Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:28695
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain ctindustries.net from 216.117.147.250 cause and error)
Message-ID: <46155EC9.1020402@ctindustries.net>
Date: Thu, 05 Apr 2007 16:40:41 -0400
User-Agent: Thunderbird 1.5.0.10 (Windows/20070221)
MIME-Version: 1.0
To: Rasmus Lerdorf <rasmus@lerdorf.com>
CC: RQuadling@googlemail.com, Ilia Alshanetsky <ilia@prohost.org>,
        Matt Wilmas <php_lists@realplain.com>, internals@lists.php.net
References: <007b01c77735$89410420$0201a8c0@pc1> <46148933.7030709@lerdorf.com>	 <646617A7-86A7-4E64-94CE-8E5D5766741A@prohost.org>	 <10845a340704050742n38ef04c0t7f6f2ad85e2711b5@mail.gmail.com>	 <46151547.6000505@lerdorf.com> <10845a340704050830p6611aa73n6f81806618d5072a@mail.gmail.com> <46153A67.3010208@ctindustries.net> <461541C1.4080404@lerdorf.com>
In-Reply-To: <461541C1.4080404@lerdorf.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] Build failure
From: rrichards@ctindustries.net (Rob Richards)

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

Rasmus Lerdorf wrote:
> That still doesn't make sense to me.  Could you try doing an sprintf()
> followed by an fwrite() then?  Is fprintf() perhaps doing something
> stupid internally?  I'd really like to avoid a lock here.
>
> -Rasmus
>
> Rob Richards wrote:
>   
>> I can confirm that it does fail with the single fprintf call without the
>> lock on Win32.
>> Using flock() does fix the problem in my tests.
>> I was able to fix win build using flock() instead of php_flock() and
>> defining HAVE_FLOCK in win32/flock.h
>>
>> Rob
>>
>> Richard Quadling wrote:
>>     
>>> This test is with the multiline fprintf and no lock.
>>>
>>> Adding the php_flock() has stalled the build process for Win32 (I
>>> can't build on Win32 as I don't know how!).
>>>
>>>
>>>
>>> On 05/04/07, Rasmus Lerdorf <rasmus@lerdorf.com> wrote:
>>>       
>>>> Yes, but again, is this test with the single fprintf call?  That's the
>>>> real fix for this problem, not the lock.
>>>>
>>>> -Rasmus
>>>>
>>>> Richard Quadling wrote:
>>>>         
>>>>> Using PHP 5.2.2-dev (cli) (built: Mar 23 2007 07:02:57) I can
>>>>> replicate the problem on Windows.
>>>>>
>>>>> Using this single line command at the CMD prompt:
>>>>>
>>>>> 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<100;$i++)error_log(str_repeat('%x',5000));"
>>>>
>>>>         
>>>>> The log file has many broken lines.
>>>>>
>>>>> On 05/04/07, Ilia Alshanetsky <ilia@prohost.org> wrote:
>>>>>           
>>>>>> Rasmus,
>>>>>>
>>>>>> Sorry for the delay in the reply. According to my tests on linux
>>>>>> using the sample script provided by the original bug reporter having
>>>>>> no lock causes a problem when the error message is >4k in length. In
>>>>>> this case multiple buffers are used and corruption can happen (it did
>>>>>> on a dual cpu machine with 10 error log writing threads running),
>>>>>> which is why I feel the lock is needed.
>>>>>>
>>>>>>
>>>>>> On 5-Apr-07, at 1:29 AM, Rasmus Lerdorf wrote:
>>>>>>
>>>>>>             
>>>>>>> Matt Wilmas wrote:
>>>>>>>               
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Maybe just a Windows problem if it wasn't noticed yet, but I was
>>>>>>>> compiling
>>>>>>>> the latest 5.2 snapshot and got:
>>>>>>>>
>>>>>>>> main.obj : error LNK2019: unresolved external symbol _php_flock
>>>>>>>> referenced
>>>>>>>> in function _php_log_err
>>>>>>>> Release_TS\php5ts.dll : fatal error LNK1120: 1 unresolved
>>>>>>>>                 
>>>> externals
>>>>         
>>>>>>>> Caused by this recent commit, http://news.php.net/php.cvs/43683,
>>>>>>>> and I
>>>>>>>> commented the php_flock line as a workaround.  The Windows 5.2
>>>>>>>> snapshots
>>>>>>>> haven't been updated because of this either, of course.
>>>>>>>>                 
>>>>>>> I see no reason for that lock at all as I commented when this was
>>>>>>> committed, but Ilia never replied.  This is a single write
>>>>>>> operation now
>>>>>>> since those fprintf's are now one, so that part of the fix is good,
>>>>>>> but
>>>>>>> the lock call is not needed since single writes in append mode are
>>>>>>> atomic, even on Windows.
>>>>>>>
>>>>>>> So, your work around is fine and should actually be committed.
>>>>>>>
>>>>>>> -Rasmus
>>>>>>>
>>>>>>> --
>>>>>>> PHP Internals - PHP Runtime Development Mailing List
>>>>>>> To unsubscribe, visit: http://www.php.net/unsub.php
>>>>>>>
>>>>>>>               
>>>>>> Ilia Alshanetsky
>>>>>>
>>>>>> --
>>>>>> PHP Internals - PHP Runtime Development Mailing List
>>>>>> To unsubscribe, visit: http://www.php.net/unsub.php
>>>>>>
>>>>>>
>>>>>>             
>>>>>           
>>>>         
>>>       
>
>