Mailing-List: contact dev-help@apr.apache.org; run by ezmlm
Precedence: bulk
Received-SPF: pass (asf.osuosl.org: local policy)
Date: Sun, 18 Sep 2005 23:27:32 +0100
From: Colm MacCarthaigh <colm@stdlib.net>
To: dev@apr.apache.org
Subject: Re: [PATCH] Configurable read/write buffer sizes
Message-ID: <20050918222732.GA27486@stdlib.net>
Reply-To: colm@stdlib.net
References: <20050906204911.GA5989@stdlib.net>
 <20050907083808.GC4636@redhat.com> <20050907092208.GA10137@stdlib.net>
 <20050909135740.GA2739@stdlib.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-15
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20050909135740.GA2739@stdlib.net>
User-Agent: Mutt/1.5.9i


Anyone had a change to look at any of the below? (Thanks for committing
the docs nit Garrett).

01-flags-stdio.patch is the one I actually have a dependency on right
now, and is also relatively simple. I'd really appreciate some feedback
on that. 

I can live with arguing a case for variable buffer sizes. If benchmarks
will help; I can run a tonne. I've seeing some pretty nifty speed-ups
when I apply large buffer sizes to parts of httpd (and sometimes even
suprising places, like rotatelogs).

If it helps, I'm usually in #apr, feel free to bug me there. 

On Fri, Sep 09, 2005 at 02:57:40PM +0100, Colm MacCarthaigh wrote:
> O.k., I've done some new patches and a few read throughs to think the
> implications of a large buffer size out, and implemented a flags-capable
> variant of the stdin/out/err openers. 
> 
> I believe apr_size_t is safe, apart from the circumstance where someone
> tries to assign a vlue to the buffer size that would overflow an
> apr_size_t anyway, so not worth worrying about.
> 
> The patches are all at;
> 
>     http://people.apache.org/~colm/apr-IO-patches/
> 
> They are all against trunk, but can be applied in sequence, there's only
> one slight offset change that patch has to figure out.
> 
>     001-flags-stdio.patch 	 Implement apr_file_open_flags_std(err|out|in)
>                                 Nothing major, but I've also set APR_READ
>                                 or APR_WRITE as appropriate, purely for the
>                                 sake of apr_file_flags_get();
> 
>     002-set-buffer.patch        APR-wide changes to implement 
>                                 apr_file_set_buffer() and apr_file_buffer_size_get(),
>                                 includes a testcase
> 
>     003-set-buffer-unix.patch   Unix implementation of same, tested on Ubuntu
> 
>     004-set-buffer-win32.patch  Win32 implementation, tested w2k3. I didn't include
>                                 the changes to the dsp files, because VS2005 to behave.
> 
>     005-set-buffer-netware.patch Netware header changes neccessary, untested
> 
>     006-set-buffer-os2.patch    OS implementation, untested, based on inspection	
> 
> -- 
> Colm MacC�rthaigh                        Public Key: colm+pgp@stdlib.net
> 

-- 
Colm MacC�rthaigh                        Public Key: colm+pgp@stdlib.net