[21.2]: M-x customize-group offers non-groups for customisation

[21.2]: M-x customize-group offers non-groups for customisation

[Marshall, Simon]

If you do M-x customize-group RET then you can enter, e.g.,
auto-compression-mode and a customisation buffer is offered up.  Problem
is, auto-compression-mode isn't a group and so the customisation buffer
is empty/confusing.

The cause is in this in customize-group:

(completing-read "Customize group: (default emacs) "
     obarray
     (lambda (symbol)
       (or (get symbol 'custom-loads)
           (get symbol 'custom-group)))
     t)

where auto-compression-mode has a non-nil custom-loads property (its
custom-group property is nil).

How hard is it to make custom only offer genuine groups?  Presumably too
hard.

Perhaps something meaningful can be done after customize-group does the
custom-load-symbol, e.g., say, "oh, it's not a group, but here's a
customisation buffer containing the list of groups to which it belongs"
or "oh, it's not a group, it's an X and here's the appropriate
customisation buffer to customise it".

Simon.

Re: [21.2]: M-x customize-group offers non-groups for customisation

[Per Abrahamsen]

"Marshall, Simon" <simon.marshall@misys.com> writes:

> How hard is it to make custom only offer genuine groups?  Presumably too
> hard.

You'd have to add the information when you generate cus-load.  I don't
think it would be hard (but I don't volunteer), the question is
whether it is worth it.

> Perhaps something meaningful can be done after customize-group does the
> custom-load-symbol, e.g., say, "oh, it's not a group, but here's a
> customisation buffer containing the list of groups to which it belongs"
> or "oh, it's not a group, it's an X and here's the appropriate
> customisation buffer to customise it".

I prefer the second option, with the caveat that it can be both a
variable and a face, simultaneously.  It isn't even a rare case.

Re: [21.2]: M-x customize-group offers non-groups for customisation

[Stefan Monnier]

>>>>> "Marshall," == Marshall, Simon <simon.marshall@misys.com> writes:
> where auto-compression-mode has a non-nil custom-loads property (its
> custom-group property is nil).
> How hard is it to make custom only offer genuine groups?  Presumably too
> hard.

Couldn't we decide that the `custom-loads' property should only
contain a list of elements if the object is a group, otherwise
it should just contain a single string (or symbol).

After all, why should a `defcustom' require loading of more
than a single file/feature ?


	Stefan

Re: [21.2]: M-x customize-group offers non-groups for customisation

[Richard Stallman]

    > How hard is it to make custom only offer genuine groups?  Presumably too
    > hard.

    You'd have to add the information when you generate cus-load.  I don't
    think it would be hard (but I don't volunteer), the question is
    whether it is worth it.

This is definitely the right solution--the other alternative is not
coherent behavior.  If it is not hard, this is what we should do.

Can you give any suggestions?

Re: [21.2]: M-x customize-group offers non-groups for customisation

[Per Abrahamsen]

Richard Stallman <rms@gnu.org> writes:

>     > How hard is it to make custom only offer genuine groups?  Presumably too
>     > hard.
>
>     You'd have to add the information when you generate cus-load.  I don't
>     think it would be hard (but I don't volunteer), the question is
>     whether it is worth it.
>
> This is definitely the right solution--the other alternative is not
> coherent behavior.  If it is not hard, this is what we should do.
>
> Can you give any suggestions?

None, except to look at the code that generates cus-load, and see if
an obvious solution present itself.  I don't have time to do look into
it myself at the present time.