bug#21464: 25.0.50; [cc-langs] void-function cadar

bug#21464: 25.0.50; [cc-langs] void-function cadar

[Mark Oteiza]

Hi,

From emacs -Q, loading

  (eval-when-compile (require 'cc-langs))
  (c-initialize-cc-mode t)
  (c-init-language-vars c-mode)

gives

"Eval error in the ‘c-lang-defvar’ or ‘c-lang-setver’ for ‘comment-start’ (source eval): (void-function cadar)"

So, similar code for some derived mode errors

In GNU Emacs 25.0.50.1 (x86_64-unknown-linux-gnu, X toolkit, cairo version 1.14.2, Xaw scroll bars)
 of 2015-09-11
Repository revision: a0ec54ae073abd671bd43002eff0267f5fe8b306

bug#21464: 25.0.50; [cc-langs] void-function cadar

[Stefan Monnier]

> "Eval error in the ‘c-lang-defvar’ or ‘c-lang-setver’ for ‘comment-start’ (source eval): (void-function cadar)"

Using cl-lib instead of cl would solve such issues easily since we could
use cl-cadar and (require 'cl-lib) at run-time instead of only requiring
cl at compile-time.

I know we've discussed it already, but I urge you (Alan) to take
another look at the idea of unconditionally using cl-lib (and just
bundling it with cc-mode).  I'd be happy to help you do that.

Then you'll be free to use CL functions (via the cl-*
namespace) anywhere without having to worry about whether it's used at
runtime or compile time.

This is bound to happen sooner or later since cl.el is slated to move from
lisp/emacs-lisp to lisp/obsolete and then to GNU ELPA.

But even if cl.el were to stay in lisp/emacs-lisp, I think CC-mode
maintenance would benefit from it.


        Stefan

bug#21464: 25.0.50; [cc-langs] void-function cadar

[Alan Mackenzie]

Hello, Stefan.

In article <mailman.969.1442020629.19560.bug-gnu-emacs@gnu.org> you wrote:
>> "Eval error in the ?c-lang-defvar? or ?c-lang-setver? for ?comment-start? (source eval): (void-function cadar)"

I've gone and replaced cadar with cadr/car in the two places it appeared
in cc-langs.el.  They were the only occurrences of c[ad]\{3,\}r in CC
Mode.  There are quite a few compositions of caar, etc., with car and cdr,
and even compositions of car and cdr, presumably dating from a time before
caar etc. existed.

> Using cl-lib instead of cl would solve such issues easily since we could
> use cl-cadar and (require 'cl-lib) at run-time instead of only requiring
> cl at compile-time.

CC Mode in Emacs (as opposed to XEmacs) has been using cl-lib instead of
cl for some weeks, now.

> I know we've discussed it already, but I urge you (Alan) to take
> another look at the idea of unconditionally using cl-lib (and just
> bundling it with cc-mode).  I'd be happy to help you do that.

Bundling cl-lib with CC Mode doesn't feel like the Right Thing to do.  It
would be bound to clash with a cl-lib in Emacs (or a cl in XEmacs)
somehow.

> Then you'll be free to use CL functions (via the cl-*
> namespace) anywhere without having to worry about whether it's used at
> runtime or compile time.

> This is bound to happen sooner or later since cl.el is slated to move from
> lisp/emacs-lisp to lisp/obsolete and then to GNU ELPA.

> But even if cl.el were to stay in lisp/emacs-lisp, I think CC-mode
> maintenance would benefit from it.

Possibly.  But there doesn't seem to be an urgent need for it at the
moment.

I'm marking this bug as closed.

>         Stefan

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#21464: 25.0.50; [cc-langs] void-function cadar

[Stefan Monnier]

>> Using cl-lib instead of cl would solve such issues easily since we could
>> use cl-cadar and (require 'cl-lib) at run-time instead of only requiring
>> cl at compile-time.
> CC Mode in Emacs (as opposed to XEmacs) has been using cl-lib instead of
> cl for some weeks, now.

By "using cl-lib" I mean "assume cl-lib is always available" and
"don't use cl anymore anywhere".

That's a lot more convenient than the current testing and using various
macro wrappers depending on the circumstance and it'd let you use
any CL functions and macros anywhere any time.

> It would be bound to clash with a cl-lib in Emacs

Your assumptions are faulty, I think.

You'd put it in a subdirectory, say "cc-cl-lib" (so it's not in
load-path by default) and in cc-compat.el you'd do something like:

  (unless (featurep 'cl-lib)
    (unless (locate-library "cl-lib")
      (add-to-list 'load-path
                   (expand-file-name "cc-cl-lib"
                                     (file-name-directory load-file-name))))
    (require 'cl-lib))

or

  (unless (featurep 'cl-lib)
    (require 'cl-lib
             (unless (locate-library "cl-lib")
               (expand-file-name "cc-cl-lib/cl-lib"
                                 (file-name-directory load-file-name)))))

> (or a cl in XEmacs) somehow.

The cl-lib.el I'm talking about is the one in GNU ELPA, which is
basically loads cl.el and then adds a whle bunch of aliases under the
new "cl-*" namespace.  So no, it doesn't clash with cl.el, instead it
relies on it.

> I'm marking this bug as closed.

Indeed, thanks,


        Stefan