When you reply to a non-ascii sender, VM has a very annoying habit of
flicking the compose buffer name back and forth between ... and ...<1>
(or whatever it is).
This is because of a couple of minor logic errors in the buffer
renaming code.
Here's a patch (against 8.2.0b).
Sorry I didn't send this in when I did it a few months ago!
--- vm-reply.el.orig 2012-01-22 13:43:14.000000000 +0000
+++ vm-reply.el 2012-01-22 13:40:47.000000000 +0000
@@ -961,10 +961,12 @@
(if (> (length prefix) 44)
(setq prefix (concat (substring prefix 0 40) " ...")))
(setq name prefix n 2)
- (while (get-buffer name)
- (setq name (format "%s<%d>" prefix n))
- (vm-increment n))
- (rename-buffer name))))))
+ (if (equal name (buffer-name))
+ t ; do nothing
+ (while (get-buffer name)
+ (setq name (format "%s<%d>" prefix n))
+ (vm-increment n))
+ (rename-buffer name)))))))
(defun vm-mail-mark-replied ()
(save-excursion
@@ -2076,9 +2078,9 @@
newbufname (funcall vm-chop-full-name-function newbufname)
newbufname (or (car newbufname) (car (cdr newbufname)))
newbufname (format fmt newbufname ellipsis
- (mail-fetch-field "Subject")))
+ (mail-fetch-field "Subject"))
+ newbufname (vm-sanitize-buffer-name newbufname))
(unless (equal newbufname curbufname)
- (setq newbufname (vm-sanitize-buffer-name newbufname))
(rename-buffer newbufname t)))))
;;;###autoload
