[cfp-interest 3950] Re: Weird Casting Error

[Joshua Cranmer]

On 6/10/2026 11:50, Vincent Lefevre wrote:
> On 2026-06-10 11:35:56 -0400, Joshua Cranmer wrote:
>> My copy of IEEE 754 has this in section 5.8:
>>
>>> Implementations shall provide conversion operations from all supported
>>> arithmetic formats to all supported
>>> signed and unsigned integer formats.
>> IEEE 754 clearly intends that conversion from floating-point formats to
>> unsigned integers is a legal (static) operation. What happens when the
>> dynamic value is out-of-range is not specified by IEEE 754 (outside of
>> invalid exception), and C opts for the UB route there.
> IEEE 754-2019 says:
>
>    When a NaN or infinite operand cannot be represented in the
>    destination format and this cannot otherwise be indicated, the
>    invalid operation exception shall be signaled. When a numeric
>    operand would convert to an integer outside the range of the
>    destination format, the invalid operation exception shall be
>    signaled if this situation cannot otherwise be indicated.
>
> So I don't think that UB is correct. However, it depends on the
> bindings (there may be a way to do the conversion that is UB and
> another way that conforms to IEEE 754).

With Annex F, the resulting value is unspecified (see F.4). Without 
Annex F, the rule is given in 6.3.14:

> When a finite value of standard floating type is converted to an 
> integer type other than bool, the
> fractional part is discarded (i.e., the value is truncated toward 
> zero). If the value of the integral part
> cannot be represented by the integer type, the behavior is undefined.