Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-developers-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin-developers/>
List-Post: <mailto:cygwin-developers AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-developers-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-developers-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin-developers AT sources DOT redhat DOT com
Date: Wed, 25 Jul 2001 21:32:44 -0400
From: Christopher Faylor <cgf AT redhat DOT com>
To: cygwin-developers AT cygwin DOT com
Subject: Re: close-on-exec handles are left open by exec parent
Message-ID: <20010725213244.B14002@redhat.com>
Reply-To: cygwin-developers AT cygwin DOT com
Mail-Followup-To: cygwin-developers AT cygwin DOT com
References: <9715860566 DOT 20010724154632 AT logos-m DOT ru>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.3.11i
In-Reply-To: <9715860566.20010724154632@logos-m.ru>; from deo@logos-m.ru on Tue, Jul 24, 2001 at 03:46:32PM +0400

On Tue, Jul 24, 2001 at 03:46:32PM +0400, egor duda wrote:
>test.exe create pipe and forks. when write end of the pipe is marked
>as close-on-exec, we DuplicateHandle () appropriately. but forked
>instance left running after execlp () is performed, and it doesn't
>close it's own copy of pipe handle. so read () blocks forever.
>
>i can see 2 ways to fix it. We can either always perform
>dtable::fixup_before_exec() and close handles marked as close-on-exec
>there, or always close all handles when exec child is started and exec
>parent cycles waiting for its child to exit.

Nice analysis.  I think the second is the way to deal with this.  The first
assumes that the exec will be successful, which is not necessarily true.
fixup_before_exec is a compromise for sockets.  I'd rather not have to
do this for other fds, too.

So, the correct fix is to close the handles after a successful CreateProcess.

I can't believe that this bug has been around for so long.  Of course, there
was a time (in B20.1, I think <gasp>) where close-on-exec didn't work at
all for non-cygwin processes...

cgf