Re: [Quickfix-developers] MS VC++ release builds

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Also, at the Dinkumware page on VC fixes (VC6 uses an old version of 
dinkumware), they give a couple of suggestions for circumventing the 
problem, one of which is to disable reference counting: 
http://www.dinkumware.com/vc_fixes.html

See the section on Fixes to <xstring>.  This will of course likely 
affect performance.

--oren

On Aug 24, 2004, at 4:52 AM, Antonio Caroselli wrote:

> Hi, Yihu and Tim.
>
> As you might have seen from my colleage's email to this newsgroup: 
> "****
> VERY IMPORTANT ***** Crashes in QuickFIX compiled with MS VC++ 6.0" 
> sent out
> yesterday, the reason for memory corruption and random crashes is the 
> use of
> STL strings in the QuickFix library when running our application in a
> multithreading environment (see
> http://support.microsoft.com/default.aspx?scid=kb;en-us;813810 ).
>
> We have been using MS VC++ 6 - SP5 to compile and link our application,
> which has been deployed under Windows 2000 SP4 on a computer with two 
> XEON
> processors plus hyperthreading enabled.
>
> Both Oren and Microsoft suggest to try a different STL from different
> suppliers (www.stlport.org [FREE] and www.dinkumware.com [NOT FREE]), 
> which
> might be a good choice if we want to keep using MS VC++ 6. On the 
> contrary,
> we might compile our application with Visual Studio .NET which seems 
> to have
> a thread-safe STL. Probably we shall try both options and go through a 
> new
> set of stress tests before taking a final decision.
>
> Anyway, we would avoid those code manipulations suggested by Microsoft 
> since
> that might unearth new problems in other areas.
>
> Bye,
>
> Antonio Caroselli
> GATE T.I.
> www.gatelab.com
>
>
>
> ----- Original Message -----
> From: "Yihu Fang" <Yih...@re...>
> To: "Timothy Yates" <Tim...@pa...>; "Antonio Caroselli"
> <ant...@ga...>
> Cc: "Oren Miller" <or...@qu...>;
> <qui...@li...>
> Sent: Friday, August 20, 2004 9:10 PM
> Subject: RE: [Quickfix-developers] MS VC++ release builds
>
>
> QuickFIX Documentation:
> http://www.quickfixengine.org/quickfix/doc/html/index.html
> QuickFIX FAQ: 
> http://www.quickfixengine.org/wikifix/index.php?QuickFixFAQ
> QuickFIX Support: http://www.quickfixengine.org/services.html
>
> Hi Antonio and Tim,
>
> It would be interesting to see what the real cause of the problems.
>
> We have been using MS VC7 compiler (Visual Studio .NET 2003) on Windows
> 2003 and XP and did not see the corrupted fields and reject messages.
> Maybe we should examine them more carefully.
>
> A few questions: (1) if the QuickFIX log shows correct FIX messages
> while the application shows error and rejects message, does it imply
> something wrong in the implementation of the application? (2) Is the
> application using the same compiler options as the QuickFIX library,
> e.g. multiplethread dll settings etc? (3) Is there any memory leak seen
> in the application after a long period of time?
>
> Thanks.
>
> -Yihu
>
> -----Original Message-----
> From: qui...@li...
> [mailto:qui...@li...] On Behalf Of
> Timothy Yates
> Sent: Friday, August 20, 2004 9:21 AM
> To: 'Antonio Caroselli'
> Cc: 'Oren Miller'; qui...@li...
> Subject: RE: [Quickfix-developers] MS VC++ release builds
>
> QuickFIX Documentation:
> http://www.quickfixengine.org/quickfix/doc/html/index.html
> QuickFIX FAQ:
> http://www.quickfixengine.org/wikifix/index.php?QuickFixFAQ
> QuickFIX Support: http://www.quickfixengine.org/services.html
>
> Hi Antonio,
>
> I mentioned two distinct problems in my original email.
>
> The first of these related to random corruptions of field data in FIX
> messages, especially during periods of heavy traffic.  These problems
> occurred in an MS VC++ quickfix application.  Once or twice, we saw the
> same
> CompID error you refer to. Typically, the corrupted field data 
> contained
