bug#12812: defsetf with third argument = t

bug#12812: defsetf with third argument = t

[Glenn Morris]

Package: emacs
Version: 24.2.50

In Emacs 24.2 and earlier, defsetf was documented [1] to accept a third
argument = t, with the same meaning as gv-define-simple-setter's
optional FIX-RETURN argument. This behaviour no longer works.

Here is an artificial example in `emacs -Q -l cl':

(defun foo (x)
  (car x))

(defun foo-set (x y)
  (setcar x y)
  nil)

(defsetf foo foo-set t)

(setq bar '(0 2 3))

(setf (foo bar) 1)   ; -> nil
bar                  ; -> (1 2 3)


The setf call correctly returns 1 in 24.2.


[1] (info "(cl)Customizing Setf")
   [...]
   As a special (non-Common-Lisp) extension, a third argument of `t'
   to `defsetf' says...

bug#12812: defsetf with third argument = t

[Stefan Monnier]

> In Emacs 24.2 and earlier, defsetf was documented [1] to accept a third
> argument = t, with the same meaning as gv-define-simple-setter's
> optional FIX-RETURN argument. This behaviour no longer works.

Indeed, thanks, fixed.


        Stefan

bug#12812: defsetf with third argument = t

[Glenn Morris]

Stefan Monnier wrote:

>> In Emacs 24.2 and earlier, defsetf was documented [1] to accept a third
>> argument = t, with the same meaning as gv-define-simple-setter's
>> optional FIX-RETURN argument. This behaviour no longer works.
>
> Indeed, thanks, fixed.

Still fails for me as originally described.
I think it's gv-define-simple-setter that's broken now, by #12813.
The FIX-RETURN case... doesn't.

bug#12812: defsetf with third argument = t

[Glenn Morris]

PS gv, cl-lib etc all looks like very nice stuff. Just reporting these
issues while trying to figure out how to document some of it.

bug#12812: defsetf with third argument = t

[Stefan Monnier]

>>> In Emacs 24.2 and earlier, defsetf was documented [1] to accept a third
>>> argument = t, with the same meaning as gv-define-simple-setter's
>>> optional FIX-RETURN argument. This behaviour no longer works.
>> Indeed, thanks, fixed.
> Still fails for me as originally described.
> I think it's gv-define-simple-setter that's broken now, by #12813.
> The FIX-RETURN case... doesn't.

Indeed, I fix it and broke it right away.  Should be fixed for good
this time.

> PS gv, cl-lib etc all looks like very nice stuff.

Thanks.  I'm particularly proud of gv.el ;-)

> Just reporting these issues while trying to figure out how to document
> some of it.

Tank you for doing that.  FWIW, I think the third arg of defsetf should
be undocumented.  I hesitated whether to fix it or not since AFAIK the
only users of that feature were in cl-macs.el and now use
gv-define-simple-setter instead.  The only reason why I installed the
fix is that it was such a trivial change to make.


        Stefan