Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:28853
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain googlemail.com designates 66.249.82.234 as permitted sender)
DomainKey-Status: bad
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=googlemail.com; s=beta;
        h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references;
        b=L9AKxGzgmdkD3VNa+R6qxVSR5VYutSCkACCUjq1eKKQfEkfS+SZWjatXX0AeFGfHjxMe/RjctdhLGG8m8+lBz+2PXPX0M2f6Rwb7z6fpxsb6e0S3jC+a6B7xA2bneEHtU4MbTXqrUGtGJEciUuHGcbmOH0NV2ZQ8EjovKgFas6k=
Message-ID: <10845a340704160155o18e0ef3dj135113d835e633d4@mail.gmail.com>
Date: Mon, 16 Apr 2007 09:55:29 +0100
Reply-To: RQuadling@GoogleMail.com
To: "Rob Richards" <rrichards@ctindustries.net>
Cc: "Ilia Alshanetsky" <ilia@prohost.org>, "Rasmus Lerdorf" <rasmus@lerdorf.com>, 
	"Matt Wilmas" <php_lists@realplain.com>, internals@lists.php.net
In-Reply-To: <4620B813.8080306@ctindustries.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <007b01c77735$89410420$0201a8c0@pc1> <46151547.6000505@lerdorf.com>
	 <10845a340704050830p6611aa73n6f81806618d5072a@mail.gmail.com>
	 <46153A67.3010208@ctindustries.net> <461541C1.4080404@lerdorf.com>
	 <46155EC9.1020402@ctindustries.net>
	 <10845a340704130333i17921209pad4538b5a3a43e7@mail.gmail.com>
	 <3FD3996A-6E63-44BF-88EF-CEB174254995@prohost.org>
	 <10845a340704130642w7c4edf5an8bc41d49c4e1dd86@mail.gmail.com>
	 <4620B813.8080306@ctindustries.net>
Subject: Re: [PHP-DEV] Build failure
From: rquadling@googlemail.com ("Richard Quadling")

That's good news. Has a patch been submitted? Even if it is initially
only a win32 compiler directive wrapping the lock?

On 14/04/07, Rob Richards <rrichards@ctindustries.net> wrote:
> I can sometimes reproduce the missing entries though never get data
> corruption anymore.
> This might now be a Windows only issue with how it is caching writes. I
> did a little debugging and no errors occurred.
> The file was opened successfully every time, the data was written
> (correct number of bytes as well) every single time, but depending upon
> the load on my system at the time it was a crap shoot whether every
> single write actually made it into the physical file or not. Adding a
> lock here did resolve it so that it worked 100% of the time.
>
> Rob
>
> Richard Quadling wrote:
> > So why are there missing entries?
> >
> > I can even get the corruption back again if I use a shorter line (100
> > rather than 5000).
> >
> >
> > On 13/04/07, Ilia Alshanetsky <ilia@prohost.org> wrote:
> >> The new implementation does not use any locks, instead it uses direct
> >> io, where locks are not necessary for append operations.
> >>
> >>
> >> On 13-Apr-07, at 6:33 AM, Richard Quadling wrote:
> >>
> >> > On 05/04/07, Rob Richards <rrichards@ctindustries.net> wrote:
> >> >> No difference using sprintf()/fwrite() instead of fprintf().
> >> >>
> >> >> I did come across a similar issue from apache:
> >> >> http://mail-archives.apache.org/mod_mbox/httpd-dev/199503.mbox/%
> >> >> 3C9502272049.aa02036@paris.ics.uci.edu%3E
> >> >>
> >> >> Changing to use VCWD_OPEN_MODE, write() and close() seems to work.
> >> >>
> >> >> Rob
> >> >>
> >> >
> >> > Using PHP 5.2.2RC2-dev (cli) (built: Apr 13 2007 04:03:02) on
> >> > Windows with
> >> >
> >> > for %x in (A B C D E F G H I J K L M N O P Q R S T U V W X Y Z) do
> >> > start php -r "ini_set('error_log','/tmp/test.log');for($i=0;$i<1000;
> >> > $i++)error_log(str_repeat('%x',5000));"`;
> >> >
> >> > And then doing some analysis of the log file I'm finding that not all
> >> > the data has been written.
> >> >
> >> > Array
> >> > (
> >> >    [A] => 4965993
> >> >    [B] => 4590918
> >> >    [C] => 4525905
> >> >    [D] => 4695939
> >> >    [E] => 4495899
> >> >    [F] => 4710942
> >> >    [G] => 4495899
> >> >    [H] => 4645929
> >> >    [I] => 4540908
> >> >    [J] => 4580916
> >> >    [K] => 4535907
> >> >    [L] => 4470894
> >> >    [M] => 4480896
> >> >    [N] => 4550910
> >> >    [O] => 4610922
> >> >    [P] => 4500900
> >> >    [Q] => 4630926
> >> >    [R] => 4480896
> >> >    [S] => 4500900
> >> >    [T] => 4535907
> >> >    [U] => 4630926
> >> >    [V] => 4470894
> >> >    [W] => 4645929
> >> >    [X] => 4825965
> >> >    [Y] => 4845969
> >> >    [Z] => 4920984
> >> > )
> >> > Lines (Should be 26 * 1000) : 23973
> >> > Longest (Should be 5024) : 5024
> >> > Shortest (Should be 5024) : 5024
> >> >
> >> > Out of the 26,000 lines expected we are missing 2027 lines.
> >> >
> >> > So no corruption of the lines, just missing ones.
> >> >
> >> > The only explanation I can think of is that the locking is working,
> >> > but nothing is waiting for the lock to become available.
> >> >
> >> > Richard.
> >> >
> >> >
> >> > --
> >> > -----
> >> > Richard Quadling
> >> > Zend Certified Engineer : http://zend.com/zce.php?
> >> > c=ZEND002498&r=213474731
> >> > "Standing on the shoulders of some very clever giants!"
> >>
> >> Ilia Alshanetsky
> >>
> >>
> >>
> >>
> >>
> >
> >
>


-- 
-----
Richard Quadling
Zend Certified Engineer : http://zend.com/zce.php?c=ZEND002498&r=213474731
"Standing on the shoulders of some very clever giants!"