Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:28859
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain googlemail.com designates 66.249.92.168 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=c+36NbTJd5O234Rv1aPbmZBsQESZEnN8f+Ko9a8AJLF85IcOFBakK4vczw4xX0DIOqRZsclNrUjdh3H4NQscx9YBzRjhvpbj1nFiDiOSlIyPDFX7u9I3wGChXSeU/NmOoX4u0bckBqmQ3Rg3UpjR5qjwqoG25TM+My09RAhjNGc=
Message-ID: <10845a340704160507s3a76c55en857999cab90feabe@mail.gmail.com>
Date: Mon, 16 Apr 2007 13:07:47 +0100
Reply-To: RQuadling@GoogleMail.com
To: "William A. Rowe, Jr." <wrowe@rowe-clan.net>
Cc: "Rob Richards" <rrichards@ctindustries.net>, 
	"Ilia Alshanetsky" <ilia@prohost.org>, 
	"Rasmus Lerdorf" <rasmus@lerdorf.com>, 
	"Matt Wilmas" <php_lists@realplain.com>, internals@lists.php.net
In-Reply-To: <462365B9.9090502@rowe-clan.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>
	 <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>
	 <10845a340704160155o18e0ef3dj135113d835e633d4@mail.gmail.com>
	 <462365B9.9090502@rowe-clan.net>
Subject: Re: [PHP-DEV] Build failure
From: rquadling@googlemail.com ("Richard Quadling")

So isn't locking the solution for Windows?

On 16/04/07, William A. Rowe, Jr. <wrowe@rowe-clan.net> wrote:
> On Windows, there is no such thing as a true open-write-for-append.
>
> Unlike Unix, write for append mode is not atomic, internally it is a
> seek + write.  Mix that with multiple writers, and you have an inherent
> race condition built it.
>
>
> Richard Quadling wrote:
> > 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!"