On Sun 11/10/2020 12:51, Daniel Dickman wrote:
> While re-testing scipy to avoid breaking Antoine's builds post-commit, it
> looks like scipy depends on a new port for pybind11 (which I just sent to
> ports).
>
> Updated scipy diff below. Also updated the minimum numpy version that's
> needed for scipy.
>
> ok on this updated scipy diff after pybind11 goes in?
One comment, see below.
> Index: math/py-scipy/Makefile
> ===================================================================
> RCS file: /cvs/ports/math/py-scipy/Makefile,v
> retrieving revision 1.46
> diff -u -p -u -r1.46 Makefile
> --- math/py-scipy/Makefile 3 Jul 2020 21:12:57 -0000 1.46
> +++ math/py-scipy/Makefile 11 Oct 2020 16:43:58 -0000
> @@ -2,10 +2,9 @@
>
> COMMENT= maths, science and engineering modules for Python
>
> -MODPY_EGG_VERSION= 1.2.3
> +MODPY_EGG_VERSION= 1.5.2
> DISTNAME= scipy-${MODPY_EGG_VERSION}
> PKGNAME= py-${DISTNAME}
> -REVISION= 2
>
> # XXX gcc 8 issue:
> # ld: error: dfft_subr_6.f:(.debug_loc+0x2A7D): has non-ABS relocation
> R_386_GOTOFF against symbol 'ntryh.3712'
> @@ -29,7 +28,7 @@ HOMEPAGE= https://scipy.org/
> MODULES= lang/python fortran
>
> FLAVORS= python3
> -FLAVOR?=
> +FLAVOR = python3
>
> MODFORTRAN_COMPILER = gfortran
>
> @@ -38,11 +37,12 @@ MODPY_SETUPTOOLS= Yes
> MODPY_DISTUTILS_BUILDARGS = --fcompiler=gnu95
>
> BUILD_DEPENDS= ${RUN_DEPENDS} \
> - ${MODFORTRAN_BUILD_DEPENDS}
> + ${MODFORTRAN_BUILD_DEPENDS} \
> + devel/py-pybind11
^ devel/py-pybind11${MODPY_FLAVOR}
Build and (lightly) run tested on amd64.
OK bket@
