From: Chuck Murcko <chuck@n2k.com>
Message-Id: <199701142020.PAA26870@spudboy.telebase.com>
Subject: Re: lingering_close
To: new-httpd@hyperreal.com
Date: Tue, 14 Jan 1997 15:20:15 -0500 (EST)
In-Reply-To: <Pine.BSF.3.95.970114122939.22187A-100000@alive.ampr.ab.ca> from
 "Marc Slemko" at Jan 14, 97 12:38:35 pm
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

Marc Slemko liltingly intones:
> 
> On Tue, 14 Jan 1997, Randy Terbush wrote:
> 
> > We currently _aren't_ calling lingering_close() in timeout().
> > I'm of the opinion that we should be. We call blclose() which
> > is closing the socket. We then longjump back to child_main()
> > where I _think_ that we are calling lingering_close() which
> > leads to the "Invalid argument" error coming from the error
> > code I added to the first shutdown() call in lingering_close().
> 
> I can't comment on where what is being done and where what should be done
> since I don't have the time to look through the code right now, but I
> iff NO_LINGCLOSE helps, then the problem is not likely to be due to not
> using lingering_close() in timeout() because we didn't do that
> before because it didn't exist before and it did work fine for
> people that are now haveing problems.  That doesn't necessarily
> mean what we are doing now is right or shouldn't be changed.
> 
Well, it looks like Randy's correct, from what I read in the code.
OTOH, what you say is true, too. The syndrome I see with 40-50 hits/sec
testing at home is that the server looks like it's closing everything,
or nearly so. That doesn't look right.

> Keep in mind that we may be seeing multiple problems here, since some
> don't seem to be helped by NO_LINGCLOSE.
> 
Yep. It's possible the patch I mentioned helps SGI folks who set
-DNO_LINGCLOSE.

*Heeeeey*. What happens if you use -DUSE_SO_LINGER with -DNO_LINGCLOSE?
The lingering close was supposed to replace using the socket option.
I think one or the other should be defined, but not both or neither. I'll
try this at home tonight.

chuck
Chuck Murcko	N2K Inc.	Wayne PA	chuck@telebase.com
And now, on a lighter note:
"This process can check if this value is zero, and if it is, it does
something child-like."
		-- Forbes Burkowski, Computer Science 454