Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

2.7.6

redhat5-i686-32bit-gcc41 & redhat6-x86_64bit-gcc44

  1. Obtain python source from python.org
  2. Ungzip/Untar source
  3. Change directory to Python-2.7.6
  4. edit setup.py, in detect_modules() change the lines from:
    add_dir_to_list(self.compiler.include_dirs, '/usr/local/include')
    add_dir_to_list(self.compiler.library_dirs, '/usr/local/lib')

    to:

    add_dir_to_list(self.compiler.include_dirs, '/usr/local/include')
    add_dir_to_list(self.compiler.library_dirs, '/usr/local/lib')
    add_dir_to_list(self.compiler.library_dirs, '/usr/lib')
    add_dir_to_list(self.compiler.include_dirs, '/usr/include')

  5. Execute ./configure --prefix /afs/slac.stanford.edu/g/glast/ground/GLAST_EXT/redhat5-i386-32bit-gcc41/python/2.7.6
  6. Execute make
  7. Execute make install
  8. Leave and erase the current directory (Python-2.7.6)

...

Download scipy-0.10.1.tar.gz from http://www.scipy.org/
Extract scipy-0.14.0.tar.gz
Change directory to scipy-0.14.0
Execute python setup.py config_fc --fcompiler=gnu95 install 
Leave and erase scipy-0.14.0
Test by starting up python and doing:
import scipy
scipy.test()  (NOTE this segfaulted on rhel5-32.. hoping that's not a big deal)

...

pip install ipython
Run the iptest located in python's bin directory.

2.7.3

Windows-i386-32bit-vc90 debug

...

to support HTML notebooks:

pyzmq which requires zeromq a C lib: http://zeromq.org/area:download

pip install pyzmq --install-option="–zmq=$GLAST_EXT/zeromq/4.0.4"

Jinga2: http://jinja.pocoo.org/docs/intro/#installation

pip install Jinja2

...

2.7.2-gl4

started really building the shared python library but also need the static lib for GR and ST too.  Need to hide the shared library during installation to avoid using it during the install process for things like numpy.

...

Unable to install numpy 1.5.1 successfully on rhel4-64, due to a known bug with 64 bit and gcc3.  The numpy unit test fails.  Fix is to move to gcc4, which we cannot do.  For now remaining on python 2.6.5.  In either case, the scipy unit tests results some errors which we seemingly have always been dealing with.  The bug report suggests doing a new install of BLAS, but according to unix-admins that is not possible.

redhat5-i686-32bit-gcc41

Same directions as redhat4-i686-32bit-gcc34.  The fortran compiler to use for numpy and scipy install is gfortran so --fcompiler=gnu95.  There should be no errors in the numpy or scipy unit tests.

...