bug#19058: [PATCH] Desktop restore fails to run mark activation hooks when it should

bug#19058: [PATCH] Desktop restore fails to run mark activation hooks when it should

[Kelly Dean]

[-- Attachment #1: Type: text/plain, Size: 1668 bytes --]

Delete your .emacs.desktop, and put in init.el:

(desktop-save-mode 1)
(add-hook 'deactivate-mark-hook (lambda () (setq cursor-type t)))
(add-hook 'activate-mark-hook (lambda () (setq cursor-type 'bar)))

Start Emacs 24.4, open a file, press C-SPC, then exit Emacs and save the desktop, then restart Emacs. Notice that the cursor type is now a block, not a bar. It should be a bar.

This bug wasn't in 24.3, and is the opposite of bug 14430. The fix for this new bug is the same patch that I already sent on May 20, 2013 for that bug (my patch fixed that bug without introducing this new bug). That patch wasn't accepted, and instead this new bug and a FIXME comment were introduced into desktop.el for 24.4.

This new bug could be fixed without applying that patch by instead applying second-restoremarkbug.bad-patch that's attached to this message (conditionally calling activate-mark as the new FIXME proposes), but that's a hack that invites future bugs when somebody modifies set-mark and forgets to correspondingly modify the code in desktop-create-buffer that duplicates the functionality of set-mark. The right fix is to not duplicate the functionality of set-mark, and instead just call set-mark and tell it whether to call activate-mark, as my original patch did.

I've been asked to not send any more patches to apply to Emacs, but this new patch is just to point out what NOT to apply, so I trust it's ok to send it.

Of course, nobody besides me cares about this bug, but its existence serves as a convenient excuse for me to ask Stefan to please add the dont-activate option to set-mark so I don't have to maintain it anymore in my personal patches. ;-)

[-- Attachment #2: second-restoremarkbug.bad-patch --]
[-- Type: application/octet-stream, Size: 730 bytes --]

Don't apply this patch! It's the wrong fix for the bug.

--- emacs-24.4/lisp/desktop.el
+++ emacs-24.4/lisp/desktop.el
@@ -1413,8 +1413,7 @@
             (if (consp desktop-buffer-mark)
                 (progn
                   (move-marker (mark-marker) (car desktop-buffer-mark))
-                  ;; FIXME: Should we call (de)activate-mark instead?
-                  (setq mark-active (car (cdr desktop-buffer-mark))))
+                  (if (car (cdr desktop-buffer-mark)) (activate-mark 'no-tmm)))
               (move-marker (mark-marker) desktop-buffer-mark)))
 	  ;; Never override file system if the file really is read-only marked.
 	  (when desktop-buffer-read-only (setq buffer-read-only desktop-buffer-read-only))

bug#19058: [PATCH] Desktop restore fails to run mark activation hooks when it should

[Stefan Monnier]

> Of course, nobody besides me cares about this bug, but its existence serves
> as a convenient excuse for me to ask Stefan to please add the dont-activate
> option to set-mark so I don't have to maintain it anymore in my personal
> patches. ;-)

But if you call set-mark and a non-nil dont-activate argument, you get
the same as (move-marker (mark-marker) <foo>), so this arg only makes
sense in the case where you dynamically want to choose whether to
activate the mark or not, which is very rare.

So I installed a patch that just calls activate-mark.


        Stefan