Mailing-List: contact cygwin-help@sourceware.cygnus.com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe@sources.redhat.com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin@sources.redhat.com>
List-Help: <mailto:cygwin-help@sources.redhat.com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner@sources.redhat.com
Delivered-To: mailing list cygwin@sources.redhat.com
Message-ID: <00d101c14cab$929c9b20$01000001@lifelesswks>
From: "Robert Collins" <robert.collins@itdomain.com.au>
To: "Stipe Tolj" <tolj@wapme-systems.de>
Cc: <cygwin@cygwin.com>
References: <3BBB6E37.DD0CD8BF@wapme-systems.de> <039001c14c57$9548da90$01000001@lifelesswks> <3BBC171E.884AB8CF@wapme-systems.de>
Subject: Re: apache-1.3.x strange behaviour on keep-alives
Date: Thu, 4 Oct 2001 18:06:48 +1000
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4133.2400
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400

----- Original Message -----
From: "Stipe Tolj" <tolj@wapme-systems.de>
To: "Robert Collins" <robert.collins@itdomain.com.au>
>
> > If you have the time to search for this, what you do is
> > * setup a cygwin1.dll build environment from CVS (as per the cygwin
web
> > site).
> > * pick a date - say half-way between 1.3.2 and 1.3.3.
> > use CVS -z3 up -Pd -D "10th June 2001" to get the CVS repository
udpated
> > to that specific date. (run this command in src/winsup, and in
> > src/newlib).
> > run make in the build dir.
> > copy new-cygwin1.dll to /usr/bin/cygwin1.dll
> > Test.
> > Iterate moving the date forward or backwards, in a binary search
pattern
> > (ie 1/4 of the date distance between 1.3.3 and 1.3.4, and the next
step
> > is 1/8th and so on.
>
> Thanks for the instructions. The problem is that the effect is shown
> after a couple of days incoming HTTP requests to the httpd. Usually
> you can not simulate it using benchmark tools like apache's ab :(( --
> so an interative approach to get back to a working DLL versions seems
> very inconvienent.

Yes thats true. That is in fact why I have not had the time to do it
yet.

> That's why I am trying to ask the Cygwin core folks to assist me.

I'm one of them. I'm assisting you :}.

The first thing _someone_ has to do is identify _when_ the bad behaviour
was introduced. Then it's realistic to scream "<cvs committers name>"
Why, Oh Why, did you do this to apache and squid!!. Until then, there is
no point.

If you want this solved, build a .dll from 1/2 way between, and then
install it. If it doesn't show the symptom, go forward 1/4 the time.

I'm willing to bet that only four iterations will be needed to isolate a
single ChangeLog entry that affects sockets/tcp. At that point,even if
there are more changes in the time period, it's pretty certain which one
will be the culprit.

Rob


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/