What version of Emacs still requires '*' in defcustom docs?

What version of Emacs still requires '*' in defcustom docs?

[Glenn Morris]

The practice of giving the docs of defcustoms a leading '*' doesn't seem
to want to die out. For those of you doing this, what version of Emacs
are you trying to support that requires this?

Eg GNU Emacs has not needed this since at least 21.1 (see user-variable-p).

RE: What version of Emacs still requires '*' in defcustom docs?

[Drew Adams]

> The practice of giving the docs of defcustoms a leading '*' 
> doesn't seem to want to die out. For those of you doing this,
> what version of Emacs are you trying to support that requires this?
> 
> Eg GNU Emacs has not needed this since at least 21.1 (see 
> user-variable-p).

Speaking for myself, Emacs 20 (to support, e.g., `set-variable').

Re: What version of Emacs still requires '*' in defcustom docs?

[Glenn Morris]

Glenn Morris wrote:

> Eg GNU Emacs has not needed this since at least 21.1 (see user-variable-p).

PS 21.1 is over 10 years old.

RE: What version of Emacs still requires '*' in defcustom docs?

[Drew Adams]

> > Eg GNU Emacs has not needed this since at least 21.1 (see 
> > user-variable-p).

Yes, and it was 21.1 that _introduced_ the bug (#10475), which is still there.
Two steps forward, one step back.  Half-baked.
 
> PS 21.1 is over 10 years old.

So are some Emacs users. ;-)

Does it bother you that `C-h v' should have to accommodate a leading `*' for a
user variable, even though `defcustom' no longer _requires_ a leading `*'?  All
it has to do is ignore (strip) it.  What's the big deal?  The fix is trivial,
and is provided in the bug report.

Besides, `C-h v' needs to do the same thing _anyway_ for a user variable defined
this way:

(defvar toto 42 "*This IS a user var. Always has been.")

`toto' is a user variable, as defined and recognized by `user-variable-p' (which
you cited) and as recognized by `set-variable'.

Circulez ; il n'y a rien a voir.

Re: What version of Emacs still requires '*' in defcustom docs?

[Chong Yidong]

Glenn Morris <rgm@gnu.org> writes:

> The practice of giving the docs of defcustoms a leading '*' doesn't
> seem to want to die out. For those of you doing this, what version of
> Emacs are you trying to support that requires this?
>
> Eg GNU Emacs has not needed this since at least 21.1 (see
> user-variable-p).

FWIW, I intend to convert user-variable-p into an alias for
custom-variable-p after the release.

RE: What version of Emacs still requires '*' in defcustom docs?

[Drew Adams]

> > The practice of giving the docs of defcustoms a leading '*' doesn't
> > seem to want to die out. For those of you doing this, what 
> > version of Emacs are you trying to support that requires this?
> >
> > Eg GNU Emacs has not needed this since at least 21.1 (see
> > user-variable-p).
> 
> FWIW, I intend to convert user-variable-p into an alias for
> custom-variable-p after the release.

So `set-variable' will be broken for user variables defined using `defvar' with
`*'.  Or do you plan to modify `custom-variable-p' so it recognizes those as
well?  In which case we will lose the distinction between the two kinds of user
variable.

In addition to providing no such details, you gave no reason for the proposed
change.  What for?  Why is it important to make such a change?

Re: What version of Emacs still requires '*' in defcustom docs?

[Stefan Monnier]

> Why is it important to make such a change?

Because complexity is a problem in itself and as maintainers our main
task is to try and keep it in check, so the code doesn't crumble under
its own weight.


        Stefan

RE: What version of Emacs still requires '*' in defcustom docs?

[Drew Adams]

> > Why is it important to make such a change?
> 
> Because complexity is a problem in itself and as maintainers our main
> task is to try and keep it in check, so the code doesn't crumble under
> its own weight.

So do _nothing_.  No added complexity for maintainers.
No loss of features for users.

This code and behavior for users have remained unchanged since at least Emacs
21.1, as some have been eager to point out.

Re: What version of Emacs still requires '*' in defcustom docs?

[Stefan Monnier]

>> > Why is it important to make such a change?
>> Because complexity is a problem in itself and as maintainers our main
>> task is to try and keep it in check, so the code doesn't crumble under
>> its own weight.
> So do _nothing_.  No added complexity for maintainers.

The complexity is there already.  It was introduced a long time ago
by defcustom.  We now know that it is not worth the effort and it's
better to merge the two concepts, hence removing some complexity.


        Stefan

Re: What version of Emacs still requires '*' in defcustom docs?

[Stephen J. Turnbull]

Stefan Monnier writes:
 > > Why is it important to make such a change?
 > 
 > Because complexity is a problem in itself and as maintainers our main
 > task is to try and keep it in check, so the code doesn't crumble under
 > its own weight.

The problem is that this is complexity for the maintainers, and it's
minor.

Making the change of `defvar ... "*..."' to defcustom is imposing
maintenance costs on 3rd-party package maintainers, and potential
additional complexity (in the custom interface) on users.  The custom
hierarchies for Gnus and VM are awful in my experience.  In reaction,
in some cases I've made a distinction between things that a very few
users might want to change, and defcustoms.

YMMV, but I think if somebody proposes porting this change to XEmacs,
I'll oppose it at least at first.