define-obsolete-variable-alias problem

define-obsolete-variable-alias problem

[Glenn Morris]

Given this sequence:

(progn
  (setq foo 1)
  (define-obsolete-variable-alias 'foo 'bar)
  (defvar bar 2))

I'd have expected the end result to be bar = foo = 1.
However, the actual result is bar = foo = 2.

This is a problem for the case where I have:

  (setq old-variable-name 1)

in my ~/.emacs, and a package uses:

  (define-obsolete-variable-name 'old-variable-name 'new-variable-name)
  (defcustom new-variable-name 2)

because it means that my customization is ignored. The only solution
seems to be to ;;;###autoload the define-obsolete-variable-name
statement, and AFAICS by extension every such statement in Emacs,
where the variable is user option that might be set in ~.emacs. Is
this really how it is supposed to work?

Re: define-obsolete-variable-alias problem

[Nick Roberts]

 > Given this sequence:
 > 
 > (progn
 >   (setq foo 1)
 >   (define-obsolete-variable-alias 'foo 'bar)
 >   (defvar bar 2))
 > 
 > I'd have expected the end result to be bar = foo = 1.
 > However, the actual result is bar = foo = 2.

Isn't this really a problem with defvaralias?  Perhaps we could test there if
new-alias is bound and if so assign the value to base-variable.

-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: define-obsolete-variable-alias problem

[Stefan Monnier]

>> Given this sequence:
>> 
>> (progn
>> (setq foo 1)
>> (define-obsolete-variable-alias 'foo 'bar)
>> (defvar bar 2))
>> 
>> I'd have expected the end result to be bar = foo = 1.
>> However, the actual result is bar = foo = 2.

> Isn't this really a problem with defvaralias?

Yes.

> Perhaps we could test there if new-alias is bound and if so assign the
> value to base-variable.

Indeed except if base-variable already exist, we can't do that.


        Stefan

Re: define-obsolete-variable-alias problem

[Glenn Morris]

Stefan Monnier wrote:

>> Isn't this really a problem with defvaralias?
>
> Yes.
>
>> Perhaps we could test there if new-alias is bound and if so assign the
>> value to base-variable.
>
> Indeed except if base-variable already exist, we can't do that.

OK, I think I've installed that.

However if it is to work for the case I posted,
define-obsolete-variable-alias must be placed _before_ an associated
defcustom. A quick glance in some random .el files shows that often it
is not. I know I usually write them the other way round. Time to move
a bunch of them?

Re: define-obsolete-variable-alias problem

[Glenn Morris]

Glenn Morris wrote:

>> Indeed except if base-variable already exist, we can't do that.
[...]
> define-obsolete-variable-alias must be placed _before_ an associated
> defcustom.

Actually, thinking about it more, I don't see why it can't be done if
base-variable already exists?

Re: define-obsolete-variable-alias problem

[Stefan Monnier]

>>> Indeed except if base-variable already exist, we can't do that.
> [...]
>> define-obsolete-variable-alias must be placed _before_ an associated
>> defcustom.

> Actually, thinking about it more, I don't see why it can't be done if
> base-variable already exists?

You can call defvaralias after the base_var has been set.
That's perfectly fine.  But if both base_var and new_alias are set when
you call defvaralias, then defvaralias can't know which of the two
values to use.


        Stefan

Re: define-obsolete-variable-alias problem

[Glenn Morris]

Stefan Monnier wrote:

> But if both base_var and new_alias are set when you call
> defvaralias, then defvaralias can't know which of the two values to
> use.

I'm probably fixated on define-obsolete-variable-alias.
Always using the value of new_alias would be the right thing there
AFAICS, and I can't see a downside to always doing so.

Re: define-obsolete-variable-alias problem

[Stefan Monnier]

>> But if both base_var and new_alias are set when you call
>> defvaralias, then defvaralias can't know which of the two values to use.

> I'm probably fixated on define-obsolete-variable-alias.
> Always using the value of new_alias would be the right thing there
> AFAICS,

No.


        Stefan


PS: BTW "new_alias" = "obsolete_name".

Re: define-obsolete-variable-alias problem

[Glenn Morris]

Stefan Monnier wrote:

> No.

OK, I'll give up on understanding this, and go away and move a bunch
of define-obsolete-variable-alias statements. (There are < 100; I'm
surprised).