> parts of other FIX fields from the same message.
>
> We were never able to track down the cause of these corruptions -- we
> found
> the problem very hard to reproduce. However, turning off compiler
> optimizations did seem to improve the situation considerably.  If you
> can
> find out the cause of this probem, I would be very interested.
>
> The second problem related to the Java version of QuickFIX on Windows
> XP.
> This problem seems to be related to hyperthreading (possibly in
> combination
> with hot-spot compilation).  When I disabled hyperthreading on the host
> machine, this problem went away -- it doesn't seem to be related to the
> first problem as I initially thought.
>
> Tim
>
> -----Original Message-----
> From: Antonio Caroselli [mailto:ant...@ga...]
> Sent: Friday, August 20, 2004 3:53 AM
> To: Timothy Yates
> Cc: 'Oren Miller'; qui...@li...
> Subject: Re: [Quickfix-developers] MS VC++ release builds
>
>
> Dear Tim,
>
> We are experiencing the same problems described in your email dated 
> June
> 14/2004.
>
> We are using MS VC++ 6 SP5 with default compilation and link options 
> and
> QuickFix 1.7.1. The target application is being run under Windows 
> Server
> 2000 SP4.
>
> In most cases we detected memory errors on incoming messages: we often
> get
> the error "Message nn rejected: CompID problem" even if the FIX message
> logged in our log file via the QuickFix log function shows correct tags
> and
> values. It may happen that we have such a problem once or twice in a 
> day
> under heavy traffic conditions, mostly on incoming messages. In few
> cases
> this possible memory problem also caused outgoing messages to be
> truncated.
>
> At times (not always) this problem is followed by an application crash:
> we
> do not get any application fault window since the application simply
> disappears from the Task Manager.
>
> I have seen Oren's suggestion about the callstack. Did that help you
> solve
> your problem ? We might try and use QuickFix 1.8.0, but the release
> notes do
> not seem to address this problem.
>
> Any help will be greatly appreciated
>
> Bye,
>
> Antonio Caroselli
> GATE T.I.
> www.gatelab.com
>
>
>
> ----- Original Message -----
> From: "Timothy Yates" <Tim...@pa...>
> To: "'Oren Miller'" <or...@qu...>
> Cc: <qui...@li...>
> Sent: Monday, June 14, 2004 5:30 PM
> Subject: RE: [Quickfix-developers] MS VC++ release builds
>
>
>> QuickFIX Documentation:
> http://www.quickfixengine.org/quickfix/doc/html/index.html
>> QuickFIX FAQ: http://www.quickfixengine.org/quickfix/doc/html/FAQ.html
>>
>> The callstack is not enabled.
>>
>> Tim.
>>
>> -----Original Message-----
>> From: Oren Miller [mailto:or...@qu...]
>> Sent: Monday, June 14, 2004 9:32 AM
>> To: Timothy Yates
>> Cc: qui...@li...
>> Subject: Re: [Quickfix-developers] MS VC++ release builds
>>
>>
>> Do you have the callstack enabled?
>>
>> --oren
>>
>> On Jun 14, 2004, at 8:39 AM, Timothy Yates wrote:
>>
>>> QuickFIX Documentation:
>>> http://www.quickfixengine.org/quickfix/doc/html/index.html
>>> QuickFIX FAQ:
> http://www.quickfixengine.org/quickfix/doc/html/FAQ.html
>>>
>>> Has anyone else had problems with apparently random crashes and
> memory
>>> corruptions in quickfix when using MS VC++, specifically using the
>>> default
>>> release build options?
>>>
>>> We have a C++ production application that uses quickfix.  We have
> had
>>> to
>>> build quickfix with all optimizations turned off.  We were unable to
>>> track
>>> down the memory corruptions that occurred with a release build.
>>>
>>> Recently, I have been using the Java version of quickfix.  I have a
>>> test
>>> application that generates MarketDataSnapshotFullRefresh messages
> and
>>> populates them with essentially random price data.  The receiving
>>> application sends a subscribe message and the test application
> starts
>>> sending refresh messages at a rate of about 20 per second.  When
> using
>>> a
