Apparently incorrect changes to custom-magic-alist.

Apparently incorrect changes to custom-magic-alist.

[Luc Teirlinck]

I propose the changes to cus-edit.el contained in the patch below.
They essentially constitute a partial revert to the previous version.

Part of recent changes to custom-magic-alist are wrong.

Do `emacs -q'.
M-x customize-group auto-revert RET

Now toggle Auto Revert Verbose to off and set for the current session.

Now erase the `5' in Auto Revert Interval.

Now it says:

you have edited something in this group, but not set anything yet.

But I _did_ set Auto Revert Verbose.

I can not give an `emacs -q' example showing that the message:

you have set something in this group, but not saved anything yet.

suffers from the same problem (because in `emacs -q', you can not
save), but in an ordinary Emacs, it may tell you that you did not save
anything, whereas you saved plenty of stuff.

===File ~/cus-edit.el-diff==================================
*** cus-edit.el	14 Feb 2005 15:38:32 -0600	1.212
--- cus-edit.el	14 Feb 2005 17:52:21 -0600	
***************
*** 1671,1680 ****
  the value displayed for this %c is invalid and cannot be set.")
      (modified "*" custom-modified-face "\
  you have edited the value as text, but you have not set the %c." "\
! you have edited something in this group, but not set anything yet.")
      (set "+" custom-set-face "\
  you have set this %c, but not saved it for future sessions." "\
! you have set something in this group, but not saved anything yet.")
      (changed ":" custom-changed-face "\
  this %c has been changed outside the customize buffer." "\
  something in this group has been changed outside customize.")
--- 1671,1680 ----
  the value displayed for this %c is invalid and cannot be set.")
      (modified "*" custom-modified-face "\
  you have edited the value as text, but you have not set the %c." "\
! you have edited something in this group, but not set it.")
      (set "+" custom-set-face "\
  you have set this %c, but not saved it for future sessions." "\
! you have set something in this group, but not saved it.")
      (changed ":" custom-changed-face "\
  this %c has been changed outside the customize buffer." "\
  something in this group has been changed outside customize.")
============================================================

Re: Apparently incorrect changes to custom-magic-alist.

[Richard Stallman]

      you have edited the value as text, but you have not set the %c." "\
    ! you have edited something in this group, but not set anything yet.")
	  (set "+" custom-set-face "\
      you have set this %c, but not saved it for future sessions." "\
    ! you have set something in this group, but not saved anything yet.")
	  (changed ":" custom-changed-face "\
      this %c has been changed outside the customize buffer." "\
      something in this group has been changed outside customize.")
    --- 1671,1680 ----
      the value displayed for this %c is invalid and cannot be set.")
	  (modified "*" custom-modified-face "\
      you have edited the value as text, but you have not set the %c." "\
    ! you have edited something in this group, but not set it.")

That string doesn't mean what you think it means.

I will make another change to deal with the error that 
you have shown.

Re: Apparently incorrect changes to custom-magic-alist.

[Luc Teirlinck]

Richard Stallman wrote:

	 the value displayed for this %c is invalid and cannot be set.")
	     (modified "*" custom-modified-face "\
	 you have edited the value as text, but you have not set the %c." "\
       ! you have edited something in this group, but not set it.")

   That string doesn't mean what you think it means.

I believe that it correctly describes the current behavior.

   I will make another change to deal with the error that you have shown.

I am not sure that they were errors.  The whole buffer State message
is problematic one way or the other, since there could be options with
nine different states in the buffer.  Short of saying how many options
of each type the buffer contains, the whole buffer state message can
not contain full information, and as such will be of limited usefulness.

I believe that the _relatively_ most useful thing for the whole State
message to do is to point out the least permanent state in the buffer,
because it is the most likely one the user might want to change.  That
is, is some option has been edited, but not set, point that out.  If
some option has been set, but not saved, point that out (if no option
has been edited but not set).  That appears to be the current behavior.

Sincerely,

Luc.