Re: missing math functions on powerpc (macppc)?

Greg Troxel <[email protected]> Thu, 17 Dec 2015 11:28:57 -0500
Newsgroups gmane.os.netbsd.ports.powerpc,gmane.os.netbsd.devel.pkgsrc.user
Message-ID <[email protected]>
"John D. Baker" <[email protected]> writes:

> While the individual object components of libvpx are linked with libm,
> the final link step creating the shared library (and the static library)
> do not include libm in their link command:
>
> libtool: link: rm -fr  .libs/libvpx.a .libs/libvpx.la .libs/libvpx.lai .libs/libvpx.lai.blsav .libs/libvpx.so .libs/libvpx.so.1 .libs/libvpx.so.1.0.5
> libtool: link: gcc -shared -fPIC -DPIC vpx/src/.libs/vpx_decoder.c.o
> [snip]
> -L/d0/build/pkgsrc/multimedia/libvpx/work/.buildlink/lib
> -Wl,-R/usr/pkg/lib -Wl,-Bsymbolic -pthread -pthread -Wl,-soname
> -Wl,libvpx.so.1 -o .libs/libvpx.so.1.0.5
> libtool: link: (cd ".libs" && rm -f "libvpx.so.1" && ln -s "libvpx.so.1.0.5" "libvpx.so.1")
> libtool: link: (cd ".libs" && rm -f "libvpx.so" && ln -s "libvpx.so.1.0.5" "libvpx.so")

That looks wrong (to omit -lm).  I have no idea why with the same
toolchain version but a different cpu type it would do that.  There may
be someting in the configure process of libvpx that keys off cputype and
makes os assumptions, or something else odd.

Next is to look at the configure log and the libvpx build system to
figure out why it's leaving it out.
signature.asc (application/pgp-signature, 180 B)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iEYEARECAAYFAlZy4skACgkQ+vesoDJhHiV3nQCaAySzLzaDwXvozD5otD1QXCoX
8c0An0YFVNeGS/QA1Pg1H9bCNJOZdAyf
=LwAl
-----END PGP SIGNATURE-----