Mailing-List:	contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe:	<mailto:cygwin-subscribe AT sources DOT redhat DOT com>
List-Archive:	<http://sources.redhat.com/ml/cygwin/>
List-Post:	<mailto:cygwin AT sources DOT redhat DOT com>
List-Help:	<mailto:cygwin-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender:	cygwin-owner AT sources DOT redhat DOT com
Delivered-To:	mailing list cygwin AT sources DOT redhat DOT com
Message-ID:	<397C737A.5408963C@cs.unc.edu>
Date:	Mon, 24 Jul 2000 12:48:58 -0400
From:	Jeffrey Juliano <juliano AT cs DOT unc DOT edu>
X-Mailer:	Mozilla 4.73 [en] (WinNT; U)
X-Accept-Language:	en
MIME-Version:	1.0
To:	cygwin AT sources DOT redhat DOT com
Subject:	ssh client (corinna's port) hangs doing 'cvs update'

Can someone please confirm that they can reproduce this problem...


In April, I reported a problem with Sergy's ssh port
http://sources.redhat.com/ml/cygwin/2000-04/msg00646.html

Unfortunately, I see the same problem when using cygwin latest and
Corinna's openSSH port.  CVS is from the cygutils.netpedia.net.  The
problem occurs on win98, but not winNT, when CVS_RSH=ssh.  Here is a
simple test case (but the problem occurs with *any* cvs operation):

$ cvs -d cystine:$NANOREPOSITORY co -c
juliano AT cystine's password: 
nano         nano
nano-all     -a quat vogl vrpn sdi nano
quat         quat
sdi          sdi
vogl         vogl
vrpn         vrpn
vrpn-all     -a vrpn vrpn_html
vrpn_html    vrpn_html


At this point, ssh (and thus cvs) is hung.  A running ps in a different
shell reveals:

$ ps -lf
    PID  PPID  PGID   WINPID  UID TTY    STIME COMMAND
   1024  1000  1024 4294669727 2388   0 10:04:33 /USR/LOCAL/BIN/RXVT.EXE
   1025  1024  1025 4294378575 2388   2 10:04:33 /usr/bin/BASH.EXE
I 20801  1025 20801 4294617787 2388   2 11:02:48 /USR/LOCAL/BIN/CVS.EXE
  20802 20801 20801 4294435463 2388   2 11:02:48 /USR/LOCAL/BIN/SSH.EXE
  20804  1006 20804 4294370715 2388   1 11:03:56 /usr/bin/PS.EXE
   1000  1000  1000 4294445443 2388   0 09:32:10 /usr/bin/BASH.EXE
   1005  1000  1005 4294478263 2388   0 09:32:20 /USR/LOCAL/BIN/RXVT.EXE
   1006  1005  1006 4294388519 2388   1 09:32:22 /usr/bin/BASH.EXE


Incidently, what is the "I" that appears at the beginning of the line for
CVS.EXE?

If I send a TERM signal to the ssh process, then the following happens:

 o The message "Killed by signal 15." appears in the window where I ran
   the cvs command.

 o ps now lists the ssh process as <defunct>

At this point, the only way I can kill the cvs process is to either press
ctrl-c in the shell's window, or send it a KILL signal from a different
shell.


Does this sound familiar to anyone?  Any known issues, or hopefully known
fixes?  Do you think it's a bug in ssh or a bug in cvs?  Signal handling
seems quirky on my win98 box, so I'm not sure if I'm right in assuming
that it's a ssh problem.  Oh, have TTY in my CYGWIN env variable, but I
believe I've tried it without TTY set.

I'm thinking about rebuilding debugable versions of ssh and cvs, but I
still haven't gotten an answer to my strace question
http://sources.redhat.com/ml/cygwin/2000-05/msg00384.html.



FWIW, I decided to look into this again when I saw the following in the
changes file[1] for SecureCRT 3.1:

- In SSH1, portforwarding could stall.  This was occasionally
    seen when sending large attachments via SMTP.  It was also
    seen when portforwarding CVS.

But I doubt that helps find the problem with openSSH.

[1] http://www.vandyke.com/products/securecrt/history.txt


thanks,

-jeff

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe AT sourceware DOT cygnus DOT com

- Raw text -