openbsd-ports/math/py-numpy/Makefile

# $OpenBSD: Makefile,v 1.43 2015/01/15 02:55:22 daniel Exp $

BROKEN-alpha =	numpy/linalg/umath_linalg.c.src:3231: internal compiler error: in extract_insn, at recog.c:2077

COMMENT=		fast array and numeric programming library for Python

MODPY_EGG_VERSION=	1.8.2
REVISION=		1
DISTNAME=		numpy-${MODPY_EGG_VERSION}
PKGNAME=		py-${DISTNAME}

CATEGORIES=		math devel

HOMEPAGE=		http://numpy.scipy.org/

MAINTAINER=		Daniel Dickman <daniel@openbsd.org>

# BSD
PERMIT_PACKAGE_CDROM=	Yes

WANTLIB=		blas lapack m pthread ${MODFORTRAN_WANTLIB} ${MODPY_WANTLIB}

MASTER_SITES=		${MASTER_SITE_SOURCEFORGE:=numpy/}

MODULES=		lang/python \
			fortran

FLAVORS=		python3
FLAVOR?=

MODPY_SETUPTOOLS=	Yes
MODPY_SETUP=		setupegg.py

TEST_DEPENDS=	devel/py-nose${MODPY_FLAVOR}

MODFORTRAN_COMPILER =	g77

BUILD_DEPENDS = ${MODFORTRAN_BUILD_DEPENDS}
LIB_DEPENDS=            math/lapack \
			${MODFORTRAN_LIB_DEPENDS}

pre-build:
	cp -f ${WRKSRC}/numpy/distutils/site.cfg ${WRKSRC}/site.cfg

do-test: fake
	${MAKE_ENV} PYTHONPATH=${WRKINST}${MODPY_SITEPKG} ${MODPY_BIN} -c \
	    'import numpy ; numpy.test()'

.include <bsd.port.mk>
Some minor numpy fixes: - add pthread to WANTLIB - backport fix for datetime64 on macppc allowing all tests to pass there - remove use of -funroll-loops with fortran (it causes scipy to segfault on macppc and it can make code bigger and slower in some cases) - "optimize function attributes" were added in gcc 4.4. silence those build-time warnings when building with earlier gcc versions. 2015-01-14 21:55:22 -05:00			`# $OpenBSD: Makefile,v 1.43 2015/01/15 02:55:22 daniel Exp $`

			`BROKEN-alpha = numpy/linalg/umath_linalg.c.src:3231: internal compiler error: in extract_insn, at recog.c:2077`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00			`COMMENT= fast array and numeric programming library for Python`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
- update numpy to 1.8.2 and take maintainer - python3 flavor contributed by Dawe with assistance from rpointel@ - tested in a bulk by landry@ - bcallah@ needs this update to update math/veusz ok landry@ 2014-09-16 20:46:13 -04:00			`MODPY_EGG_VERSION= 1.8.2`
Some minor numpy fixes: - add pthread to WANTLIB - backport fix for datetime64 on macppc allowing all tests to pass there - remove use of -funroll-loops with fortran (it causes scipy to segfault on macppc and it can make code bigger and slower in some cases) - "optimize function attributes" were added in gcc 4.4. silence those build-time warnings when building with earlier gcc versions. 2015-01-14 21:55:22 -05:00			`REVISION= 1`
update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00			`DISTNAME= numpy-${MODPY_EGG_VERSION}`
REVISION 2010-07-18 06:50:11 -04:00			`PKGNAME= py-${DISTNAME}`
- update numpy to 1.8.2 and take maintainer - python3 flavor contributed by Dawe with assistance from rpointel@ - tested in a bulk by landry@ - bcallah@ needs this update to update math/veusz ok landry@ 2014-09-16 20:46:13 -04:00
update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00			`CATEGORIES= math devel`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
- update py-numpy to 1.3.0 - update homepage "looks good" djm@ (MAINTAINER) 2009-09-28 11:15:37 -04:00			`HOMEPAGE= http://numpy.scipy.org/`
update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00
- update numpy to 1.8.2 and take maintainer - python3 flavor contributed by Dawe with assistance from rpointel@ - tested in a bulk by landry@ - bcallah@ needs this update to update math/veusz ok landry@ 2014-09-16 20:46:13 -04:00			`MAINTAINER= Daniel Dickman <daniel@openbsd.org>`

