Date: Fri, 8 Dec 2017 10:21:38 +1100 (EST) From: Bruce Evans <brde@optusnet.com.au> To: Dimitry Andric <dim@freebsd.org> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r326669 - head/lib/msun/man Message-ID: <20171208095030.X1048@besplex.bde.org> In-Reply-To: <201712072038.vB7KccuT009226@repo.freebsd.org> References: <201712072038.vB7KccuT009226@repo.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 7 Dec 2017, Dimitry Andric wrote: > Log: > Remove the sentence in math(3) about some long double math functions not > being available. > > Submitted by: Steve Kargl > MFC after: 3 days > > Modified: > head/lib/msun/man/math.3 > > Modified: head/lib/msun/man/math.3 > ============================================================================== > --- head/lib/msun/man/math.3 Thu Dec 7 20:09:17 2017 (r326668) > +++ head/lib/msun/man/math.3 Thu Dec 7 20:38:37 2017 (r326669) > @@ -28,7 +28,7 @@ > .\" from: @(#)math.3 6.10 (Berkeley) 5/6/91 > .\" $FreeBSD$ > .\" > -.Dd December 5, 2010 > +.Dd December 7, 2017 > .Dt MATH 3 > .Os > .Sh NAME > @@ -235,12 +235,6 @@ and > .Vt long double > values, were written for or imported into subsequent versions of FreeBSD. > .Sh BUGS > -Some of the > -.Vt "long double" > -math functions in > -.St -isoC-99 > -are not available. > -.Pp > Many of the routines to compute transcendental functions produce > inaccurate results in other than the default rounding mode. > .Pp The removal is technically correct, since math(3) doesn't cover complex functions and the "imprecise" functions powl() and tgammal() are available. The unusuability of the imprecise functions is just not documented anywhere (their man page gives a generic description of error bounds that is off by a factor of 2**60 for imprecise functons on ld128). This man page gives stale documentation about the bad implementation of trig functions on i386. I fixed that a year ago in r305384, but didn' notice that it is documented here. The documentation only claims inaccurate argument reduction for huge arguments. Actually, the reduction was inaccurate for all args (about N bits wrong for an error of 2**N ulps for args near 2**N times Pi/2). Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20171208095030.X1048>