I encountered the following line of (C++) code today, along with a few more like it:

      totalcost = rnd((totalcost += cost), 2);

 

Now, as far as I can see, there's a few things wrong:

1. The += is pointless - it should just be a +
2. rnd() does not return a random number like in every other program/language I've seen. It rounds the 1st argument to the number of decimal places specified by the 2nd.
3. totalcost and cost are double-precision floating point numbers. Ever rounded a floating point number and got the expected result? rnd() actually converts the double to a decimal, rounds it, then converts it back to a double.

mxsscott:

The += is pointless - it should just be a +

No, it's definitely very pointful. It tries to invoke undefined behaviour but fails because the function call creates a sequence point.

Ah indeed. I forgot to mention my quick research into sequence points after thinking undefined thoughts!

Vempele:

mxsscott:

The += is pointless - it should just be a +

No, it's definitely very pointful. It tries to invoke undefined behaviour but fails because the function call creates a sequence point.

 

It would suck, though, if rnd() were redefined as a macro...

alegr:

Vempele:

mxsscott:

The += is pointless - it should just be a +

No, it's definitely very pointful. It tries to invoke undefined behaviour but fails because the function call creates a sequence point.

 

It would suck, though, if rnd() were redefined as a macro...

 

(Un)fortunately rnd() is defined as a function in a file that gets passed through an Informix ESQL/C precompiler because it calls deccvdbl(), decround() then dectodbl() in the Informix client library. A macro sounds like too easy a way to break the calculation.