Re: closing all inactive server connections

[Katsumi Yamaoka <yamaoka@jpl.org>]

[-- Attachment #1: Type: text/plain, Size: 1491 bytes --]

>>>>> Ted Zlatanov wrote:

>>> (defun gnus-server-close-all-servers ()
>>> "Close all servers."

[...]

KY> I can agree to this change if it is really helpful to you.  But
KY> you still need to type `C-g' for almost servers when the VPN is
KY> disconnected, don't you?  Even so, is it really helpful to you?
KY> I don't have a problem like yours and don't know how to reproduce
KY> it without VPN, so don't know how to solve it so far.

> Yes, I need C-g, but I don't need to

> 1) go to each opened server
> 2) press C
> 3) press C-g

> (repeat 3-4 times for each foreign server that's hung)

> now I just

> 1) M-c
> 2) press C-g for whatever hangs.

> Much better user experience.  Of course, avoiding the C-g would be even
> better; my other question in the original e-mail was if I can tell Emacs
> all these connections are truly dead so Gnus doesn't hang trying to
> close them.

I see.  Though I've never looked into this problem throughly,
the cause is often `accept-process-output' to which the timeout is
not specified.  So, I guess `nntp-wait-for-string' is the culprit
(I have no idea for IMAP connections).  To verify it, we probably
need to remove all the `condition-case' forms and `ignore-errors'
forms from nntp.el.  But, er, it might be hard to do.  How about
deleting the processes immediately instead of quitting mannerly?
The following patch forces Gnus to delete the nntp process if
the server doesn't respond when Gnus tries to open it for sending
the QUIT command:


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Type: text/x-patch, Size: 1034 bytes --]

--- nntp.el~	2007-07-18 12:07:39 +0000
+++ nntp.el	2007-07-31 02:54:25 +0000
@@ -978,8 +978,12 @@
       (or (nntp-find-connection nntp-server-buffer)
 	  (nntp-open-connection nntp-server-buffer)))))
 
+(defvar nntp-force-close-server nil
+  "Non-nil means force server to close if it doesn't respond immediately.")
+
 (deffoo nntp-close-server (&optional server)
-  (nntp-possibly-change-group nil server t)
+  (let ((nntp-force-close-server t))
+    (nntp-possibly-change-group nil server t))
   (let ((process (nntp-find-connection nntp-server-buffer)))
     (while process
       (when (memq (process-status process) '(open run))
@@ -1728,7 +1732,10 @@
     (while (and (setq proc (get-buffer-process buf))
 		(memq (process-status proc) '(open run))
 		(not (re-search-forward regexp nil t)))
-      (accept-process-output proc)
+      (if nntp-force-close-server
+	  (unless (accept-process-output proc 1)
+	    (delete-process proc))
+	(accept-process-output proc))
       (set-buffer buf)
       (goto-char (point-min)))))
 

[-- Attachment #3: Type: text/plain, Size: 245 bytes --]

KY> In addition, isn't it unnecessary to use `let'?  I mean:

[...]

> I was debugging, so the variable was handy.  I am OK with your version.

I see.  If you think the variable will be useful in the future
too, please leave it as is.

Regards,