Duplicated outline-cycle binding, and problems with the new one

[Yuan Fu]

When outline-cycle was originally added, I think we only added this

(defvar outline-mode-map
  (let ((map (make-sparse-keymap)))
    (define-key map "\C-c" outline-mode-prefix-map)
    (define-key map [menu-bar] outline-mode-menu-bar-map)
    ;; Only takes effect if point is on a heading.
    (define-key map (kbd "TAB")
      `(menu-item "" outline-cycle
                  :filter ,(lambda (cmd)
                             (when (outline-on-heading-p) cmd))))
    (define-key map (kbd "<backtab>") #'outline-cycle-buffer)
    map))

So ‘outline-cycle’ takes effect when point is on a heading. Now in addition to 
this, I also see

(defvar outline-minor-mode-cycle-map
  (let ((map (make-sparse-keymap)))
    (outline-minor-mode-cycle--bind map (kbd "TAB") #'outline-cycle)
    (outline-minor-mode-cycle--bind map (kbd "<backtab>") 
#'outline-cycle-buffer)
    map)
  "Keymap used by `outline-minor-mode-cycle'.”)

Which presumably are applied as keymap text properties to headings in a buffer. 
I’m having problems with this: some text that are not headings in my buffer are 
incorrectly propertied with this keymap, and when I try to indent with TAB, 
outline-cycle is invoked.

Overall, from my limited knowledge, I think the old approach is more reliable: 
I wouldn’t have this problem with the old approach. And the new functionality 
added by the new approach and outline-minor-mode-cycle-filter can be easily 
implemented in the old approach. We don’t need to fiddle with 
font-lock-keywords with the old approach, either. How about we go back to the 
old approach? Juri, WDYT?

Yuan