Greetings, and thanks for the feedback! Indeed, we just forward the C code result here. The '-' sign has recently appeared in newer gcc. Unlike the case with negative zero, which does denote a lisp number, +-inf/nan have no lisp counterpart, so it seems best just to follow ieee here. Compliant lisp code should not depend on the printed representation, as far as I understand. This is why we only print these when the special variable si::*print-nans* is set to non-nil.
Take care, James Cloos <[email protected]> writes: >>>>>> "RD" == Robert Dodier <[email protected]> writes: > >>> >(setq b (- a) c (- a a)) >>> #<-nan> > > RD> Minor wart here -- should be #<nan>, NaN doesn't have a sign. > > they do at the hw level (per ieee754). > > other langs support signed nans. i presume gcl must be passing thru > what the c code returns. > > -JimC -- Camm Maguire [email protected] ========================================================================== "The earth is but one country, and mankind its citizens." -- Baha'u'llah
