Re: [Chicken-users] long pointer equivalent

[felix]

> "felix" <address@hidden> writes:
> > > Well, it can't be a float in that if you're doing lseeks you really
> > > want to specify which block in the disk it is you're looking for, not
> > > which general section. :)
> > 
> > I don't understand. As long as you return/pass a float with 
> > a zero fraction, everything should be ok:
> > 
> > (lseek 545485845.0 seek_cur)
> > 
> > It only has to be *representable* as a 64-bit number. And that
> > can be checked in the call.
> 
> All fractionless zero-exponent 64 bit floats can be represented as 64
> bit integers. The problem is that the reverse is not true -- all 64
> bit integers cannot be represented exactly as the the mantissa of some
> double precision float.

Of course you're right. My apologies for being so daft. ;-)

Hm, how much can then be represented in an IEEE double?
Would it be sufficient for, say, a range of 48 bit integers?


cheers,
felix