4 replies to this topic

[PixelOp]

Hey guys, Whenever I use this piece of code:

srand(time(0));

I always get this error: : warning C4244: 'argument' : conversion from 'time_t' to 'unsigned int', possible loss of data Could anyone explain to me exactly what it means? Thanks!

[superpig]

I would guess that if you track down the definition of the type time_t, it's defined as a long (or unsigned long). On some systems, longs are larger types than ints, so when you try moving a value from one type to the other, part of it has to be thrown away.

In this particular case, it doesn't matter - all you want is some unpredicatable value to start the random number generator with. If you want to silence the warning:

srand((unsigned int)time(0));

[doynax]

It simply means that the type time_t (returned by time) isn't guaranteed to be equivalent to a normal int variable.
In this case a loss of precision wouldn't cause any harm but in others it might, especially when writing portable code. The "solution" is to cast the return time into a basic integer.

srand((unsigned) time(0));

The time_t type could correspond to any arithmetic type. And sending, for example, a floating point value to a function receiving an integer without explicitly casting it down is seen by the compiler as a potential source of bugs.

[Telastyn]

time(0) returns a time_t.
srand() takes an unsigned int.

The compiler is nice and auto-converts from time_t to unsigned int. The warning comes that a time_t isn't an unsigned int and has different maximums/minimums. So what happens if the time_t is greater than the unsigned int's maximum?

[PixelOp]

Ahh, I was trying to mess around with the unsigned a bit but couldn't quite put it in the right place.

Thanks a lot guys!