Mailing-List: contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
Sender: cygwin-owner AT sourceware DOT cygnus DOT com
Delivered-To: mailing list cygwin AT sourceware DOT cygnus DOT com
From: Chris Faylor <cgf AT cygnus DOT com>
Date: Thu, 25 Mar 1999 14:09:14 -0500
To: Richard DOT Hirst AT net-tel DOT co DOT uk
Cc: cygwin AT sourceware DOT cygnus DOT com
Subject: Re: B20.1: Returning from pause() before signal handler runs?
Message-ID: <19990325140914.A3064@cygnus.com>
References: <"SLOE:0092-990325140215-0004*/G=Richard/S=Hirst/O=NET-TEL Computer Systems Ltd/PRMD=NET-TEL/ADMD=GOLD 400/C=GB/"@MHS>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-Mailer: Mutt 0.95.3i
In-Reply-To: <"SLOE:0092-990325140215-0004*/G=Richard/S=Hirst/O=NET-TEL Computer Systems Ltd/PRMD=NET-TEL/ADMD=GOLD 400/C=GB/"@MHS>; from Richard.Hirst@net-tel.co.uk on Thu, Mar 25, 1999 at 02:02:53PM +0000

On Thu, Mar 25, 1999 at 02:02:53PM +0000, Richard DOT Hirst AT net-tel DOT co DOT uk wrote:
>Hi,
>  I am running "sendmail -q1" under B20.1 on NT4.0 SP3.  This should poll a
>queue directory once a minute, but I find that after a few polls it goes to
>sleep.  I have provided a small test program which shows the problem.
>The parent installs ALRM and CHLD signal handlers, forks a child,
>requests an alarm tick., and calls pause().  The child exits.  Sometimes
>the child exits before the parent calls pause(), sometimes not.  The
>signal handlers increment counts each time they run.  The parent notes
>these counts before calling pause, and again after returning from pause().
>Sometimes the parent finds that the counts havn't changed; it does a
>printf() and checks the counts again - now they have changed.
>
>It looks to me as though the parent is returning from pause() before
>the signal handler has been run.  
>
>I can provoke the problem within a few cycles if I run a couple of
>copies of the program and then start up a few other programs (DevStudio,
>MSDOS boxes, etc).  This is on a PII-233 with 64MB, NT4.0 SP3.
>
>If anyone could reproduce the problem, and say whether it looks like
>a cygwin bug or not, I'd appreciate it.

Looking at the code, it does appear to be a bug.  I've fixed the problem
and it will be in the next snapshot.

cgf

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