Subject: Re: UnixWare sockets patch - example
To: new-httpd@hyperreal.com
Date: Tue, 7 May 1996 23:02:33 +0100 (BST)
From: Ben Laurie <ben@gonzo.ben.algroup.co.uk>
In-Reply-To: <199605072220.SAA11045@telebase.com.> from "Chuck Murcko" at May
 7, 96 06:20:19 pm
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 4891      
Message-ID: <9605072302.aa20421@gonzo.ben.algroup.co.uk>
Sender: owner-new-httpd@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

Chuck Murcko wrote:
> 
> Ben Laurie liltingly intones:
> > 
> > Chuck Murcko wrote:
> > > 
> > > 2) Imagemap files must contain absolute URLs, with FQDN, etc. for virtual
> > >    hosts accessed my MSIE and Spry Mosaic. Netscape works OK.
> > > 
> I have MSIE to take home tonight and see if I can track this down on one
> of our inside servers.
> > >    ...
> > > 
> > > 4) Apache signal handling is still weird on some OS (IRIX, UnixWare).
> > >    SIGTERM doesn't, occasionally. SIGHUP sometimes ends up in server
> > >    terminating at log rollover time here, under BSDI. Partial work-
> > >    around is the log rotation util in /support, but that doesn't help
> > >    when reconfiguring servers.
> > 
> > As I've said before, I suspect that switching to sigaction instead of signal
> > may cure this ... shall I work up a patch?
> > 
> By all means!
> 
> > > 
> > > 5) UnixWare has broken socket library, which requires different options
> > >    to get working.
> > 
> > Patches?
> > 
> Here's what's come in before (connection abstractions, anyone?):
> 
> Rob forwarded:
> 
> I know this has been discussed. Can someone who knows what
> was decided, respond to this user please.

I have a horrible suspicion that SCO 5 may need this patch, too. I may try it.

Cheers,

Ben.

> 
> To: apache-bugs@apache.org
> cc: Joseph Urban <jurban@norden1.com>
> Subject: Patch to Apache 1.1b2
> Date: Sat, 27 Apr 1996 20:43:41 -0400
> From: Kimmo Suominen <kim@tac.nyc.ny.us>
> 
> ------- =_aaaaaaaaaa0
> Content-Type: text/plain; charset="us-ascii"
> Content-ID: <15351.830652000.1@grendel.gw.com>
> 
> Dear Sirs,
> 
> On most SVR4/386 systems based on USL SVR4.0 there is a bug in the STREAMS
> implementation of sockets.  When a socket is closed, any pending data will
> be lost because the STREAMS structures holding the data are destroyed.  If
> the SO_LINGER option is set on the socket with a long enough time, one can
> usually avoid the problem.
> 
> I've attached the necessary patch below.  I use it on Dell SVR4 Issue 2.2,
> but it is necessary for any issue of Dell SVR4, Esix, early UnixWare and
> other USL SVR4.0/386 based systems.
> 
> I'm hoping you would include this patch in the release version of Apache.
> I've submitted the same fix for 1.0.3 earlier.  I am not using the proxy
> module, but if it has outbound sockets created, the same fix is needed in
> all those places.  I would recommend writing a small function that returns
> a socket with all the proper options set, and using that in all modules.
> 
> The Makefile flags I use are:
> 
>     AUX_CFLAGS= -DSVR4 -DLINGER
>     AUX_LIBS= -lsocket -lnsl
> 
> Sincerely,
> Kimmo Suominen
> 
> ------- =_aaaaaaaaaa0
> Content-Type: text/plain; charset="us-ascii"
> Content-ID: <15351.830652000.2@grendel.gw.com>
> Content-Description: Patch to src/http_main.c
> 
> Index: src/http_main.c
> ===================================================================
> RCS file: /home/grendel/cvs/apache/src/http_main.c,v
> retrieving revision 1.1.1.2
> retrieving revision 1.3
> diff -u -r1.1.1.2 -r1.3
> --- http_main.c	1996/04/25 21:45:48	1.1.1.2
> +++ http_main.c	1996/04/25 22:14:03	1.3
> @@ -1240,16 +1240,31 @@
>      
>      if((setsockopt(s, SOL_SOCKET,SO_REUSEADDR,(char *)&one,sizeof(one)))
>         == -1) {
> -        perror("setsockopt");
> -        fprintf(stderr,"httpd: could not set socket option\n");
> +        perror("setsockopt(SO_REUSEADDR)");
> +        fprintf(stderr,"httpd: could not set socket option SO_REUSEADDR\n");
>          exit(1);
>      }
>      if((setsockopt(s, SOL_SOCKET,SO_KEEPALIVE,(char *)&keepalive_value,
>          sizeof(keepalive_value))) == -1) {
> -        perror("setsockopt"); 
> +        perror("setsockopt(SO_KEEPALIVE)"); 
>          fprintf(stderr,"httpd: could not set socket option SO_KEEPALIVE\n"); 
>          exit(1); 
>      }
> +
> +#ifdef LINGER   /* If puts don't complete, you could try this. */
> +    {
> +	struct linger li;
> +	li.l_onoff = 1;
> +	li.l_linger = 900;
> +
> +	if (setsockopt(s, SOL_SOCKET, SO_LINGER,
> +	    (char *)&li, sizeof(struct linger)) < 0) {
> +	    perror("setsockopt(SO_LINGER)");
> +	    fprintf(stderr,"httpd: could not set socket option SO_LINGER\n");
> +	    exit(1);
> +	}
> +    }
> +#endif  /* LINGER */
>  
>      if(bind(s, (struct sockaddr *)server,sizeof(struct sockaddr_in)) == -1)
>      {
> 
> ------- =_aaaaaaaaaa0--
> 
> 
> 
> chuck
> Chuck Murcko	N2K Inc.	Wayne PA	chuck@telebase.com
> And now, on a lighter note:
> Aleph-null bottles of beer on the wall,
> Aleph-null bottles of beer,
> 	You take one down, and pass it around,
> Aleph-null bottles of beer on the wall.

-- 
Ben Laurie                  Phone: +44 (181) 994 6435
Freelance Consultant and    Fax:   +44 (181) 994 6472
Technical Director          Email: ben@algroup.co.uk
A.L. Digital Ltd,           URL: http://www.algroup.co.uk
London, England.