GNU bug report logs - #640
Forward: Buffer being killed not current in kill-buffer-hook

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 640 in the body.
You can then email your comments to 640 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Max Mikhanosha <max <at> openchat.com>
To: emacs-pretest-bug <at> gnu.org
Subject: Forward: Buffer being killed not current in kill-buffer-hook 
Date: Fri, 01 Aug 2008 12:22:52 -0400

[Message part 1 (text/plain, inline)]

I posted this to emacs-dev, but there was no response. This is still
happenning to me with latest CVS build.

[Message part 2 (message/rfc822, inline)]

From: Max Mikhanosha <max <at> openchat.com>
To: emacs-devel <at> gnu.org
Subject: Buffer being killed not current in kill-buffer-hook 
Date: Sat, 19 Jul 2008 16:29:25 -0400

When using multi-tty emacsclient, the following piece of code demostrates the problem:

DISPLAY is unset, CVS Emacs started with --no-init-file

(require 'server)
(progn (setq server-name "test")
       (server-start))
(defun print-current-buffer ()
  (message "current-buffer=%s" (current-buffer)))
(add-hook 'kill-buffer-hook 'print-current-buffer)

Now from a different screen session, connect with emacsclient like so:
"lib-src/emacsclient -c -s test foo". New tty frame will be created. 
Immeditely kill the buffer, it will ask "Buffer has existing clients"
say "yes". emacsclient will disconnect from a server.

In the *Messages* buffer on the server frame:

When done with a buffer, type C-x #
(No files need saving)
current-buffer=*scratch*

All the code that uses kill-buffer-hook rightfully assumes that the
buffer being killed will be current buffer when called, as that is
what is stated in the kill-buffer-hook documentation. For example
saveplace does not work because of this.. ERC leaves a channel if 
unrelated buffer is killed on a different terminal, etc.

I had fixed this with the following patch, altho I don't know enough
about server.el logic to figure out if what it does is ok? Seems to
have fixed the bug and not broke anything else:

Index: lisp/server.el
===================================================================
RCS file: /cvsroot/emacs/emacs/lisp/server.el,v
retrieving revision 1.166
diff -u -b -r1.166 server.el
--- lisp/server.el	6 May 2008 07:57:49 -0000	1.166
+++ lisp/server.el	19 Jul 2008 20:28:26 -0000
@@ -1084,7 +1084,8 @@
 	  ;; tell it that it is done, and forget it entirely.
 	  (unless buffers
 	    (server-log "Close" proc)
-	    (server-delete-client proc)))))
+            (save-current-buffer 
+              (server-delete-client proc))))))
     (when (and (bufferp buffer) (buffer-name buffer))
       ;; We may or may not kill this buffer;
       ;; if we do, do not call server-buffer-done recursively

Message #10 received at 640-done <at> emacsbugs.donarmstrong.com (full text, mbox):

From: Chong Yidong <cyd <at> stupidchicken.com>
To: Max Mikhanosha <max <at> openchat.com>
Cc: 640-done <at> debbugs.gnu.org
Subject: Re: Buffer being killed not current in kill-buffer-hook 
Date: Wed, 24 Sep 2008 16:14:17 -0400

> I posted this to emacs-dev, but there was no response. This is still
> happenning to me with latest CVS build.

Sorry for the delay.  I've just checked in a slightly more conservative
version of you patch (since I'm not sure it's correct to apply
save-current-buffer in all situations).  Thanks for the report and the
patch.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.