...

1. Execute ./configure --prefix /afs/slac.stanford.edu/g/glast/ground/GLAST_EXT/redhat6-x86_64-64bit-gcc44/python/2.7.2/gcc34 -fPIC
2. Execute make
3. Execute make install
4. Leave and erase the current directory (Python-2.7.2)

5. Wiki Markup

   install additional python modules the same way as python 2.7.1 instruction. As for matplotlib, most of matplolib unit tests requires functions from PIL imaging module which in turn requirez zlib.crc32 which has been an issue for us. After consulting with the users, we agreed that if in the future the users start to use functions in PIL, we will come back to re-visit the PIL issue again. For now, we will just skip the PIL module installation as we always have. Also there is a Scipy unit test need to be fixed after the Scipy installation. Edit scipy/stats/tests/test_morestats.py, change line #10 from "assert_raises" to "assert_raises, assert_" and change line #72 from "assert_array_less(crit\[:-1\], A)" to "assert_(A > crit\[-1\])".

pipcurl https://raw.github.com/pypa/pip/master/contrib/get-pip.py | python

NumPy - 1.6.1

Still installing by hand to include --fcompiler flag to force correct fortran compiler to be used.  This is dictated by whatever blas and lapack are linked against.  For more information, see the numpy INSTALL.txt file. On RHEL6-64:
glastrm@noric22 $ ldd /usr/lib64/liblapack.so
linux-vdso.so.1 => (0x00007fff49dff000)
libblas.so.3 => /usr/lib64/libblas.so.3 (0x00007fac2ac9d000)
libm.so.6 => /lib64/libm.so.6 (0x00007fac2aa18000)
libgfortran.so.3 => /usr/lib64/libgfortran.so.3 (0x00007fac2a726000)
libc.so.6 => /lib64/libc.so.6 (0x00007fac2a394000)
/lib64/ld-linux-x86-64.so.2 (0x00007fac2b72d000)

where libgfortran means we should use the gnu95 flag

1. Download numpy-1.6.1.tar.gz from http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103&release_id=628858
2. Extract numpy-1.6.1.tar.gz
3. Change directory to numpy-1.5.1
4. python setup.py build --fcompiler=gnu95   (for gfortran)
5. python setup.py install
6. Run the test routine by starting up python and doing:
   import numpy
   numpy.test()
7. Leave and erase the current directory (numpy-1.5.1) 

2.7.1

Adding in the following modules:

...