Re: [gentoo-embedded] Re: Howto guide or something?

[Ned Ludd]

On Thu, 2004-11-11 at 09:23, Peter S. Mazinger wrote:
> On Thu, 11 Nov 2004, Ned Ludd wrote:
>
> > > X (tested older XFree86 and xorg-x11-6.8.1) can't be built against uclibc
> > > for now (missing math functions)
> > >
> > > solar,vapier:
> > > I could add to the masked uclibc
> > > use X && (apply X needed math funcs)
> > > and/or
> > > use java && (apply sun-jdk math funcs)
> > > the java condition is not quite true, they are needed to build sun-j2sdk
> >
> > Well here is what I'm thinking in this respect. mjn3 is working on a new
> > libm but it's going to be a while before he is done. Reading the
> > uClibc/TODO states that libm is targeted for 1.0 which we know is going
> > to be a ways out. So if we have replacement functions then lets post
> > them here for review and consider patching out own -lm till such time as
> > mjn3 has it ready and it works for us.
> >
> > http://uclibc.org/cgi-bin/cvsweb/uClibc/TODO?&view=auto
> > TODO list for the uClibc 1.0.0 release:
> > *) Implement the float and long double versions of math funcs
> > 	using wrappers on top of the double versions (size / precision
> > 	trade off where size clearly wins).
>
> look in the current uclibc patches tarball subdir math, those are that I
> needed until now, they are edited versions of net/openbsd (as almost the
> whole original libm stuff in uclibc)
> sinf/cosf/tanf/atan2f/sqrtf/powf/fabsf are implemented


> I'll add IUSE="X" and use X && epatch patch/math to the new uclibc ebuild

USE=X is not logical and should not be the controlling factor which
decide if libm is built.

Please consider.

if use build || use bootstrap; then
 # It's ok to skip it here.
else
 # It really should be enabled here.
fi


-- 
Ned Ludd <solar@g.o>
Gentoo (hardened,security,infrastructure,embedded) Developer