double to unsigned int / char

Question

I read here, that:

According to C99 §6.3.1.4 footnote 50:

The remaindering operation performed when a value of integer type is converted to unsigned type need not be performed when a value of real floating type is converted to unsigned type. Thus, the range of portable real floating values is (−1, Utype_MAX+1).

Now, I am interested in the subtle difference (this time for C++ 03!) between:

double d1 = 257;
double d2 = -2;

unsigned char c1 = d1; // undefined, since d1 > 256
unsigned char c2 = d2; // undefined, since d2 < -1

and

double d1 = 257;
double d2 = -2;

unsigned int i1 = d1; // defined, since d1 <= 2^32
unsigned int i2 = d2; // still undefined, right?

unsigned char c1 = i1; // defined, modulo 2^8, so c1 == 1

So the first c1 and the second c1 are not guaranteed to compare as equal, right? Is the citation from above also valid for C++03, or are there other rules?

Edit:

and for making c2 defined (for -(2^31-1) <= d2 < 0) would this be necessary to do?

double d2 = -2;
int sign = (d2<0 ? -1 : 1);

unsigned char c2 = sign * (int)abs(d2); // defined, c2 == 2^8-2 ?

Answer 1

Yes, the same rule applies for C++. (However, I am going by a 2010 draft of the C++ standard; C++ 2003 is old. Also, I am using N3092, not the official draft.) Clause 4.9, paragraph 1 says “The behavior is undefined if the truncated value cannot be represented in the destination type.”

Unsigned integer arithmetic does wrap; it is performed modulo 1 more than the maximum value of the type. However, this applies to arithmetic within the type. Conversion from floating-point to unsigned integer is not part of this.

Your code for converting d2 seems more complicated than necessary. If d2 is within the range of an int, you could simply use unsigned char c2 = (int) d2;. (Although conversion from int to unsigned int is also outside homogeneous unsigned integer arithmetic, the specification for this conversion does say it is reduced the same way unsigned integer arithmetic is.)