Re: [bug-gawk] Overflow to Infinity

[Andrew J. Schorr]

On Mon, Jul 16, 2018 at 10:04:46PM +0300, Eli Zaretskii wrote:
> > From: "Andrew J. Schorr" <address@hidden>
...
> > I could be mistaken, but I think that NaN was defined by IEEE 754, not C.  
> > So
> > Are we mimicking C or are we attempting to comply more closely with
> > the IEEE standard that defined NaN in the first place?
> > 
> > https://en.wikipedia.org/wiki/IEEE_754-1985
> 
> C just gives you what the FP instructions produce, and the
> instructions were made to match IEEE.
> 
> > Do other languages handle NaN differently than C, or does IEEE really
> > define this behavior?
> 
> I don't know.

As far as I know, NaN is a Frankenstein created by IEEE, and programmers expect
NaN to behave the particular way that IEEE defined it to behave.  So why is it
desirable for gawk to have NaN behavior that doesn't match the standard? Isn't
that more confusing?

Regards,
Andy