socket problem

I have this function where something very strange happens:

Code:

int DGRAMSocketRead_Wait ( int sock, char* iBuffer , unsigned int iToReadByteNumber,  	struct timeval * TimeOut, UINT32 * ReadByteNumber)
{
	//to be tested with signaling simulation and partial read (client sends only part of the message or too slowly)
 
#if WINDOWS || WINDOWSCE
	int errno;
#endif

 //	int 		ToReadByteNumber;
 	int			RdByteNumber;
 //	unsigned char*		Buffer;
 	
 	fd_set mono_fd_set, read_fd_set;
 	int DataReady;
 	
 	struct timeval  TempTO;
 	
 	Type_LogString LogString;
	Type_PrefixString Prefix;
	snprintf(Prefix,sizeof(Prefix),"DGRAMSocketRead_Wait::");
 	
  struct sockaddr from;
  int sizeoffrom;


 	
    FD_ZERO (&mono_fd_set);
    FD_SET (sock, &mono_fd_set);
 
 	//ToReadByteNumber = iToReadByteNumber;
 	RdByteNumber = -1;
 	//Buffer           = iBuffer;


	
    read_fd_set = mono_fd_set; 
 	TempTO=*TimeOut;
    //to be done upper limit on TimeOut in case of signal
	DataReady= select (FD_SETSIZE, &read_fd_set, NULL, NULL, &TempTO);
    while (DataReady < 0) 
    { 
		#if WINDOWS || WINDOWSCE
			errno=WSAGetLastError();
//{snprintf(LogString,sizeof(LogString),"dbg:: errno %d",errno); Log(&Lg,Prefix,LogString,2,LOG_ON_ERROR_ALSO);}
			//In the previous line : if I put snprintf out of comment it is not executed but the whole function works correctly reading the packet of the size expected!!??
			//if I put snprintf under comment strangely enough I get : Log(&Lg,Prefix,"TIME OUT",2,LOG_ON_ERROR_ALSO);return MYSOCKET_USER_DEFINED_TIMEOUT;
					#endif
#if LINUX
		if (errno!=EINTR)
#endif
			{
	     	 		Log(&Lg,Prefix,"Read Error",2,LOG_ON_ERROR_ALSO);
	           		return MYSOCKET_ERROR;
	      	}
	        read_fd_set = mono_fd_set; 
	     	TempTO=*TimeOut;//TempT0;//it is not POSIX compliant to use TempTO cause it is not decremented in all implementations
    		DataReady=select (FD_SETSIZE, &read_fd_set, NULL, NULL, &TempTO);
    }  
    
    if (!DataReady)
    {
    	Log(&Lg,Prefix,"TIME OUT",2,LOG_ON_ERROR_ALSO);
    	*TimeOut=TempTO;
  		return MYSOCKET_USER_DEFINED_TIMEOUT;
    }
    
    
     while ( RdByteNumber < 0 )
     {        
 		//in case of EINTR chars should not be lost before the first read

        
	    	RdByteNumber = recvfrom( sock , iBuffer , iToReadByteNumber, NULL, &from, &sizeoffrom  );
		#if WINDOWS || WINDOWSCE
			errno=WSAGetLastError();
		#endif
	    	if 	( RdByteNumber < 0 )
	       	{
	    		if (errno==MYEAGAIN || errno==MYEWOULDBLOCK )//some Linux do this if the data was lost after select() for some (strange) reason, like check sum
	    		{
	    			*ReadByteNumber=0;
	    			return MYSOCKET_DATA_FALLEN;
	    		}
	    		else
#if LINUX
					if	( errno != EINTR )//also SERVER_NON_CATASTROFIC_ERRNO must cause an error return code here
#endif
	          	{
	       			Log(&Lg,Prefix,"Read Error",2,LOG_ON_ERROR_ALSO);
	          		return MYSOCKET_ERROR;
	          	}
	       	}
	    	else if ( ReadByteNumber == 0 )
	       	{
	    		Log(&Lg,Prefix,"Connection closed",2,LOG_ON_ERROR_ALSO);
	       		return MYSOCKET_CONNECTION_CLOSED;
	       	}
	    	else ;
	    	//{
		    //	ToReadByteNumber = ToReadByteNumber - ReadByteNumber;
		    //	Buffer = Buffer + ReadByteNumber;
	    	//}

        
    }
    *ReadByteNumber=RdByteNumber;
 	return MYSOCKET_OK;
 
}

where the strange behaviour is this in short:

Code:

//{snprintf(LogString,sizeof(LogString),"dbg:: errno %d",errno); Log(&Lg,Prefix,LogString,2,LOG_ON_ERROR_ALSO);}
			//In the previous line : if I put snprintf out of comment it is not executed but the whole function works correctly reading the packet of the size expected!!??
			//if I put snprintf under comment strangely enough I get : Log(&Lg,Prefix,"TIME OUT",2,LOG_ON_ERROR_ALSO);return MYSOCKET_USER_DEFINED_TIMEOUT;

it is very likely - you have some buffer overrun that corrupts your stack... when sprintf is uncommented - the corruption does not prevent function from executing what it should, when you uncomment that line - something essential is overwritten.

I would started with checking that all your automatic vars have enough space for storing what you pass into them...

Originally Posted by vart

it is very likely - you have some buffer overrun that corrupts your stack... when sprintf is uncommented - the corruption does not prevent function from executing what it should, when you uncomment that line - something essential is overwritten.

I would started with checking that all your automatic vars have enough space for storing what you pass into them...

i agree

Originally Posted by vart

it is very likely - you have some buffer overrun that corrupts your stack... when sprintf is uncommented - the corruption does not prevent function from executing what it should, when you uncomment that line - something essential is overwritten.

I would started with checking that all your automatic vars have enough space for storing what you pass into them...

i have disabled the directive /GL (link time code generation) (I am using MS Visual 2008 Express Ed. with Berkley sockets) and the problem seem to have vanished, but perhaps just in a way similar to excluding the comment of the snprintf (perhaps it didn't really solved)..

i didn't find any memory leak or problem for the moment..