Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe AT cygwin DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin AT cygwin DOT com>
List-Help: <mailto:cygwin-help AT cygwin DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
Delivered-To: mailing list cygwin AT cygwin DOT com
Message-ID: <3E9C73FB.F74E775A@precidia.com>
Date: Tue, 15 Apr 2003 17:04:59 -0400
From: Brian White <bcwhite AT precidia DOT com>
Organization: Precidia Technologies  http://www.precidia.com/
X-Accept-Language: en
MIME-Version: 1.0
To: Max Bowsher <maxb AT ukf DOT net>
CC: cygwin AT cygwin DOT com
Subject: Re: tcp RST instead of FIN if child exits after parent closes path
References: <3E9C374E DOT 4256358D AT precidia DOT com> <004001c30372$9b95d9b0$235e893e AT pomello> <3E9C42AF DOT F8A6AEA1 AT precidia DOT com> <008701c30386$64cc56c0$5c51893e AT pomello>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

> >> This is a bug/feature of Windows Sockets. I think it would be possible to
> >> craft a workaround in Cygwin, but it would require some non-trivial IPC.
> >> (IIUC, you would need to maintain a cross-process refcount for the
> socket).
> >
> > Hmmm...  From my work with Amanda, the child process can continue to
> > send data via the socket even after the parent has closed it's handle
> > on that path, so there must be some sort of reference counting already.
> 
> Yes, but the wrong sort.
> 
> To guarantee avoiding the RST, you need to shutdown(SHUT_WR) the socket. And
> that does stop all processes from writing to the socket. So, you would need
> to refcount the socket, and issue a shutdown just before the last close. (I
> think.)

I don't follow.  The path is still open and able to send the data from the
child even though the parent has closed it.  It's only when the child exits
(note: the child just exited; it didn't do an explicit "close") that the
connection was killed, so there the system already knows when the last close
occurs.  What is being done to the socket that is different depending upon
whether the final close comes from the process that originally opened it or
another one?


> For a quick hack, you could try sleeping for a second before exiting. Should
> work most of the time.

For Amanda, I fixed it by moving the close within the main application until
after all the messages from the child had been parsed and the child had
exited.

                                          Brian
                                 ( bcwhite AT precidia DOT com )

-------------------------------------------------------------------------------
"Everything that can be invented already has been."  --U.S. Patent Office, 1899

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/