Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Received-SPF: pass (athena.apache.org: domain of apache-dev@m.gmane.org
 designates 80.91.229.12 as permitted sender)
To: dev@httpd.apache.org
From: Dan Poirier <poirier@pobox.com>
Subject: Re: Per module LogLevel configuration
Date: Sun, 27 Dec 2009 08:39:48 -0600
Lines: 32
Message-ID: <yuaoclkqy4r.fsf@slappy.gateway.2wire.net>
References: <200912231314.47323.sf@sfritsch.de>
	<4B3553BC.5050900@rowe-clan.net>
	<yuabphlsrla.fsf@slappy.gateway.2wire.net>
	<4B3703C1.5040303@rowe-clan.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.1 (darwin)
Cancel-Lock: sha1:yd8o4c3McBdSBoQ9KS+XGlaP59o=
Sender: news <news@ger.gmane.org>

"William A. Rowe Jr." <wrowe@rowe-clan.net> writes:

> Dan Poirier wrote:
>> "William A. Rowe Jr." <wrowe@rowe-clan.net> writes:
>> 
>>> One thing we should refactor is 'debug' logging.  Proper debug
>>> logging is log early and often, but there is overhead involved
>>> in preparing the args and submitting the log request, only to have
>>> it fall on deaf ears.
>>>
>>> If we are doing any significant 2.0 refactoring, toggling the
>>> truly-debug log level processing at compile time would be a big win.
>> 
>> Do we have some measurements of how much overhead this adds?
>
> Of course not, since it varies wildly by what is being logged.  But just
> the preparation can consume double buffers/invoke a copy with the simple
> presence of an LF character.

Looking at log_error_core(), it appears that if the logging level is set
to disallow a particular message from being logged, that
log_error_core() returns before doing any argument processing.  So the
overhead would only consist of a few function calls, and my sense would
be that saving that wouldn't make a big difference.

Of course common sense is notoriously unreliable in making judgments
about performance :-), but this does make me even more interested in
seeing some evidence of significant improvement before we make major
changes in this area at the expense of making problem determination more
difficult.

Dan