MB>
  >    #include 
  >    #include      // for UINT_MAX
  >    #include 
  >
  >    srand((unsigned int)(time(NULL) % UINT_MAX));
MB>

  I didn't want to spoil the inpact of my previous message by complicating
  the issue too far.

    srand(time(0))

  is correct and well-formed for all OS/2 C and C++ compilers.

  However, I just need to refute your claim that the above is portable. (-:

  It isn't portable.  ISO Standard C says that time_t is an arithmetic
  type, you see.  On a (hypothetical) platform where time_t was a
  non-integral arithmetic type, such as float or double, the code violates
  a diagnosable semantic restriction, and possibly invokes undefined
  behaviour.

  In ISO Standard C++ the operands of the `%' operator *must* be of
  integral type.  Applying the `%' operator to the return value of
  time_t() on such a platform results in a violation of a semantic
  restriction, and a diagnostic message.

  In ISO Standard C++ converting an floating point value to an integral
  value results in undefined behaviour if the floating point value happens
  to be too large.

  To summarise :

      srand(time(0)) works on all C and C++ compilers for OS/2, where
      time_t is an integral type.  Your code is superfluous.

      Where time_t isn't an integral type, your code is wrong anyway.

  Ah pedantry.  (-:

  > JdeBP <
___
 X MegaMail 2.10 #0:Dr Who is back on U.K. television!
--- Maximus/2 3.01