[cfp-interest 3854] Re: Annex F - Special Cases & Introduction - Updates against latest C2Y

[Damian McGuckin]

On Fri, 3 Apr 2026, Damian McGuckin wrote:

> On Fri, 3 Apr 2026, Damian McGuckin wrote:
>
>>  On Fri, 3 Apr 2026, Damian McGuckin wrote:
>>
>>>    [if] x is not a signaling NaN
>>
>>  This has no parantheses which is a bonus in my opinion.
>
> The following explanatioons are now in F.2.2#3
>
> While this annex does not require the full support for signaling NaNs 
> specified in ISO/IEC 60559, it does use the phrase "x is not a signaling NaN" 
> to imply that either x is a quiet NaN or -INF <= x <= +INF. For brevity, this 
> annex also uses the term NaN, unless explicitly qualified, to denote a quiet 
> NaN, and consequently uses the phrase "x is a NaN" to imply that x belongs to 
> the set of all quiet NaNs.

Currently, C2Y says what is NOT. This is not always helpful. In an attempt 
to minimize mention of a signaling NaN which may not even be supported, I 
have simply put

 	-INF <= x <= +INF or x is a quiet NaN

directly in the domain. Much clearer. No need for explanation in F.2.2. I 
use the symbol for infinity in the text.

I have also eliminated mentions of signaling NaNs from all Annex F domains 
except for

         fabs(x)
         copysign(x, y)
         canonicalize(cxp, xp)
         totalorder/totalordermag(x, y)

All of these must mention the abnormal case where an exception is not 
raised for signaling NaN input.

Any mentions of a NaN within a domain which would imply a quiet NaN (as 
per the statement of F.2.2) have been changed to explicitly say a quiet 
NaN. There were not so many and it avoids any confusion.

Note that the inference of a unqualified NaN being a quiet NaN is still
used for arguments/parameters and results. Too many and if destroys the
readability.

Uploaded the updates:

  https://cfp-wiki.esi.com.au/pub/CFP/WebHome/C2Y-F-Introduction-20260401.pdf

  https://cfp-wiki.esi.com.au/pub/CFP/WebHome/C2Y-F-SpecialCases-20260401.pdf

Thanks - Damian