Problem building NumPy with information from ~/.numpy-site.cfg

I am trying to build numpy 1.1.1 with ATLAS support. the f77blas
library I am trying to use is named f77blas_gfortran to distinguish it
From builds for other compilers. First from the documentation it is
unclear whether the default section is named DEFAULT or ALL (I tried
both). So I set

[DEFAULT]
library_dirs = /usr/local/gltools/kinux/lib:/usr/local/gltools/linux_x86_64/lib
include_dirs = /usr/local/gltools/include:/usr/local/gltools/linux/include:/usr/local/gltools/linux_x86_64/include

[blas_opt]
libraries = f77blas_gfortran, cblas, atlas

[lapack_opt]
libraries = lapack_gfortran, f77blas_gfortran, cblas, atlas

in my ~/.numpy-site.cfg but the build process does not find LAPACK
support and builds lapack_lite.so with the built in slow lapack.

/usr/local/gltools/linux_x86_64/lib/libf77blas_**.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_g77.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_gfortran.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_ifc100.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_ifc101.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_ifc90.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_ifc91.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_pg52.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_pg60.so
/usr/local/gltools/linux_x86_64/lib/libf77blas_pg61.so

Re: Problem building NumPy with information from ~/.numpy-site.cfg

On Mon, 2008-09-01 at 11:26 +0200, Berthold "Höllmann" wrote:
> I am trying to build numpy 1.1.1 with ATLAS support. the f77blas
> library I am trying to use is named f77blas_gfortran to distinguish it
> From builds for other compilers. First from the documentation it is
> unclear whether the default section is named DEFAULT or ALL (I tried
> both). So I set
> 

Hi Berthold,

	To help you, we need more information: the build log would be useful.
Please re-run the python setup.py build from scratch (deleting the build
directory):

python setup.py build &> build.log

Thank you,

David

_______________________________________________
Numpy-discussion mailing list
Numpy-discussion <at> scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: Problem building NumPy with information from ~/.numpy-site.cfg

_______________________________________________
Numpy-discussion mailing list
Numpy-discussion <at> scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: Problem building NumPy with information from ~/.numpy-site.cfg

Nadav Horesh <nadavh <at> visionsense.com> writes:

> In the [DEFAULTS]  library_dirs you wrote kinux instead of linux --- is
> it on purpose?

It is not on purpose, but for my case it should not matter, because
the libraries in `/usr/local/gltools/linux/lib` are 32Bit where I am
building a 64Bit version of numpy where the libraries are in
`/usr/local/gltools/linux_x86_64/lib`. I am trying to write a
'~/.numpy-site.cfg` valid for out 32Bit Linux Systems as well as for
our 64Bit Linux System. I get the same NFS mounted home directory on
all or our machines.

Kind regards

Berthold Höllmann
--

-- 
Germanischer Lloyd AG
CAE Development
Vorsetzen 35
20459 Hamburg
Phone: +49(0)40 36149-7374
Fax: +49(0)40 36149-7320
e-mail: berthold.hoellmann <at> gl-group.com
Internet: http://www.gl-group.com

_______________________________________________
Numpy-discussion mailing list
Numpy-discussion <at> scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

how to freeze a numpy based script ?

    Hi,

I would like to freeze a numpy based script in order to have an
application which could run without having to install numpy and cie.

Indeed, I'm not root on the targeted computer and so, I can't easily
make a complete install of numpy and scipy.

So I decided to test the freeze.py tool shipped with python2.5. To
complicate matters, I must say that I only have a local installation
of python2.5 and numpy.

I used the following command line :
> python2.5 ../Python-2.5.1/Tools/freeze/freeze.py ~/Python/numpy/test_freeze.py

At first sight, it seems to be fine, as I saw numpy in the liste of
frozen dependancies :
        [...]
freezing numpy ...
freezing numpy.__config__ ...
freezing numpy._import_tools ...
freezing numpy.add_newdocs ...
freezing numpy.core ...
freezing numpy.core._internal ...
freezing numpy.core.arrayprint ...
freezing numpy.core.defchararray ...
freezing numpy.core.defmatrix ...
freezing numpy.core.fromnumeric ...
freezing numpy.core.info ...
freezing numpy.core.memmap ...
freezing numpy.core.numeric ...
freezing numpy.core.numerictypes ...
freezing numpy.core.records ...
freezing numpy.ctypeslib ...
              [...]
freezing numpy.version ...

