Date: Tue, 14 Nov 2006 17:54:03 -0500 From: Diane Bruce <db@db.net> To: python@freebsd.org Subject: [rafan@FreeBSD.org: Re: sci-py updated] Message-ID: <20061114225403.GA72433@night.db.net>
next in thread | raw e-mail | index | archive | help
----- Forwarded message from Rong-En Fan <rafan@FreeBSD.org> -----
Date: Mon, 13 Nov 2006 21:45:03 +0800
From: Rong-En Fan <rafan@FreeBSD.org>
To: Diane Bruce <db@db.net>
Cc: Rong-En Fan <rafan@FreeBSD.org>,
"Li-Lun Wang (Leland Wang)" <llwang@infor.org>,
tonymaher@optusnet.com.au
Subject: Re: sci-py updated
In-Reply-To: <20061113133311.GA18170@night.db.net>
User-Agent: Mutt/1.5.13 (2006-08-11)
X-UIDL: TH##!TIU!!Z5"#!2K'"!
X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.0.3
On Mon, Nov 13, 2006 at 08:33:11AM -0500, Diane Bruce wrote:
> On Mon, Nov 13, 2006 at 06:37:37PM +0800, Rong-En Fan wrote:
> > On Sun, Nov 12, 2006 at 05:04:16PM -0500, Diane Bruce wrote:
> > > On Mon, Nov 13, 2006 at 03:01:43AM +0800, Rong-En Fan wrote:
> > > > Hi all,
> ...
> > I'm a bit confused. wsjt depends on py-scipy, and py-scipy
> > depends on py-numpy. IIRC, wsjt works with scipy 0.3.x, not
>
> No.
>
> PORTNAME= scipy
> PORTVERSION= 0.3.2
> ...
>
> BUILD_DEPENDS= ${PYNUMERIC} \
>
> What has happened is this.
> 1) the py-scipy that wsjt depends on now, uses py-numeric
>
> PORTNAME= scipy
> PORTVERSION= 0.3.2
> ...
> MAINTAINER= db@db.net
> COMMENT= Scientific tools for Python
>
> BUILD_DEPENDS= ${PYNUMERIC} \
>
> The new scipy 0.5.1 Now depends on numpy. I would assume the python
> folks @ freebsd.org (pointy hat?) should introduce a ${PYNUMPY} as
> we have a ${PYNUMERIC}
> in /usr/ports/Mk/bsd.python.mk
>
> # Python 3rd-party modules
> PYDISTUTILS= ${PYTHON_LIBDIR}/distutils/core.py:${PYTHON_PORTSDIR}
> PYNUMERIC= ${PYTHON_SITELIBDIR}/Numeric/Numeric.py:${PORTSDIR}/math
> /py-numeric
>
> PORTNAME= scipy
> PORTVERSION= 0.5.1
> ...
> BUILD_DEPENDS= ${PYTHON_SITELIBDIR}/numpy/core/numeric.py:${PORTSDIR}/math/py-n
> umpy \
>
> i.e. in /usr/ports/Mk/bsd.python.mk
> PYNUMPY= ${PYTHON_SITELIBDIR}/numy/core/numeric.py:${PORTSDIR}/math
> /py-numpy
>
> Something like that.
>
> 2) upstream has renamed py-numeric to py-numpy
> 3) If you read the pkg-descr for py-numeric /usr/ports/math/py-numeric/pkg-descr
>
> ...
> WWW: http://numeric.scipy.org/
> _____________
> Note: Development for Numeric has ceased, and users should transisition to
> NumPy as quickly as possible.
>
> i.e. present py-scipy is deprecated
> If you go to http://numeric.scipy.org/ you now get the site for
> http://numpy.scipy.org
>
> i.e. py-numeric is living on borrowed time.
>
> 4) Unfortunately, wsjt will not compile with the f2py supplied with
> the new numpy and hence is not usable with the new py-scipy.
> At the present. I am working with the upstream folks at wsjt
> trying to get wsjt to compile under the new scipy and numpy.
>
> 4)
> > 0.5.x. So, if we update scipy to 0.5.x and numpy to 1.0r1,
> > will wsjt with scipy03 break or not?
>
> If we rename the present scipy to scipy03 and keep it dependent
> on numeric wsjt will compile with scipy03, but I will have to rename the
> dependancy in the wsjt port. So both should be changed at or about
> the same time.
