Re: python 2.7 instead of 2.6?

[Erik B. <eb...@fs...>]

On Thu, 12 Aug 2010, Peter Kuma wrote:
> Hi Erik,
> The PyNIO binary should match exactly with the version of Python
> installed. The easiest way would be to install Python 2.6.2. Precompiled
> version of PyNIO for Python 2.7 is not yet available, and compiling it
> with all the third-party libraries it needs would take a long time.
> Regards,
> Peter Kuma

Hi Peter,

I cleared the cruft from all the software, and I compiled Python 2.6.2. I 
then re-extracted and recompiled the other software: numpy, matplotlib, 
and basemap. Things seemed to go well enough and any tests ran ok. 
However, when I installed PyNIO and typed '# /usr/local/bin/python -c 
"import Nio;"' the error was essentially identical. To debug, I typed '# 
/usr/local/bin/python -v -c "import Nio;"'. Here is the output just 
before the text of the error output I emailed you last time:
[...]
dlopen("/usr/local/lib/python2.6/lib-dynload/datetime.so", 2);
import datetime # dynamically loaded from 
/usr/local/lib/python2.6/lib-dynload/datetime.so
# /usr/local/lib/python2.6/site-packages/PyNIO/_xarray.pyc matches 
/usr/local/lib/python2.6/site-packages/PyNIO/_xarray.py
import _xarray # precompiled from 
/usr/local/lib/python2.6/site-packages/PyNIO/_xarray.pyc
Traceback (most recent call last):
   File "<string>", line 1, in <module>
   File "/usr/local/lib/python2.6/site-packages/PyNIO/Nio.py", line 145, in 
<module>
     _is_new_ma = get_integer_version(np.__version__) > 10004
[...]

I'm not sure what this tells us except it seems to be loading, then 
crashes.

This is a non-sequiter, but could this be numpy related? I only ask 
because I saw during compile on numpy (using --fcompiler=gnu95) this 
remark:
[...]
   FOUND:
     libraries = ['blas']
     library_dirs = ['/usr/lib64']
     language = f77
[...]
Yet the libraries seem dependent on gfortran, not f77 (an alias to g77).
# ldd libblas.so.3.0.3
         libm.so.6 => /lib64/libm.so.6 (0x00002b12ae958000)
         libgfortran.so.1 => /usr/lib64/libgfortran.so.1 
(0x00002b12aebdb000)
         libc.so.6 => /lib64/libc.so.6 (0x00002b12aee72000)
         /lib64/ld-linux-x86-64.so.2 (0x0000003a3f800000)

I've verified all my blas and lapack shared object files are dependent on 
libgfortran.

By the way, I had installed lapack-devel and blas-devel rpms before 
recompiling numpy. (I'm not using Atlas. From what I read at the Atlas 
site, it provides a speedier blas/lapack.) Another oddity when 
compiling numpy is:
[...]
customize GnuFCompiler
gnu: no Fortran 90 compiler found
gnu: no Fortran 90 compiler found
customize GnuFCompiler using config
C compiler: gcc -pthread -fno-strict-aliasing -DNDEBUG -g -fwrapv -O3 
-Wall -Wstrict-prototypes -fPIC
[...]
There are other warnings of various kinds while building numpy. But again, 
the test in the README.txt runs with 4 failures of 2962 tests and 
$?=0.

Erik