But at the end I saw this message :
Warning: unknown modules remain: _bisect _csv _ctypes _curses _hashlib
_heapq
              [...]
numpy.core._dotblas numpy.core._sort numpy.core.multiarray
numpy.core.scalarmath numpy.core.umath numpy.fft.fftpack_lite
numpy.lib._compiled_base numpy.linalg.lapack_lite numpy.random.mtrand
operator parser pyexpat readline
              [...]
Now run "make" to build the target: test_weibull

I runned make without any problem but the final application didn't
work :
% ./test_freeze
Traceback (most recent call last):
  File "/home/loic/Python/numpy/test_freeze.py", line 8, in <module>
    import numpy             as np
             [...]
  File "/home/loic/tmp/bluelagoon/lib/python2.5/site-packages/numpy/
core/__init__.py", line 5, in <module>
    import multiarray
ImportError: No module named multiarray

Is there any known issue when freezing a numpy based script ?
I should add that I configured my PYTHONPATH to match my local
installation
 > echo $PYTHONPATH
/home/loic/lib/python:/home/loic/tmp/bluelagoon/lib/python2.5:/home/
loic/tmp/bluelagoon/lib/python2.5/site-packages/:

and this local installation work fine :
> python2.5 -c 'import numpy; print numpy.__version__; import numpy.core.multiarray; print "no pb"'
1.2.0b2
no pb

Have you got any recipe to freeze numpy based script ?

Regards,

--
LB

Windows_XP_x86_64_MSVC buildbot needs clean install directory

Hi,

The buildbot slave Windows_XP_x86_64_MSVC seems to have some bogus files 
left over in its installation directory (in numpy\doc\) which cause the 
tests to fail.

Removing the installation directory probably helps.

--

-- 
Pauli Virtanen

numpy failures for rc1

I was hoping to tag the 1.2.0rc1 soon, but I noticed that a couple of
failures on the buildbots:
  http://buildbot.scipy.org/builders/Windows_XP_x86_64_MSVC/builds/701/steps/shell_2/logs/stdio
  http://buildbot.scipy.org/builders/FreeBSD_64/builds/621/steps/shell_2/logs/stdio

Are there any other issues that need to be resolved before tagging the rc?

Thanks,

--

-- 
Jarrod Millman
Computational Infrastructure for Research Labs
10 Giannini Hall, UC Berkeley
phone: 510.643.4014
http://cirl.berkeley.edu/

Re: numpy failures for rc1

Jarrod Millman wrote:
> I was hoping to tag the 1.2.0rc1 soon, but I noticed that a couple of
> failures on the buildbots:
>   http://buildbot.scipy.org/builders/Windows_XP_x86_64_MSVC/builds/701/steps/shell_2/logs/stdio
>   http://buildbot.scipy.org/builders/FreeBSD_64/builds/621/steps/shell_2/logs/stdio
>
>   

The FreeBSD one is a bogus test (again, my fault), which should be
trivial to fix. For the Win64 problem, I don't understand how it would
appear only on windows; maybe a build bot configuration problem ?

cheers,

David

Re: Windows_XP_x86_64_MSVC buildbot needs clean install directory

Pauli Virtanen schrieb:
> Hi,
> 
> The buildbot slave Windows_XP_x86_64_MSVC seems to have some bogus files 
> left over in its installation directory (in numpy\doc\) which cause the 
> tests to fail.
> 
> Removing the installation directory probably helps.
> 

It did.

--

-- 
Thanks,
Thomas

Re: Problem building NumPy with information from ~/.numpy-site.cfg

David Cournapeau <cournapeau <at> cslab.kecl.ntt.co.jp> writes:

> On Mon, 2008-09-01 at 11:26 +0200, Berthold "Höllmann" wrote:
>> I am trying to build numpy 1.1.1 with ATLAS support. the f77blas
>> library I am trying to use is named f77blas_gfortran to distinguish it
>> From builds for other compilers. First from the documentation it is
>> unclear whether the default section is named DEFAULT or ALL (I tried
>> both). So I set
>> 
>
> Hi Berthold,
>
> 	To help you, we need more information: the build log would be useful.
> Please re-run the python setup.py build from scratch (deleting the build
> directory):
>
> python setup.py build &> build.log

I tried to send the logfile to the list, but the mail still awaits
moderation.

Kind regards

Berthold Höllmann
--

-- 
Germanischer Lloyd AG
CAE Development
Vorsetzen 35
20459 Hamburg
Phone: +49(0)40 36149-7374
Fax: +49(0)40 36149-7320
e-mail: berthold.hoellmann <at> gl-group.com
Internet: http://www.gl-group.com

_______________________________________________
Numpy-discussion mailing list
Numpy-discussion <at> scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion