Re: Nonsensical byte compiler warning.

[Alan Mackenzie <acm@muc.de>]

Hi Chong and David!

On Sun, Apr 01, 2007 at 02:10:18PM -0400, Chong Yidong wrote:
> David Kastrup <dak@gnu.org> writes:

> > when compiling the latest source (with "make cvs-update" in the lisp
> > subdirectory), I get

> > Compiling /home/tmp/emacs/lisp/progmodes/cc-cmds.el...

> > In c-end-of-defun:
> > progmodes/cc-cmds.el:1612:4:Warning: value returned by `char-after' is not
> >     used
> > Wrote /home/tmp/emacs/lisp/progmodes/cc-cmds.elc

> > I have no idea where this compiler warning is supposed to come from.

> I don't know why the byte-compiler printed this confusing message, but
> it seems to be issuing a real warning.  There is a `when' statement at
> cc-cmds.el:1633 that has no effect, because the return value of the
> surrounding `if' form is discarded.

Yes.  There's something stupid about the code there - it looks as though
I've half deleted something at some stage.  I'll sort it out early this
week, sometime.  (Other things are of higher priority at the moment.)

> This is probably just a bit of cruft, and easily corrected
> (eliminating the warning in the process).  Alan, could you verify
> this?

No, this is a long standing problem with cc-cmds.el.  I first tried to
track it down (compiled with Emacs 21) in April 2006, but didn't succeed,
beyond identifying the line which I needed to comment out to get rid of
the warning.  The line was "(eq (char-before) ?\})" and was at the time
in `c-end-of-defun' (which has since been radically changed).  That line
is now L1635; commenting it out silences the warning.

The bug never seemed important enough to lose a lot of sleep over, but it
does irritate.  The (more detailed) warning message reported by Emacs 21
is:

    While compiling toplevel forms:
      ** `(char-after (1- (point)))' called for effect

Seemingly, `char-before' is implemented as a macro in the byte-compiler
in Emacs 21.

I strongly believe that cc-cmds.el gets compiled correctly, and that the
spurious warning is a bug in the byte compiler, triggered by something
unusual in cc-cmds.el.

Chong, could you ask the byte-compiler hacker to look at this, please?
(I don't know who that is).


> *** emacs/lisp/progmodes/cc-cmds.el.~1.55.~	2007-03-30 18:31:07.000000000 -0400
> --- emacs/lisp/progmodes/cc-cmds.el	2007-04-01 14:05:04.000000000 -0400
> ***************
> *** 1630,1639 ****
>   	      (setq arg (1+ arg)))
>   	  (if (< arg 0)
>   	      (setq arg (c-backward-to-nth-BOF-{ (- arg) where)))
> ! 	  (when (and (= arg 0)
> ! 		     (c-syntactic-skip-backward "^}")
> ! 		     (eq (char-before) ?\}))
> ! 	    t))

>         ;; Move forward to the } of a function
>         (if (> arg 0)
> --- 1630,1637 ----
>   	      (setq arg (1+ arg)))
>   	  (if (< arg 0)
>   	      (setq arg (c-backward-to-nth-BOF-{ (- arg) where)))
> ! 	  (if (= arg 0)
> ! 	      (c-syntactic-skip-backward "^}")))

>         ;; Move forward to the } of a function
>         (if (> arg 0)

-- 
Alan.