>
> 5) Finally: The new py-numpy introduces f2py as part of its package, which
> immediately conflicts with /lang/f2py
>
> 6) As noted from the wsjt-devel list, Debian has both the new py-numpy
> which replaces the old py-numeric and f2py; and the older py-numeric
> and separate standalone f2py.
>
> 7) It's a PITA because of three different port committers for all three
> ports, (scipy, numpy, f2py) trying to synchronise the changes needed.
> I'm perfectly willing to give up maintainership in order to have all
> three ports under one maintainer as this would make much more sense.
> But we still have not heard back from Tony. As a side effect it now
> impacts wsjt in ports.
Tony is already in maintainer timeout mode. So, I can directly
commit to py-numpy. I'm thinking that perhaps we can transfer
wsjt and numpy to freebsd-python@ (I think you know that, right?)
> I hope that was clear this time. *sigh*
Now I really understand everything.
Thanks,
Rong-En Fan
>
> >
> > Thanks,
> > Rong-En Fan
> >
> > >
> > > Now for *everyones* interest (from the wsjt-devel list)
> > >
> > > "
> > > On Sun, Nov 12, 2006 at 04:37:42PM -0500, Diane Bruce wrote:
> > > > It depends on what Ubuntu has installed. f2py + py-numeric
> > > > or py-numpy. I'm curious, what did Ubuntu install?
> > >
> > > FWIW, Debian has both, though they conflict so that only one can be
> > > installed at any time.
> > >
> > > (Which is a nuisance because I use standalone f2py to build wsjt, but
> > > have numpy/scipy installed for some other coding I'm doing.)
> > > "
> > >
> > > So ideally, it would be *nice* if we could somehow switch between
> > > numpy f2py and lang f2py at run time or even with a name change
> > > of the binary. It's complicated here because we have three different
> > > maitnainers. :-(
> > >
> > > Comments?
> > >
> > >
> > > > Regards,
> > > > Rong-En Fan
> > > >
> > > > On Sun, Oct 22, 2006 at 09:09:30PM -0600, Diane Bruce wrote:
> > > > > Hi,
> > > > >
> > > > > Ok this is the latest port for an updated py-scipy. The old py-scipy
> > > > > then gets repomoved to py-scipy03
> > > > > If leland wants port maintainership, he is welcome to it.
> > > > > Note that this port continues to have the make option for using atlas
> > > > > or not, that the present py-scipy port has.
> > > > >
> > > > > I have made two tiny changes to Lelands original PR.
> > > > > 1) atlas as options
> > > > > 2) missing man page put into /usr/local
> > > > >
> > > > > # This is a shell archive. Save it in a file, remove anything before
> > > > > # this line, and then unpack it by entering "sh file". Note, it may
> > > > > # create directories; files and directories will be owned by you and
> > > > > # have default permissions.
> > > > > #
> > > > > # This archive contains:
> > > > > #
> > > > > # .
> > > > > # ./Makefile
> > > > > # ./distinfo
> > > > > # ./pkg-descr
> > > > > # ./files
> > > > > #
> > > > > echo c - .
> > > > > mkdir -p . > /dev/null 2>&1
> > > > > echo x - ./Makefile
> > > > > sed 's/^X//' >./Makefile << 'END-of-./Makefile'
> > > > > X# ex:ts=8
> > > > > X# Ports collection makefile for: SciPy
> > > > > X# Date created: Jun 8, 2004
> > > > > X# Whom: ijliao
> > > > > X#
> > > > > X# $FreeBSD: ports/science/py-scipy/Makefile,v 1.7 2006/08/28 21:21:16 pav Exp $
> > > > > X#
> > > > > X
> > > > > XPORTNAME= scipy
> > > > > XPORTVERSION= 0.5.1
> > > > > XCATEGORIES= science python
> > > > > XMASTER_SITES= ${MASTER_SITE_SOURCEFORGE}
