[ruby-dev:12390] Re: NaN, Infinity (Bug?)

むらけんです.  みなさん, 初めまして!!

At Wed, 7 Mar 2001 15:29:18 +0900,
gotoken wrote:
> >newlib/libm/math/w_log.c:
> >>RETURNS
> >>Normally, returns the calculated value.  When <[x]> is zero, the
> >>returned value is <<-HUGE_VAL>> and <<errno>> is set to <<ERANGE>>.
> >>When <[x]> is negative, the returned value is <<-HUGE_VAL>> and
> >><<errno>> is set to <<EDOM>>.  You can control the error behavior via
> >><<matherr>>.
> 
> なんでわざわざこんな変なことするんだろう…

僕も不思議だったので cygwin-ml で質問してみたところ, しがない
英語の投稿に対して親切に解答してくださったので引用します.

At Wed, 7 Mar 2001 04:59:46 -0800,
Tim Prince wrote:
> The math library adopted by cygwin is part of newlib, an entirely
> separate project not even oriented to the same type of usage as cygwin
> or even the glibc.  I imagine it may be intended not to depend on
> running on targets which implement IEEE standard.  If you're using gcc
> or g++, it's easy to supply your own behavior by linking your own
> version first, or by supplying an in-line definition.  IIRC, all the
> math functions test out fairly well on elefunt, except for asinf/acosf.
> I don't know why you would write an app to depend on the NaN behavior;
> you would at least have to forgo -ffast-math or supply and use your own
> isnan() and take other special precautions to make that work.

-- 
Kenta MURATA <[e-mail]> "http://village.infoweb.ne.jp/~muraken/"