Am Di., 27. Okt. 2020 um 18:37 Uhr schrieb Wolfgang Corcoran-Mathe
<xxxxxx@sigwinch.xyz>:
>
> On 2020-10-27 07:58 +0100, Marc Nieper-Wißkirchen wrote:
> > Am Di., 27. Okt. 2020 um 02:18 Uhr schrieb Wolfgang Corcoran-Mathe
> > <xxxxxx@sigwinch.xyz>:
> > >
> > > On 2020-10-26 18:18 -0400, John Cowan wrote:
> > > > On Mon, Oct 26, 2020 at 2:07 PM Wolfgang Corcoran-Mathe <xxxxxx@sigwinch.xyz>
> > > > wrote:
> > > >
> > > > Would it be sufficient to provide C versions of
> > > > > nan-quiet?, nan-negative?, etc., operating on doubles?
> > > > >
> > > >
> > > > That's exactly what I had in mind.  Create a union type in C that can be
> > > > either a double or a uint64, put in the double, get out the uint64, and
> > > > bit-twiddle it, probably 10-15 lines of code.
> > >
> > > This would be undefined behavior in C89, but, after a lot of
> >
> > Is it undefined in C89? I think it is just implementation-defined behavior.
>
> You're right, it's implementation-defined.
>
> > > > If the member used to access the contents of a union object is not
> > > > the same as the member last used to store a value in the object, the
> > > > appropriate part of the object representation of the value is
> > > > reinterpreted as an object representation in the new type as
> > > > described in 6.2.6 (a process sometimes called "type punning").
> > > > This might be a trap representation.
> >
> > "Reinterpreted as an object ..." is again implementation-defined,
> > isn't it? Is there a way to find out how an IEEE double is mapped to
> > an array of chars?
>
> Yes, it seems so.  (With ISO C, there's always more.  I'm not really
> sure now whether their adding that footnote in the 2007 draft
> clarified anything.)

I don't think ISO C can do much about it as the binary representation
of doubles is implementation-dependent. So, there cannot be a portable
implementation in C. To demonstrate what is possible, one could take a
Scheme implementation that has a fixed set of backends (say Chez or
Loko) and develop the architecture-specific C code for each backend.
However, a Scheme that is supposed to run whenever a C compiler is
available (like Chibi) probably cannot implement SRFI 208 while
maintaining this guarantee.