> > > > > XMASTER_SITE_SUBDIR= scipy
> > > > > XPKGNAMEPREFIX= ${PYTHON_PKGNAMEPREFIX}
> > > > > XLATEST_LINK= py-${PORTNAME}
> > > > > XPORT_DBDIR= ${DESTDIR}/var/db/ports
> > > > > XOPTIONSFILE= ${PORT_DBDIR}/py-numpy/options
> > > > > X
> > > > > XMAINTAINER= db@db.net
> > > > > XCOMMENT= Scientific tools for Python
> > > > > X
> > > > > XBUILD_DEPENDS= ${PYTHON_SITELIBDIR}/numpy/core/numeric.py:${PORTSDIR}/math/py-numpy \
> > > > > X ${LOCALBASE}/lib/libdjbfft.a:${PORTSDIR}/math/djbfft \
> > > > > X f2py:${PORTSDIR}/lang/f2py
> > > > > XLIB_DEPENDS= fftw.2:${PORTSDIR}/math/fftw
> > > > > X.if exists(${PYTHON_SITELIBDIR}/wx.pth)
> > > > > XBUILD_DEPENDS+= ${PYTHON_SITELIBDIR}/wx.pth:${PORTSDIR}/x11-toolkits/py-wxPython26
> > > > > X.elif exists(${PYTHON_SITELIBDIR}/wx/__init__.py)
> > > > > XBUILD_DEPENDS+= ${PYTHON_SITELIBDIR}/wx/__init__.py:${PORTSDIR}/x11-toolkits/py-wxPython24
> > > > > X.else
> > > > > XBUILD_DEPENDS+= ${PYTHON_SITELIBDIR}/wx.pth:${PORTSDIR}/x11-toolkits/py-wxPython26
> > > > > X.endif
> > > > > X
> > > > > XUSE_PYTHON= 2.3+
> > > > > XUSE_PYDISTUTILS= yes
> > > > > XOPTIONS= ATLAS "Use optimized blas library" OFF
> > > > > X
> > > > > Xpost-patch:
> > > > > X @${GREP} -lR "malloc\.h" ${WRKSRC} | ${XARGS} ${REINPLACE_CMD} \
> > > > > X -e "s@malloc\.h@stdlib.h@"
> > > > > X
> > > > > X.include <bsd.port.pre.mk>
> > > > > X
> > > > > X.if defined(WITH_ATLAS)
> > > > > XLIB_DEPENDS+= atlas.1:${PORTSDIR}/math/atlas
> > > > > X.if !exists(${LOCALBASE}/lib/libalapack.a)
> > > > > XIGNORE= Atlas needs to be built with WITH_STATICLIB for scipy to function properly
> > > > > X.endif
> > > > > X.endif
> > > > > X.if defined(WITHOUT_ATLAS)
> > > > > XLIB_DEPENDS+= lapack.3:${PORTSDIR}/math/lapack \
> > > > > X blas.1:${PORTSDIR}/math/blas
> > > > > X.endif
> > > > > X
> > > > > Xpost-install:
> > > > > X @${FIND} ${PYTHON_SITELIBDIR}/scipy ! -type d | \
> > > > > X ${SED} 's,^${PREFIX}/,,' >> ${TMPPLIST}
> > > > > X @${FIND} ${PYTHON_SITELIBDIR}/scipy -type d | ${SORT} -r | \
> > > > > X ${SED} 's,^${PREFIX}/,@dirrm ,' >> ${TMPPLIST}
> > > > > X
> > > > > X.include <bsd.port.post.mk>
> > > > > END-of-./Makefile
> > > > > echo x - ./distinfo
> > > > > sed 's/^X//' >./distinfo << 'END-of-./distinfo'
> > > > > XMD5 (scipy-0.5.1.tar.gz) = 48442a427f0556ad2ad1721dd62e401c
> > > > > XSHA256 (scipy-0.5.1.tar.gz) = 312e7562aa04923c02c346866e06fd44e4fcf872c518f64c0f9dabece5ff860a
> > > > > XSIZE (scipy-0.5.1.tar.gz) = 4141008
> > > > > END-of-./distinfo
> > > > > echo x - ./pkg-descr
> > > > > sed 's/^X//' >./pkg-descr << 'END-of-./pkg-descr'
> > > > > XSciPy is an open source library of scientific tools for Python. SciPy
> > > > > Xsupplements the popular Numeric module, gathering a variety of high level
> > > > > Xscience and engineering modules together as a single package.
> > > > > X
> > > > > XSciPy includes modules for graphics and plotting, optimization, integration,
> > > > > Xspecial functions, signal and image processing, genetic algorithms, ODE
> > > > > Xsolvers, and others.
> > > > > X
> > > > > XWWW: http://www.scipy.org/
> > > > > END-of-./pkg-descr
> > > > > echo c - ./files
> > > > > mkdir -p ./files > /dev/null 2>&1
> > > > > exit
> > > > >
> > >
> > >
> > >
> > > --
> > > - db@db.net http://www.db.net/~db
>
>
>
> --
> - db@db.net http://www.db.net/~db
----- End forwarded message -----
--
- db@db.net http://www.db.net/~db
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20061114225403.GA72433>