ls --color=auto "issue" (GNU fileutils) 4.0

[R. Bernstein]

ls --color=auto will try to colorize when LS_COLORS is null. LS_COLORS
is set null for example by the dircolors program when there is a
terminal (like dumb) that is not known to be able to color.

Below is a patch that works, although I'm don't think it is the *best*
patch.

The idea in the patch below is that if parse_ls_color finds the null
string it sets the variable to colorize (print_with_color) off.

My taste would be to change parse_ls_color into a function which
returns false if parsing fails or the LS_COLORS environment variable
is the null string.

So instead of:

if (print_with_color) 

in the fileutils 4.0 code, we might have 

if (print_with_color && prep_non_filename_text()) 

I've checked the fileutils 4.0x pachage on alpha.gnu.org and the
situation is the same there.

*** ls-orig.c  Sat Sep 19 13:09:23 1998
--- ls.c       Sun Aug  6 15:13:47 2000
***************
*** 709,715 ****
    if (print_with_color)
      {
        parse_ls_color ();
!       prep_non_filename_text ();
      }
  
    format_needs_stat = sort_type == sort_time || sort_type == sort_size
--- 709,715 ----
    if (print_with_color)
      {
        parse_ls_color ();
!       if (print_with_color) prep_non_filename_text ();
      }
  
    format_needs_stat = sort_type == sort_time || sort_type == sort_size
***************
*** 1413,1421 ****
    struct col_ext_type *ext;   /* Extension we are working on */
    struct col_ext_type *ext2;  /* Extra pointer */
  
!   if ((p = getenv ("LS_COLORS")) == NULL || *p == '\0')
      return;
  
    ext = NULL;
    strcpy (label, "??");
  
--- 1413,1426 ----
    struct col_ext_type *ext;   /* Extension we are working on */
    struct col_ext_type *ext2;  /* Extra pointer */
  
!   if ((p = getenv ("LS_COLORS")) == NULL)
      return;
  
+   if (*p == '\0') {
+     print_with_color = 0;
+     return;
+   }
+ 
    ext = NULL;
    strcpy (label, "??");