>>> release build of quickfix_jni.dll, the receiving application often
>>> crashes
>>> after several hundred messages due to an access violation.  This
>>> typically
>>> occurs in the destructor call from Message.finalize(), other times
> in
>>> Group.finalize(), and othertime some place else.   I do a session
> reset
>>> between test runs.  Sometimes the application does not crash at all.
>>> I
>>> have tried disabling all processing of refresh messages by the
>>> receiving
>>> application (so it just returns from fromApp without doing
> anything),
>>> and
>>> the crashes still happen.
>>>
>>> The random nature of the crashes leads me to suspect a threading
> issue.
>>>
>>> When I use a debug build of quickfix_jni.dll, I can never get it to
>>> crash.
>>> This might just be luck, or it might indicate a problem with
> compiler
>>> optimizations.
>>>
>>> I think this problem may have something to do with repeating groups
> as
>>> I
>>> have only started noticing it since implementing market data
> support.
>>> However, it could just be that I have started stressing quickfix
> more
>>> than
>>> before.
>>>
>>> Interestingly, the corruptions we were experiencing in our
> production
>>> C++
>>> application also occurred when there was a large amount of market
> data
>>> traffic.
>>>
>>> I am using quickfix 1.7.0.
>>>
>>> Tim Yates
>>> Lead Developer
>>> Patsystems (US) LLC
>>> 141 West Jackson Boulevard
>>> Chicago 60604, USA
>>> Tel +1 (312) 542-1336
>>> www.patsystems.com
>>>
>>>
>>>
>>>
>>>
>>> -------------------------------------------------------
>>> This SF.Net email is sponsored by the new InstallShield X.
>>> From Windows to Linux, servers to mobile, InstallShield X is the
>>> one installation-authoring solution that does it all. Learn more and
>>> evaluate today! http://www.installshield.com/Dev2Dev/0504
>>> _______________________________________________
>>> Quickfix-developers mailing list
>>> Qui...@li...
>>> https://lists.sourceforge.net/lists/listinfo/quickfix-developers
>>>
>>
>>
>> -------------------------------------------------------
>> This SF.Net email is sponsored by the new InstallShield X.
>>> From Windows to Linux, servers to mobile, InstallShield X is the
>> one installation-authoring solution that does it all. Learn more and
>> evaluate today! http://www.installshield.com/Dev2Dev/0504
>> _______________________________________________
>> Quickfix-developers mailing list
>> Qui...@li...
>> https://lists.sourceforge.net/lists/listinfo/quickfix-developers
>>
>
>
> -------------------------------------------------------
> SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media
> 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33
> Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift.
> http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285
> _______________________________________________
> Quickfix-developers mailing list
> Qui...@li...
> https://lists.sourceforge.net/lists/listinfo/quickfix-developers
>
>
>
> -----------------------------------------------------------------
>         Visit our Internet site at http://www.reuters.com
>
> Get closer to the financial markets with Reuters Messaging - for more
> information and to register, visit http://www.reuters.com/messaging
>
> Any views expressed in this message are those of  the  individual
> sender,  except  where  the sender specifically states them to be
> the views of Reuters Ltd.
>
>
>
> -------------------------------------------------------
> SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media
> 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33
> Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift.
> http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285
> _______________________________________________
> Quickfix-developers mailing list
> Qui...@li...
> https://lists.sourceforge.net/lists/listinfo/quickfix-developers
>
>
>