From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S267671AbUHMXHm (ORCPT ); Fri, 13 Aug 2004 19:07:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S267726AbUHMXHm (ORCPT ); Fri, 13 Aug 2004 19:07:42 -0400 Received: from the-village.bc.nu ([81.2.110.252]:50906 "EHLO localhost.localdomain") by vger.kernel.org with ESMTP id S267671AbUHMXHa (ORCPT ); Fri, 13 Aug 2004 19:07:30 -0400 Subject: Re: select implementation not POSIX compliant? From: Alan Cox To: Nick Palmer Cc: Manfred Spraul , Linux Kernel Mailing List In-Reply-To: <411D2227.2060500@sluggardy.net> References: <411A8646.1030205@colorfullife.com> <411D2227.2060500@sluggardy.net> Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1092434702.24989.23.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Fri, 13 Aug 2004 23:05:02 +0100 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Gwe, 2004-08-13 at 21:18, Nick Palmer wrote: > Actually Solaris and Linux are consistent in terms of the behavior of > select in this respect. I suspect that the first select is blocking the > socket from being used at all, so the second select can't tell that it > is closed. The objects are refcounted so the socket hasn't gone away until the point the select returns. Alan