Dear Sir/Madam,

I’m developing a source code in C++ for RTU (Remote Terminal Unit). It will gather data of IO modules of RTU and then transmit them to SCADA center with standard communication protocol (IEC-870-5-101). In first step, I developed it in windows in visual C++ environment. For testing its functionality, I arranged test vectors instead of IO cards of RTU in windows. I trigger test vector via timers in windows. Actually, it is running well without exception. I used try catch in any thread in my program. For example I arranged it in DI thread as follows:

// Digital Input process thread
while(Loop){
try{
GetMsgQue()->PopWait(Msg);
Msg.OpenMsg();
bFunctionCode = (DITHREAD_EVENTS_ENUM_) Msg.GetFunCode();
switch(bFunctionCode){
case DITH_EVENT_SHOUTDOWN_: // ShutDown message to Kill thread
Loop = FALSE_S;
DestroyThread(m_pthread_t);

break;
case DITH_EVENT_HWMCHKMSG_: // Hardware Manager Health Check message to DI Thread
...
break;
case DITH_EVENT_PCIFREEZE_: // PCI Freeze message
...

break;
default :
m_pApp->GetGate()->ErrorMsg("DITH, Unknown FunCode received to DI thread!");
}
}
catch(CScadaException* e){
m_pApp->GetGate()->ErrorMsg(e->GetErrorMessage()+"!(DI_ProcessLoop)");
delete e;
}
catch(...){
m_pApp->GetGate()->ErrorMsg("Unknown exception occurred in DI_ProcessLoop!");
}
}

Now, I’m porting it to Linux to run in embedded system (RTU). It is going well so far but I am hitting a segmentation-fault and each time error occurs, my application will shut off and Aborted (Segmentation Fault) is appear on the terminal. It seems, in Windows catch(...) is a fail safe that catches everything, but in Linux catch(...) doesn't seem to catch anything.


My question is how to control this thing? Let’s say after error occurs, just prompt the error message and never do that action so that the app will be ok and still running.

Best Regards
Ali

A seg fault is typically caused by accessing an area of memory that has not been assigned to the program. So it is not a case of avoiding the problem but locating it and modifying the code.

To help you solve teh problem look at something like gdb a debugger that will be able to tell you where the error is occuring.

Hi, akzare -

I understand your frustration. The problem is that Microsoft has an OS-level feature called "Structured Exception Handling" (SEH) that their VC++ compiler (or, more precisely, their MSVC runtime) uses to implement C++ try/catch blocks.

The result is non-portable code. ANSI C++ try/catch blocks are NOT intended to catch OS-level events like segmentation faults or floating point exceptions. But, let's face it - Microsoft's Visual Studio, SEH-based implementation is really convenient (to say nothing of a lot more efficient)!

Probably the best approach in your case is to use "signal()" to trap OS-level events (SIGSEGV, SIGFPE, etc) that might occur. Perhaps you want to "#ifdef" these calls, so they're only compiled in the non-Windows versions of the code.

Although I wasn't able to find any good links that directly talk about "pure ANSI C++ vs. Microsoft SEH", these should give you the flavor of the problem:

http://www.dbforums.com/t502031.html
http://rds.yahoo.com/_ylt=A0geurvc7u...onhandling.htm

http://msdn.microsoft.com/chats/tran...io_022703.aspx

'Hope that helps .. PSM