import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00			`# BSD`
			`PERMIT_PACKAGE_CDROM= Yes`
Remove '-funroll-loops' to workaround a segmentation fault on powerpc when building py-scipy. maintainer timeout discussed with and ok eric@, ok jasper@ 2010-01-17 14:45:47 -05:00
Some minor numpy fixes: - add pthread to WANTLIB - backport fix for datetime64 on macppc allowing all tests to pass there - remove use of -funroll-loops with fortran (it causes scipy to segfault on macppc and it can make code bigger and slower in some cases) - "optimize function attributes" were added in gcc 4.4. silence those build-time warnings when building with earlier gcc versions. 2015-01-14 21:55:22 -05:00			`WANTLIB= blas lapack m pthread ${MODFORTRAN_WANTLIB} ${MODPY_WANTLIB}`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00			`MASTER_SITES= ${MASTER_SITE_SOURCEFORGE:=numpy/}`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
module fortran 2010-05-23 06:25:55 -04:00			`MODULES= lang/python \`
			`fortran`

- update numpy to 1.8.2 and take maintainer - python3 flavor contributed by Dawe with assistance from rpointel@ - tested in a bulk by landry@ - bcallah@ needs this update to update math/veusz ok landry@ 2014-09-16 20:46:13 -04:00			`FLAVORS= python3`
			`FLAVOR?=`

update to numpy-1.2.1, set MODPY_EGG_VERSION + MODPY_SETUPTOOLS reorder to Makefile.template 2009-01-19 02:11:42 -05:00			`MODPY_SETUPTOOLS= Yes`
Use numpy's special setuptools file when building (and regen PLIST). Build breakage pointed out by jasper@; fix by me. ok jasper@, ajacoutot@ 2009-01-22 14:23:10 -05:00			`MODPY_SETUP= setupegg.py`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
- update numpy to 1.8.2 and take maintainer - python3 flavor contributed by Dawe with assistance from rpointel@ - tested in a bulk by landry@ - bcallah@ needs this update to update math/veusz ok landry@ 2014-09-16 20:46:13 -04:00			`TEST_DEPENDS= devel/py-nose${MODPY_FLAVOR}`
Add py-nose to REGRESS_DEPENDS. Before: ImportError: Need nose >= 0.10.0 for tests - see http://somethingaboutorange.com /mrl/projects/nose After: Ran 1128 tests in 5.271s OK 2009-01-27 16:12:02 -05:00
replace MODFORTRAN_WANTG77 with MODFORTRAN_COMPILER 2010-10-23 11:09:57 -04:00			`MODFORTRAN_COMPILER = g77`
module fortran 2010-05-23 06:25:55 -04:00
			`BUILD_DEPENDS = ${MODFORTRAN_BUILD_DEPENDS}`
missed a few depends 2010-11-22 04:30:38 -05:00			`LIB_DEPENDS= math/lapack \`
module fortran 2010-05-23 06:25:55 -04:00			`${MODFORTRAN_LIB_DEPENDS}`
make use of lapack/blas ok djm@ 2009-02-16 05:10:09 -05:00
don't link against blas/lapack for now. ok espie@ djm@ 2007-08-07 01:25:58 -04:00			`pre-build:`
			`cp -f ${WRKSRC}/numpy/distutils/site.cfg ${WRKSRC}/site.cfg`

PERMIT_* / REGRESS->TEST sweep 2013-03-11 07:20:26 -04:00			`do-test: fake`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00			`${MAKE_ENV} PYTHONPATH=${WRKINST}${MODPY_SITEPKG} ${MODPY_BIN} -c \`
Adapt some changes from the upstream repository for cpu detection. I have only waited a short while for djm (maintainer)'s ok, commiting anyway as it takes out a lot of the tree on !{i386,amd64}. Fix the regression tests while there. ok jasper@. 2009-10-14 14:44:04 -04:00			`'import numpy ; numpy.test()'`
import of numpy-0.9.8 with some fixes from numpy SVN -current: > NumPy is a Python library for numeric and scientific computing, it > contains a powerful N-dimensional array object, basic linear algebra > functions, basic Fourier transforms, sophisticated random number > capabilities and tools for integrating Fortran code. feedback and ok alek@ 2006-06-25 20:45:37 -04:00
			`.include <bsd.port.mk>`