From: W.
Sent: Thursday, August 04, 2011 5:34 PM
To: Zbynek Zahradnik
Subject: ....

Hello Zybnek,

[...] the [...] customer [...] was doing multithreading and believes the last problems were in his application as we thought.

.....
W.

Win,

No, I haven’t seen this behavior. The problem description is, however, still more user-oriented rather than developer-oriented. If the customer’s application appears “hung”, and he is developing it, he should be able to figure out where it is hung – the actual method call.

If it appears hung on creation of EasyDAClient object, then it would point towards a bug (some kind of deadlock) in the component. If it is hung in other methods calls (such as ReadItemValue), it could be waiting before a timeout occurs.

<span style="font-family: "Tahoma","sans-serif"; font-size: 10pt; mso-fareast-font-family: 'Times New Roman'">From:<span style="font-family: "Tahoma","sans-serif"; font-size: 10pt; mso-fareast-font-family: 'Times New Roman'"> W.
Sent: Friday, July 29, 2011 3:53 PM
To: Zbynek Zahradnik
Subject: RE: DataCOM and Reconnecting

Hello Zbynek,

Sorry for the poor level of detail I sent. The customer clarified it for me.

<span style="font-family: "Microsoft Sans Serif","sans-serif"; color: black; font-size: 8.5pt">When I restart the application, the screen never appears - It is almost as if it is hung up somewhere. But when I kill eopcdal.exe and then run it, the screen comes right up.

Have you seen any behavior like this in the past.

Thanks,

Hello W.,

With regard to CSingleLock, I can only tell that if the code had a synchronization bug, then of course it could have caused various crashes. I cannot tell if this was the cause or not.


Killing “eopcdal.exe” should not be necessary – if it is, something is wrong. I do not understand, however, what is meant by “I could not restart the application”. What were the symptoms?

From: W.
Sent: Thursday, July 28, 2011 4:28 PM
To: Zbynek Zahradnik
Subject: RE: DataCOM and Reconnecting

Hello Zbynek,

I just heard back from the customer and he has a question. He is asking about the eopcdal.exe. Let me know if you have any feedback on this.


Thanks for the info. I did make a change yesterday having to do with the CSingleLock stuff because I remembered in the past having an issue with the MFC lock stuff. I was using it slightly differently in the alive thread than we use it in other places, so I changed it to be exactly like the other places. I installed this change on our prototype machine and our production machine yesterday and it has not crashed since. So we will keep our fingers crossed that that took care of the problem.

One notable observation is that when it did crash, I could not restart the application without first killing the "eopcdal.exe" task. Any ideas on that?


W.

From: W.
Sent: Thursday, July 28, 2011 4:25 PM
To: Zbynek Zahradnik
Subject: RE: DataCOM and Reconnecting

Hello Zbynek,

I will pass this information to the end user and see if he can follow any of our recommendations.

Thanks,

W.,

The easiest way to troubleshoot this would be to build the application in Debug configuration, and let it run under debugger. If it crashes, collect as much information available – and specifically, where it has crashed, and the call stack of the executing thread (or other relevant threads). If the problem is access violation (such as dereferencing an invalid pointer), then also at look values of pointers used at moment of the crash.

The crash message is coming from MSVCR90.DLL, which is Microsoft Visual C++ runtime library for Visual Studio 2008. This implies that it is coming from the user’s code, because QuickOPC-COM (OPCData-COM) is built using Visual Studio 2010 since version 5.01; and of course, in addition, if the component is indeed running out-of-process, a crash of the user’s application cannot be directly caused by that other process. It likely that there is some bug or at least incomplete error checking in the user’s code. I have inspected the code visually and could not spot it, but I believe it is there.

From: W.
Sent: Wednesday, July 27, 2011 3:34 PM
To: Zbynek Zahradnik
Subject: RE: DataCOM and Reconnecting

Hello Zbynek,

The customer had another crash last night. Attached are the errors and his code. His code does the following every second.

Read ValueA, Write ValueB, Write ValueC, Sleep for 1 second

What information do you want from me to troubleshoot this crash?

Thanks,

W.
[for privacy, attachment containing the source code is not included]

W.,

An upgrade to latest build in that major version (5.02.455.1) would not harm, but I actually do not see any compelling reason to upgrade, with the current knowledge of the problem.

Regards,

From: W.
Sent: Thursday, July 14, 2011 10:47 PM
To: Zbynek Zahradnik
Subject: RE: DataCOM and Reconnecting

Hello Zbynek,

Unfortunately the customer cannot give me steps to reproduce the problem. He has tried the steps you did and everything reconnects correctly. Under certain circumstances (which we don’t know) he will get this failure. Or in the last email I sent you he actually saw a crash in a Microsoft DLL.

I have asked him to turn on the OPC Diagnostics in the TOP Server tonight. If a failure occurs it is critical we see where it is failing.

Also he is using v5.1.274. Should I have him upgrade? If so which version should he be using?

Thanks,

W.