bug#13676: 24.2.93; Compiler warnings with (defalias #'symbol ...)

bug#13676: 24.2.93; Compiler warnings with (defalias #'symbol ...)

[Michael Heerdegen]

Hi,

after reading (long ago) in (elisp) Anonymous Functions:

,----------------------------------------------------------------------
|  -- Special Form: function function-object
|
|  [...]  But unlike `quote', it also serves as a note to the Emacs
|evaluator and byte-compiler that FUNCTION-OBJECT is intended to be used
|as a function.  [...]
`----------------------------------------------------------------------

I often used something like this:

--8<---------------cut here---------------start------------->8---
;; my package --- doing cool stuff

(defalias #'my-package-end-of-line
  (if (fboundp #'end-of-visual-line)
      #'end-of-visual-line
    #'end-of-line))

(defun my-package-do-cool-stuff ()
  "Doc..."
  (ding)
  (goto-char (point-min))
  (my-package-end-of-line))
--8<---------------cut here---------------end--------------->8---

If I compile this, I get a warning:

,----------------------------------------------------------------------
| Compiling file /home/micha/today/my-package.el at Sun Feb 10 17:54:08 2013
| 
| In end of data:
| my-package.el:13:1:Warning: the function `my-package-end-of-line' is not known
|     to be defined.
`----------------------------------------------------------------------

If the first arg of `defalias' is quoted with just quote (instead of
`#''), however, you don't get a warning.

Dunno if it's no good style to use `#'' at that place - OTOH, it works,
and the warning is not useful.

A fix would presumably need to change the first condition of the `pcase' in
`byte-compile-file-form-defalias', probably to something like

  `(,_ ,(or `',name `(function ,name)) ,arg . ,rest)


Thanks,

Michael.


In GNU Emacs 24.2.93.1 (x86_64-unknown-linux-gnu, GTK+ Version 2.24.10)
 of 2013-02-09 on drachen
Bzr revision: cyd@gnu.org-20130209044342-t2b063zu2fqy8nud
Windowing system distributor `The X.Org Foundation', version 11.0.10707000
System Description:	Debian GNU/Linux 7.0 (wheezy)

Configured using:
 `configure '--prefix=/usr/local/built/''

Important settings:
  value of $LC_ALL: de_DE.utf8
  value of $LC_TIME: C
  value of $LANG: de_DE.utf8
  locale-coding-system: utf-8-unix
  default enable-multibyte-characters: t

bug#13676: 24.2.93; Compiler warnings with (defalias #'symbol ...)

[Stefan Monnier]

> |  [...]  But unlike `quote', it also serves as a note to the Emacs
> |evaluator and byte-compiler that FUNCTION-OBJECT is intended to be used
> |as a function.  [...]

Right, this means that #'foo is a reference to the function stored in `foo'.

> (defalias #'my-package-end-of-line

So, this is weird, because you're not referring to the function stored
in `my-package-end-of-line' here, instead you really want to set the
`symbol-function' slot of the symbol, so the argument you want to pass
to `defalias' is not a function but a symbol.

Both work in practice, but the intention expressed by the use of #' is
incorrect here.

>   (if (fboundp #'end-of-visual-line)

Similarly, here it should be (fboundp 'end-of-visual-line) since
fboundp's argument should be a symbol, not a function.


        Stefan

bug#13676: 24.2.93; Compiler warnings with (defalias #'symbol ...)

[Michael Heerdegen]

Hello Stefan,

thanks for the explanation.  I'll close it.


Michael.

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> > |  [...]  But unlike `quote', it also serves as a note to the Emacs
> > |evaluator and byte-compiler that FUNCTION-OBJECT is intended to be used
> > |as a function.  [...]
>
> Right, this means that #'foo is a reference to the function stored in
> foo'.
>
> > (defalias #'my-package-end-of-line
>
> So, this is weird, because you're not referring to the function stored
> in `my-package-end-of-line' here, instead you really want to set the
> `symbol-function' slot of the symbol, so the argument you want to pass
> to `defalias' is not a function but a symbol.
>
> Both work in practice, but the intention expressed by the use of #' is
> incorrect here.
>
> >   (if (fboundp #'end-of-visual-line)
>
> Similarly, here it should be (fboundp 'end-of-visual-line) since
> fboundp's argument should be a symbol, not a function.
>
>
>         Stefan