Re: math/blas not compiling due to gcc7 error

[Chavdar Ivanov <ci4ic4%gmail.com@localhost>]

On Mon, 24 Aug 2020 at 06:44, Martin Husemann <martin%duskware.de@localhost> wrote:
>
> On Sun, Aug 23, 2020 at 09:31:08PM +0200, Connor McLaughlan wrote:
> > bash-5.0# nm /lib/libm.so |grep cab
> > 000000000001f300 T __c99_cabs
> > 000000000001f2e0 T __c99_cabsf
> > 000000000001f2c0 T __c99_cabsl
> > 000000000001f340 T cabs
> > 000000000001f320 T cabsf
> >
> > So it seems the problem is cabsl missing in /lib/lim.so which then is
> > a problem of the base system?
>
> The same is true for both amd64 and sparc64 in -current, the symbol used
> in libm is __c99_cabsl, our <complex.h> has:
>
>         long double cabsl(long double complex) __RENAME(__c99_cabsl);
>
> but apparently gfortran does not know about that. How does it work on
> NetBSD/amd64? Or does it fail there just the same?

FWIW I have

PKGSRC_FORTRAN=gfortran
GFORTRAN_VERSION=8

in /etc/mk.conf and math/blas builds just fine under -current amd64.

>
> Martin

Chavdar




-- 
----