BUG: port numpy should depend on python24

David Tremouilles david.trem at gmail.com
Fri Jul 27 07:21:24 PDT 2007


Hello,

Chris, I think you asked me directly but I your email get lost... I'm
sorry about that :-(

Here is a reply to the g95 related problem (see next paragraph)
With this "trick" I was able to successfully compile scipy svn 0.5.3 dev3099
Somebody else (Punch) reported me additional problems with scipy 0.5.2
he did not solved.
He find a solution using a precompiled scipy version for python 2.5 at :
http://idisk.mac.com/fonnesbeck-Public/?view=web

 The one with g95 was really tricky to fix:
The problem is related to numpy disutils that does not work for MacOsX.
A dirty fix is to edit the g95.py file in
/opt/local/lib/python2.4/site-packages/numpy/distutils/fcompiler
and modified it to obtain the following:

executables = {
       'version_cmd'  : ["g95", "--version"],
       'compiler_f77' : ["g95", "-ffixed-form"],
       'compiler_fix' : ["g95", "-ffixed-form"],
       'compiler_f90' : ["g95"],
       'linker_so'    : ["g95", '-undefined', 'dynamic_lookup',
'-bundle'], # ["g95", "-shared"] #here is the mod
       'archiver'     : ["ar", "-cr"],
       'ranlib'       : ["ranlib"]
       }

and then it should compile smoothly :-)

Hope you will also succeed !

Regards,

David


2007/7/27, Christopher W. MacMinn <cmac at mit.edu>:
> Mark -
>
> I tried to install the port py-scipy on an Intel Mac last week with
> similar results.
>
> >> a lot of dependecies were
> >> installed, but python24 wasn't (at least at the time that numpy was
> >> built). So I guess that numpy needs a dependency on python24.
>
> >> Error: The following dependencies failed to build: py-numpy
> >> Error: Status 1 encountered during processing.
>
> I did not run into problems with the python24 dependency, but I may
> have already had it installed.
>
> > now I'm stuck at building scipy itself.
>
> > In scipy-0.5.2/Lib/odr/setup.py, Line 9 method "get_path" is
> > imported from some numpy module, but that method doesn't exist there.
>
> I definitely encountered this one too -- after digging around for
> awhile and posting to several lists, I filed a bug report in the
> MacPorts Trac system (#12331, http://trac.macports.org/projects/
> macports/ticket/12331).  I also contacted Erick Tryzelaar, the
> maintainer for py-numpy and py-scipy, earlier this week, and he
> seemed to be aware of this bug and said that he hoped to have an
> update out soon.
>
> > Replacing
> >
> > from numpy.distutils.misc_util import get_path, Configuration,
> > dot_join
> >
> > by
> >
> > from numpy.distutils.misc_util import Configuration, dot_join
> >
> > solves the issue. However, I don't know why it was imported because
> > it is not used in Lib/odr/setup.py at all.
>
> I resolved this differently than you did (by replacing it with a call
> to get_path_from_frame instead of simply removing it -- see my bug
> report above), but the end result seems to be the same for both of us:
>
> > Now, having that fixed, the next error seems to be a linking error.
> > Obviously someone forgot to link against the Python Framework /
> > libpython2.4.dylib. Linking happens to fail for every Fortran module,
> > cluster and io on the other hand are built without error. Could
> > someone shed light into this?
> >
> >> --->  Building py-scipy with target build
>
> >> error: Command "g95 -shared build/temp.macosx-10.3-i386-2.4/build/
> >> src.macosx-10.3-i386-2.4/Lib/fftpack/_fftpackmodule.o build/
> >> temp.macosx-10.3-i386-2.4/Lib/fftpack/src/zfft.o build/
> >> temp.macosx-10.3-i386-2.4/Lib/fftpack/src/drfft.o build/
> >> temp.macosx-10.3-i386-2.4/Lib/fftpack/src/zrfft.o build/
> >> temp.macosx-10.3-i386-2.4/Lib/fftpack/src/zfftnd.o build/
> >> temp.macosx-10.3-i386-2.4/build/src.macosx-10.3-i386-2.4/
> >> fortranobject.o -L/opt/local/lib -Lbuild/temp.macosx-10.3-i386-2.4 -
> >> ldfftpack -lfftw3 -o build/lib.macosx-10.3-i386-2.4/scipy/fftpack/
> >> _fftpack.so" failed with exit status 1
>
> I'm getting this too, and here I am also stuck.  Any leads?  I think
> I have exhausted my meager MacPorts skills.  I'll let you know if I
> hear anything else from Erick, and please do let me know if you
> figure out a fix in the meantime.
>
> Best
> - Chris MacMinn
> _______________________________________________
> macports-users mailing list
> macports-users at lists.macosforge.org
> http://lists.macosforge.org/mailman/listinfo/macports-users
>



More information about the macports-users mailing list