variable `font-use-system-font' not bound

[Jason Rumney]

Jan Djärv wrote:
> Eli Zaretskii skrev:
>> Building Emacs on MS-Windows produces this message at loadup time:
>>
>>     Loading cus-start...
>>     Note, built-in variable `font-use-system-font' not bound
>>
>
> What does that even mean?  Why does it complain on that variable and 
> not any of the other X-specific variables in cus-start?

The other variables start with "x-", as caught by the code about a page 
below the entry for font-use-system-font starting with this comment:

      ;; Don't complain about missing variables which are
      ;; irrelevant to this platform.

Re: built-in variable `font-use-system-font' not bound

[Jan Djärv]

Jason Rumney skrev 2009-11-20 18.19:
> Jan Djärv wrote:
>> Eli Zaretskii skrev:
>>> Building Emacs on MS-Windows produces this message at loadup time:
>>>
>>> Loading cus-start...
>>> Note, built-in variable `font-use-system-font' not bound
>>>
>>
>> What does that even mean? Why does it complain on that variable and
>> not any of the other X-specific variables in cus-start?
>
> The other variables start with "x-", as caught by the code about a page
> below the entry for font-use-system-font starting with this comment:
>
> ;; Don't complain about missing variables which are
> ;; irrelevant to this platform.

Aha.  I just thought that we should avoid x-, since there are many x- in use 
in the w32 port.  But maybe that was functions?  I'll fix this.

	Jan D.

Re: built-in variable `font-use-system-font' not bound

[Stefan Monnier]

> Aha.  I just thought that we should avoid x-, since there are many x- in use
> in the w32 port.  But maybe that was functions?  I'll fix this.

Regardless of how things currently are (we currently have a big mess in
this respect), we should follow the following convention:
- if it's a feature specific to X11, W32, NS, ... then use a name that
  starts with "x-", "w32-", "ns-", ...
- if it's a feature that makes sense everywhere, then don't use such
  a prefix (even if it's only implemented for some of the backends).


        Stefan

Re: built-in variable `font-use-system-font' not bound

[Jason Rumney]

Stefan Monnier wrote:
> Regardless of how things currently are (we currently have a big mess in
> this respect), we should follow the following convention:
> - if it's a feature specific to X11, W32, NS, ... then use a name that
>   starts with "x-", "w32-", "ns-", ...
> - if it's a feature that makes sense everywhere, then don't use such
>   a prefix (even if it's only implemented for some of the backends).
>   

If we follow that convention, I think we need another prefix for 
functions/variables that are common across window systems that can be 
used in the test to avoid warnings/errors in cus-start.el for tty only 
builds.   Currently we usually use x- for these, since historically most 
such features started on X and were ported to other window systems later.

Re: built-in variable `font-use-system-font' not bound

[Stefan Monnier]

>> Regardless of how things currently are (we currently have a big mess in
>> this respect), we should follow the following convention:
>> - if it's a feature specific to X11, W32, NS, ... then use a name that
>> starts with "x-", "w32-", "ns-", ...
>> - if it's a feature that makes sense everywhere, then don't use such
>> a prefix (even if it's only implemented for some of the backends).

> If we follow that convention, I think we need another prefix for
> functions/variables that are common across window systems that can be used
> in the test to avoid warnings/errors in cus-start.el for tty only builds.
> Currently we usually use x- for these, since historically most such features
> started on X and were ported to other window systems later.

So you mean features that only make sense for GUIs but not for
text-based terminals like ttys?  I can't think of any such feature
which couldn't be treated similarly to "a feature that makes sense
everywhere but hasn't yet been implemented in all terminal types".


        Stefan