[cfp-interest 3898] Re: convertFrom and signaling NaNs

[Damian McGuckin]

On Thu, 16 Apr 2026, RAJAN BHAKTA wrote:

> I was giving a specific case of value, but the general case I believe holds:

> The value of a NaN (signaling or otherwise) is a NaN (not a number).

Correct. When it comes to their "value", all you can say is that the value 
of a NaN does not compare equal to itself. That is not to say that the 
bit-by-bit representation of a NaN does not compare equal to itself. And 
likewise, the payloads of an sNaN and a qNaN may be identical. And 
likewise, the 'totalOrder()' operation is defined for any two memory (or 
register) objects which are a NaN.  An there are multiple bit-wise 
representations which are a NaN which is different to say an Infinity or
a Zero of which there are only two of each and they are distinguishes from 
each other by the sign. Remember the sign of a NaN has no special meaning.

I had a look in

 	https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2026/p3938r1.html

But I do not have time to review it before your meeting next week.

One thing jumped out at me however:

 	The value of a floating-point-literal never has negative sign bit.

A value can only be positive or negative. That said, at least in C, the
negative sign in front of say

 	1234.5678

is not part of the definition of a literal so that mathematically, a 
literal in C, is always a positive number although because

 	0.0

is a literal, a literal is not always strictly positive.

I could always be wrong. It has happened way too many times in the past!

Thanks - Damian