[BUG] INT_MIN..=INT_MAX is broken

[GrigorenkoPV]

main: () = {
	MIN: u8 = 1;
	MAX: u8 = 255;
	for MIN..=MAX do (i) {
		std::cout << i as u32 << std::endl;
	}
}

correctly prints numbers from 1 to 255, while

main: () = {
	MIN: u8 = 0;
	MAX: u8 = 255;
	for MIN..=MAX do (i) {
		std::cout << i as u32 << std::endl;
	}
}

prints nothing.

(Not discovered by me).

If I have read the implementation right, the boundary conditions are not considered. For example, we have for 0..=UINT32_MAX. And the program will go into an infinite loop or do nothing, because for

    range(
        T const& f, 
        T const& l, 
        bool     include_last = false
    ) 
        : first{ f }
        , last{ l } 
    { 
        if (include_last) { 
            ++last; 
        }
    }

curr will never be (or will immediately be) equal to last.

[sookach]

Thank you for the bug report and especially for the detailed explanation.

[hsutter]

Yes, thanks!

Both examples now work correctly.

The fix I decided to use is that if the range type is integral, I widen the stored values so that it's always possible to represent it as a half-open range, which keeps the code simple (or at least simpler, no range/iterator is "simple" exactly 😉 ).

So now the only numeric range I don't support is one whose end value is numeric_limits< [size_t | ptrdiff_t ] >::max(). I think not supporting that case is fine... those are typically used as invalid/sentinel values, not valid program values that you'd want to include in a closed range, I think.

Again, thanks!