[Cfp-interest] printf, NaN, infinity

[Mike Cowlishaw]

For what it's worth, decNumber uses a quite simple NAN -> chars conversion
that is bounded; being essentially

  [signbit] ['s'] NaN [payload]

where signbit is '-' if the sign bit is 1, 's' is added for signalling NaN,
and payload is the payload as an integer (omitted if the payload is 0).

e.g.:

   NaN 
   NaN101
  -Nan12345
   sNaN
   sNaN12

All simply parsed, etc -- many examples in my testcases.  Perhaps something
similar would work for binary NaNs?

Mike

 

> -----Original Message-----
> From: cfp-interest-bounces at oakapple.net 
> [mailto:cfp-interest-bounces at oakapple.net] On Behalf Of Fred 
> J. Tydeman
> Sent: 26 November 2018 16:39
> To: Jim Thomas
> Cc: CFP
> Subject: Re: [Cfp-interest] printf, NaN, infinity
> 
> On Mon, 26 Nov 2018 08:27:15 -0800 Jim Thomas wrote:
> >
> >Is this proposal at the request of WG14? If so, what exactly 
> was requested?
> 
> Not exactly.  Martin Sebor presented a paper showing that the 
> output of the form NaN(chars) is unbounded, so is a security 
> problem.  He presented an idea on how to limit that output.  
> The committee did not like his idea.  So, I decided to come 
> up with my own solution.
> 
> 
> 
> ---
> Fred J. Tydeman        Tydeman Consulting
> tydeman at tybor.com      Testing, numerics, programming
> +1 (702) 608-6093      Vice-chair of PL22.11 (ANSI "C")
> Sample C99+FPCE tests: http://www.tybor.com Savers sleep 
> well, investors eat well, spenders work forever.
> 
> _______________________________________________
> Cfp-interest mailing list
> Cfp-interest at oakapple.net
> http://mailman.oakapple.net/mailman/listinfo/cfp-interest
>