Depending on MATLAB version and platform and compile flags, mwSize can either be an int (a signed integer) or a size_t (an unsigned integer ... key point here). For your case you have mwSize as size_t, an unsigned integer. So t and T are unsigned. Thus the result of the test t>=0 will ALWAYS be true. Once you get past the t=0 case, the expression t-- actually does a wrap-around in modulo arithmetic to get a very large unsigned value (the max possible then it keeps going). Then you print out the results but use the wrong format code for this, %d which is for signed values. So the mexPrintf essentially interpretes this very large unsigned bit pattern (the most significant bit is set) as a signed integer and prints it out.
To demonstrate that mwSize is in fact unsigned you can try this:
if( (t = -1) > 0 ) {
mexPrintf("mwSize is unsigned\n");
}
