bug#14464: 24.3; require sets deactivate-mark causing problems with some plugins

[Glenn Morris <rgm@gnu.org>]

[-- Attachment #1: message body text --]
[-- Type: text/plain, Size: 1160 bytes --]


Please keep the debbugs address cc'd.
I am resending your comments so that they are visible to others.


Date: Fri, 24 May 2013 17:47:31 -0500
From: Donald Curtis <dcurtis@milkbox.net>
Subject: Re: bug#14464: 24.3; require sets deactivate-mark causing problems
 with some plugins

Yes, this is what my initial intuition was. I have traced this error down
to something much more obscure. I don't understand why it's happening. But
let me give you some code to reproduce the odd behavior without concern for
`multiple-cursors`.

Make sure both files are in the same directory.

# Working Correctly

1. open `sample.el` with: emacs -Q sample.el
2. eval just the `progn` portion of the provided code.
3. transient mark something
4. hit C-.

the active-mark is still there. now exit emacs and start again,

# Odd Behavior

1. open `sample.el` with: emacs -Q sample.el
2. eval the `defadvice`
3. eval the `progn` portion of the provided code.
4. transient mark something
5. hit C-.

now the active-mark disappears.

AND what is even weirder is that if you remove the `compile` option from
the `defadvice` then there is no issue. but there is something weird here!


[-- Attachment #2: sample.el --]
[-- Type: application/octet-stream, Size: 367 bytes --]

;; scratch.el file

(defadvice kill-line (after kill-line-cleanup-whitespace activate compile)
  "Cleanup whitespace on `kill-line'."
  (if (not (bolp))
      (delete-region (point) (progn (skip-chars-forward " \t") (point)))))

(progn (add-to-list 'load-path ".")
       (autoload 'testtest "autoloadme" "\(fn)" t nil)
       (global-set-key (kbd "C-.") 'testtest))

[-- Attachment #3: autoloadme.el --]
[-- Type: application/octet-stream, Size: 304 bytes --]

;;;###autoload

(defun testtest (arg)
  (interactive "p")
  (message "deactivate-mark: %s   arg: %s" deactivate-mark arg))

(defadvice read-char (around test-read-char activate)
  (message "read-char %s" deactivate-mark)
  ad-do-it)

(message "deactivate-mark: %s" deactivate-mark)
(provide 'autoloadme)

[-- Attachment #4: message body text --]
[-- Type: text/plain, Size: 294 bytes --]


On Fri, May 24, 2013 at 4:02 PM, Glenn Morris <rgm@gnu.org> wrote:

> I think it's unlikely that require itself messes with the mark.
> The thing that you are requiring can of course do anything at all.
> Sounds like this could be a problem in "multiple-cursors", which isn't
> part of Emacs.