Server throws AccessViolation Error when started up

[Sendil kumar]

I'm a software tester. My recent task is to test the exposed methods of our C# server which is using WCF as communication layer.

I wrote a piece of C# code to tets our server. for one of the test, I had to test whether the server is throwing up AddressAlreadyUsed exception when multiple instance of the server is started up.

Ex, my pseudocode
while(till count reaches a limit)
{
Process.Start("server info);
}

During the execution, many instance of the server was started up and it throwed AddressAlreadyUsed exception as expected. But, for some server instance it throwed up an AccessViolation exception.

ex code
21-Jul-2009 15:26:32.71 Error: Unhandled Exception: System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
at sFabServer.returnEntry(sFabServer* , sTString<char>* )
at SSM.BCT.Legacy.FabServerConfig.ReturnEntry(String name)
at SSM.BCT.ProcessServer.Server.ServerContext..ctor(P rocessServerConfig config,

Can any one let me know why this was thrown up and is there any un-safe code/thread in the server that spaned up but not killed?

[Plater]

I would guess it has to do with one of the Servers being at a certain state when the next one called.
Try doing a thread sleep in your loop and see if that lowers the number that produce AccessViolations?

[Sendil kumar]

Hi Platter,

Thanks for your reply. I will try your suggestion.

I have a question.

I just wondered about what this "System.Runtime.InteropServices.SEHException: External component has thrown an exception." could be, which was thrown for this scenario). Going through the internet related to this error, I came across some people complaining about similar problem and found out that the issue was because of the CLR of the dotnet.

link:
http://objectmix.com/dotnet/97977-i-...pservices.html

I'm not sure about this because those people are using com components which is not in our case, but we are making a call to unmanaged C++ from our C#.
So do you think that this could be a reason for this problem?

We are using CLR version 3.5, but the problem was reported against CLR 2.0.
Is the CLR 3.5 also contains same problem?

[Plater]

unmanaged code can be considered a COM object for these purposes I think

[Sendil kumar]

Hi Plater,

I tried your suggestion, introduced a sleep between the subsequent calls of the server, surprisingly the problem didn't occur with this scenario.

So, What could be the problem? Is it my code?

Thanks,
Sendil

[Plater]

What object you are starting, is in the middle of "starting up" when the next one tries to start. Since the first one hasn't finished "starting up", it hasn't finished releasing its startup data (which is also probably used in the next object's startup). Depending on how far the first one has gotten at starting up could change the exception you are receiving. (AccessViolation vs AddressInUse)
You could maybe try to put a mutex around the data that is used in startup, but I am rather unfamiliar with how all that works