From owner-freebsd-hackers  Sat Jun 21 02:26:30 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id CAA20741
          for hackers-outgoing; Sat, 21 Jun 1997 02:26:30 -0700 (PDT)
Received: from scanner.worldgate.com (scanner.worldgate.com [198.161.84.3])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id CAA20736
          for <hackers@FreeBSD.ORG>; Sat, 21 Jun 1997 02:26:25 -0700 (PDT)
Received: from znep.com (uucp@localhost)
	by scanner.worldgate.com (8.8.5/8.8.5) with UUCP id DAA03471;
	Sat, 21 Jun 1997 03:26:18 -0600 (MDT)
Received: from localhost (marcs@localhost) by alive.znep.com (8.7.5/8.7.3) with SMTP id DAA23462; Sat, 21 Jun 1997 03:26:49 -0600 (MDT)
Date: Sat, 21 Jun 1997 03:26:48 -0600 (MDT)
From: Marc Slemko <marcs@znep.com>
To: sthaug@nethelp.no
cc: hackers@FreeBSD.ORG
Subject: Re: gethostbyname() and fds >255 
In-Reply-To: <23110.866883503@verdi.nethelp.no>
Message-ID: <Pine.BSF.3.95.970621031533.21696N-100000@alive.znep.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-hackers@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

On Sat, 21 Jun 1997 sthaug@nethelp.no wrote:

> > > For comparison, it works fine on BSD/OS 2.1.  Does it work in -current? 
> > 
> > It would appear to:
> > 
> > 396: host = 198.161.84.2 (valis.worldgate.com)
> > 397: host = 198.161.84.2 (valis.worldgate.com)
> > 398: host = 198.161.84.2 (valis.worldgate.com)
> > 399: host = 198.161.84.2 (valis.worldgate.com)
> > root@make-> uname -a
> > FreeBSD make.ican.net 3.0-CURRENT FreeBSD 3.0-CURRENT #0: Mon Jun 16 13:10:14 EDT 1997     jkh@make.ican.net:/a/local/src/sys/compile/MAKE  i386
> > 
> > It does indeed fall over on my 2.2-stable system.
> 
> The problem is in res_send(). In lib/libc/net/res_send.c line 577 in
> FreeBSD-current we have:

Sigh.  Yes, that's it.  I found it an hour or so ago looking through
things.  For a test, rebuilding libc with FD_SETSIZE=1024 then linking it
against my test program worked fine.  

That isn't very nice.  It appears the only reason it works in -current
(well, it will break at 1024 fds there...) is that FD_SETSIZE has been
increased.

Guess I'll have to double check what else was broken when FD_SETSIZE was
increased and make a patch for 2.2 that I can point people to.  My biggest
concern in this is that Apache 1.2 will not work with >~255 fds used for
logs without recompiling libc with a larger FD_SETSIZE.