Re: long_to_cons doesn't always return a cons

[Alan Donovan]

On Sun, Jun 17, 2007 at 05:49:12PM -0400, Richard Stallman wrote:
> Does this fix it?
>
> *** fileio.c  22 Mar 2007 11:13:17 -0400      1.580
> --- fileio.c  17 Jun 2007 10:56:38 -0400
> ***************
> *** 5692,5702 ****
>   See Info node `(elisp)Modification Time' for more details.  */)
>        ()
>   {
> !   Lisp_Object tcons;
> !   tcons = long_to_cons ((unsigned long) current_buffer->modtime);
> !   if (CONSP (tcons))
> !     return list2 (XCAR (tcons), XCDR (tcons));
> !   return tcons;
>   }
>
>   DEFUN ("set-visited-file-modtime", Fset_visited_file_modtime,
> --- 5692,5700 ----
>   See Info node `(elisp)Modification Time' for more details.  */)
>        ()
>   {
> !   if (! current_buffer->modtime)
> !     return make_number (0);
> !   return make_time ((time_t) current_buffer->modtime);
>   }
>
>   DEFUN ("set-visited-file-modtime", Fset_visited_file_modtime,

I confess I'm not building from CVS but using the emacs-snapshot
package instead, so I haven't tested this patch.  From the point of
view of visited-file-modtime, this change looks good to me.

Of course, long_to_cons still doesn't meet its spec in the case where
the top 16 bits are zero, so any other place where this function is
called might suffer from a similar problem.  If I were you, I would
additionally update the docstring for long-to-cons and cons-to-long to
mention the special treatment of values in this range.

cheers
alan