[SciPy-user] compiling scipy on Fedora Core 4 -- undefined reference to PyInt_Type --follow-up

David Mobley dmobley at gmail.com
Fri Feb 3 13:23:01 CST 2006


Following up on the below problems, I'm still trying to compile scipy
0.4.4under RedHat Fedora Core 4. I'm now using gcc
4.0.2 and atlas 3.6.0-8, blas 3.0-36, lapack 3.0-35, numpy 0.94, numarray
1.4.0, and Numeric 24.2. We've also tried on an Opteron (fedora core 4) with
no success. I'm still getting the same error when I try to compile as
before:

> The first error is "In function 'int_from_pyobj':
build/src/Lib/fftpack/_fftpackmodule.c:109: undefine d reference to
'PyInt_Type'.

Any further suggestions?

This is all because I'm hoping to install f2py (although I anticipate using
scipy for other reasons later) to be able to interface with some fortran
code I'm trying to use for my research relating to computational drug
design.

Anyway, any help will be greatly appreciated. I still suspect I've done
something fundamental wrong, as it doesn't seem like this should be so
difficult.

Thanks,
David


On 2/2/06, David Mobley <dmobley at gmail.com> wrote:
> > Dear All,
> >
> >  I'm sure I've done something stupid, but I haven't been able to turn up
> > anything useful with Google, so I need to appeal to the list for help.
> >
> >  I'm trying to install scipy under RedHat Fedora Core 4 on a Xeon, and
> the
> > build crashes with a whole bunch of "undefined reference to..." errors
> when
> > it gets to the step of running g77 on various fft and fftpack object and
> > shared object libraries. The first error is "In function
> 'int_from_pyobj':
> > build/src/Lib/fftpack/_fftpackmodule.c:109: undefined
> > reference to 'PyInt_Type'.
> >
> >  Can anyone help? I already have NumPy installed and in my PYTHONPATH,
> and
> > also turned up on the web various suggestions about using
> lapack/blas/atlas
> > even though they're not absolutely required, so I tried with those as
> well
> > as without them (first with just lapack and blas, then adding ATLAS) and
> > encountered the same problem every time.
> >
> >  Am I missing something eseential?
> >
> >  Thanks very much,
> >  David Mobley
> >
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.net
> > http://www.scipy.net/mailman/listinfo/scipy-user
> >
> >
> >
>
> Which version of scipy are you trying to build? tried the 0.4.4?
>
> be sure to yum update everything first. also the Numeric package
> (python-numeric-23.7-2) for FC4 is kinda out of date, grab the latest
> one from sourceforge. Be sure to upgrade your gcc to at least 4.0.2.
>
> I am using FC4 on a P4 machine, the following is what i have, not all
> of them are needed.
>
> blas-3.0-36.fc4
> atlas-3.6.0-8.fc4
> lapack-3.0-36.fc4
> gcc-4.0.2-8.fc4
> python-numarray-1.3.3-4.fc4
>
> fftw-2.1.5-8
> fftw3-3.0.1-3
> Numeric 24.2
>
> scipy-0.4.4
>
>
> --
> iCy-fLaME
> The body maybe wounded, but it is the mind that hurts.
>
> ------------------------------
>
> Message: 9
> Date: Thu, 02 Feb 2006 11:30:54 -0800
> From: Gregory Novak <novak at ucolick.org>
> Subject: [SciPy-user] What shape should arg 2 of integrate.trapz be?
> To: SciPy Users List <scipy-user at scipy.net>
> Message-ID: <m2acd94ky9.fsf at euterpe.local>
> Content-Type: text/plain; charset=us-ascii
>
> If I do this:
>
> (Note: nrange is like arange, except you specify the number of samples
> instead of dx).
>
> In [474]: x = nrange(-1,1,4)
> In [475]: y = nrange(-1,1,5)
> In [476]: z = reshape(arange(20.), (4,5))
> In [479]: shape(integrate.trapz(z,axis=0))
> Out[479]: (5,)
>
> the output is as I expect: I integrate away the (length 4) x dimension
> and am left with an answer for each y value.
>
> However, if I try to specify the x axis, I get the following traceback:
>
> In [480]: shape(integrate.trapz(z,x,axis=0))
>
> ---------------------------------------------------------------------------
> /sw/lib/python2.3/site-packages/scipy/integrate/quadrature.py in trapz(y,
> x, dx, axis)
>      96     slice1[axis] = slice(1,None)
>      97     slice2[axis] = slice(None,-1)
> ---> 98     return add.reduce(d * (y[slice1]+y[slice2])/2.0,axis)
>      99
>     100 def cumtrapz(y, x=None, dx=1.0, axis=-1):
>
> ValueError: frames are not aligned
>
>
> I can make this work again by explicitly making the x array be 2dim
> instead of 1dim:
>
> In [481]: shape(integrate.trapz(z,x[:,NewAxis],axis=0))
> Out[481]: (5,)
>
> The question is: Is this the desired behavior?  It seems to me that
> In[479] is well-specified and should give the correct result.  Is
> there some reason that one would want to pass in a multi-d array as
> the grid for the independent variable?
>
> Thanks,
> Greg
>
>
>
> ------------------------------
>
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-user
>
>
> End of SciPy-user Digest, Vol 30, Issue 3
> *****************************************
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.scipy.net/pipermail/scipy-user/attachments/20060203/fa02c623/attachment.htm


More information about the SciPy-user mailing list