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
Date: Sat, 14 Dec 2002 00:27:44 -0500
From: Christopher Faylor <cgf AT redhat DOT com>
To: cygwin AT cygwin DOT com
Subject: More pipe (and other) improvements in snapshot
Message-ID: <20021214052744.GA17108@redhat.com>
Reply-To: cygwin AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
References: <20021211041250 DOT GA31215 AT redhat DOT com> <094401c2a1da$954ab980$1d6cba8c AT sfdev3> <20021212165736 DOT GA14986 AT redhat DOT com> <H000076f17389de6 DOT 1039774000 DOT eux100 DOT sgp DOT st DOT com AT MHS> <3DF9F33C DOT 6060906 AT st DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <3DF9F33C.6060906@st.com>
User-Agent: Mutt/1.5.1i

On Fri, Dec 13, 2002 at 03:48:28PM +0100, Pavel Holejsovsky wrote:
>nhv AT cape DOT com wrote:
>>Christopher Faylor wrote:
>>>On Thu, Dec 12, 2002 at 07:32:16AM -0500, Norman Vine wrote:
>>>>Any 'tips' as to how to best debug this appreciated 
>>>
>>>- Attach to the hung process with gdb and see where it is hung.
>>>
>>>- Provide cygcheck output.
>>>
>>>- Run under strace and see if you can infer where hangs or problems
>>> are occurring.
>
>The hung process (sed) is actually not hung, but connected to stdin 
>instead of file. The root cause is that when config.status is processed 
>by bash, then sometimes `` construct forgets all output, causing 
>different generated file names to be empty, thus connecting sed to stdin 
>instead of some file. Following command line reproduces the bug for me:
>
>while true; do test "`echo foo`" = "foo" || echo failed; done
>
>When this command is run inside bash, then "failed" lines appear quite 
>regularly. The bug is reproducible only with /bin/bash, /bin/sh works 
>reliably.

I meant to test the above before I made some more changes to cygwin
to attempt to avoid data loss but I didn't get around to it.

So, there is a snapshot up there which may or may not solve the problem.
The above command doesn't die for me now, so I guess that's something.

I'll check the prevous snapshot tomorrow to see if it actually fails for
me.  I've been a little sick for a couple of weeks now and attempting to
stay with my "work on real job in the day and cygwin at night" scenario
is not working too well.

In the meantime, please try the current snapshot.  It is probably
slightly (but possibly unoticeably) slower but it should be very much
less likely to lose data in a pipe.

The latest snapshot also has some new inline assembly functions courtesy
of some cygwin-developers (Gary R.  Van Sickle and Thomas Pfaff) which
might cause a little bit of a boost, too.

cgf

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/