bug#28631: 25.3.50; Deactivate mark with Ctrl-g mess up the primary-selection

[Tino Calancha]

On Fri, 13 Oct 2017, Noam Postavsky wrote:

> Eli Zaretskii <eliz@gnu.org> writes:
>
> > >             ((and (/= (region-beginning) (region-end))
> > >                   (or (gui-backend-selection-owner-p 'PRIMARY)
> > > -                     (null (gui-backend-selection-exists-p 'PRIMARY))))
> > > +                      (null (gui-backend-selection-exists-p 'PRIMARY)))
> > > +                  ;; `winner-undo' shouldn't update the selection (Bug#28631).
> > > +                  (not (eq this-command 'winner-undo)))
> > >              (gui-set-selection 'PRIMARY
> > >                                  (funcall region-extract-function nil)))))
>
> > Is there really no way to solve this in winner?  It seems like a
> > winner bug/misfeature, and I'm worried by the possible effect of this
> > patch on use cases that have nothing to do with the specific scenario
> > of this bug.  deactivate-mark is used a lot in places and ways we
> > cannot possibly predict.
>
> That patch only has affect during winner-undo, no?
I think so.

> > +      ;; `winner-undo' shouldn't update the selection (Bug#28631) when
> > +      ;; select-enable-primary is non-nil.
> > +      (unless select-enable-primary
> > +        (save-current-buffer
> > +         (cl-loop for buf in buffers
> > +                   for entry = (cadr (assq buf winner-point-alist))
> > +                   do (progn (set-buffer buf)
> > +                             (set-mark (car entry))
> > +                             (setf (winner-active-region) (cdr entry))))))
>
> Maybe only the (setf (winner-active-region) (cdr entry)) part should be
> skipped?
We need to ban
(set-mark (car entry))
as well, because it updates the primary selection.

emacs  -Q -eval '(winner-mode 1)' \
-eval "(customize-set-variable 'select-enable-clipboard nil)" \
-eval "(customize-set-variable 'select-enable-primary t)"

M-<
M-: (set-mark 15) RET
M-: C-y ; It inserts in minibuffer ";; This buffer"

I prefer this patch because:
1) Make the fix inside winner.el
2) The fix takes effect just if user has enabled select-enable-primary

I don't fully understand the purpose of that part of the code, but
I assume if someone have set select-enable-primary, then she
probably doesn't want winner-undo to change her selection.