bug#48478: 28.0.50; yank-from-kill-ring and kill-ring-yank-pointer

[Juri Linkov]

> If one uses "C-y M-y M-y ... C-y", the last C-y will yank the
> same kill-ring entry that the M-y before it.  But if one uses
> "M-y M-y ... C-y", the last C-y will yank not the entry produced
> by the last M-y, but the one after it.  Which means
> yank-from-kill-ring leaves the kill-ring-yank-pointer at a different
> entry than yank-pop does.
>
> Bug or feature?  If the latter, what is the rationale for the
> different operation?

This is an unfinished feature, i.e. currently yank-from-kill-ring
simply has no special handling of kill-ring-yank-pointer.
When I tried to add it, I noticed that it has many possible
variants of implementation, so I stuck with analysis paralysis.

Perhaps we should have a new option with several choices:

1. Don't change kill-ring-yank-pointer (as it does now).
   This is useful when the user wants to type a predictable
   number of M-p in the minibuffer history to yank the same
   previously-killed text from the kill-ring several times.
2. Adjust kill-ring-yank-pointer to point to the selected text.
3. Still unclear how to adjust kill-ring-yank-pointer
   when the user selects a previous text
   then edits it before yanking.

Regarding 7b82584c69, actually the description of 'M-y' in
(info "(emacs) Isearch Yank") was accurate.  It correctly
described the behavior of isearch-yank-pop that activates
the minibuffer.  The problem is that after recent controversy,
'M-y' was rebound to the backward-compatible command
isearch-yank-pop-only.

A FIXME from 7b82584c69 is implemented by this patch:

diff --git a/lisp/isearch.el b/lisp/isearch.el
index 095f8ba145..c52718d94a 100644
--- a/lisp/isearch.el
+++ b/lisp/isearch.el
@@ -2583,7 +2585,7 @@ isearch-yank-from-kill-ring
   "Read a string from the `kill-ring' and append it to the search string."
   (interactive)
   (with-isearch-suspended
-   (let ((string (read-from-kill-ring)))
+   (let ((string (read-from-kill-ring "Yank from kill-ring: ")))
      (if (and isearch-case-fold-search
               (eq 'not-yanks search-upper-case))
          (setq string (downcase string)))
diff --git a/lisp/simple.el b/lisp/simple.el
index 90942df0da..08c7c903a8 100644
--- a/lisp/simple.el
+++ b/lisp/simple.el
@@ -5587,7 +5587,8 @@ yank-pop
 property, in the way that `yank' does."
   (interactive "p")
   (if (not (eq last-command 'yank))
-      (yank-from-kill-ring (read-from-kill-ring) current-prefix-arg)
+      (yank-from-kill-ring (read-from-kill-ring "Yank from kill-ring: ")
+                           current-prefix-arg)
     (setq this-command 'yank)
     (unless arg (setq arg 1))
     (let ((inhibit-read-only t)
@@ -5676,7 +5677,7 @@ rotate-yank-pointer
   (current-kill arg))
 
 (defvar read-from-kill-ring-history)
-(defun read-from-kill-ring ()
+(defun read-from-kill-ring (prompt)
   "Read a `kill-ring' entry using completion and minibuffer history."
   ;; `current-kill' updates `kill-ring' with a possible interprogram-paste
   (current-kill 0)
@@ -5721,11 +5722,7 @@ read-from-kill-ring
              (define-key map "?" nil)
              map)))
       (completing-read
-       ;; FIXME: This prompt is specific to using this function from
-       ;; yank-related commands, but the function could be used in
-       ;; other contexts.  Should the prompt be passed via an
-       ;; argument?
-       "Yank from kill-ring: "
+       prompt
        (lambda (string pred action)
          (if (eq action 'metadata)
              ;; Keep sorted by recency
@@ -5755,7 +5752,8 @@ yank-from-kill-ring
 beginning of the inserted text and mark at the end, like `yank' does.
 
 When called from Lisp, insert STRING like `insert-for-yank' does."
-  (interactive (list (read-from-kill-ring) current-prefix-arg))
+  (interactive (list (read-from-kill-ring "Yank from kill-ring: ")
+                     current-prefix-arg))
   (push-mark)
   (insert-for-yank string)
   (if (consp arg)