From: Geoff Wood (geoffrey.wood_at_epfl.ch)
Date: Mon Sep 24 2007 - 15:05:54 CDT

Hello,

I am writing with regards to the problem associated with python as
outlined by John Stone (see below).

I am wanting to use the IED program and so it looks like I need to go
for option 1 that you have outlined below. We are running Redhat
Enterprise on an x86_64 system.
I have managed to compile Python 2.5.1 and numpy 24.2 but now I am
have trouble building vmd.
First of all I couldn't build the plugins. I set the variable
PLUGINDIR and then tried make LINUXAMD64 but got the error:

/usr/bin/ld cannot find -lnetcdf

So I thought (probably incorrectly) that since I already have this
precompiled in the binary distrubution of vmd I could use them by
copying the directory to the appropriate place. So after running ./
configure LINUXAMD64 PYTHON NUMPY I tried to run make in the src
directory but now I have the error:

Spaceball.h:33:51: sball.h: No such file or directory

Any help on building vmd against my versions of python and numpy
would very much be appreciated.

Cheers, Geoff.

From: John Stone (johns_at_ks.uiuc.edu)
Date: Tue Apr 04 2006 - 16:00:09 CDT

     * Next message: L. Michel Espinoza-Fonseca: "ied on vmd"
     * Previous message: John Stone: "Re: alternative display mode of
HBonds"
     * In reply to: Sacerdoti, Federico: "python Numeric library
problem"
     * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
[ attachment ]

Hi Federico,
   The VMD binaries we distribute are all compiled against the same
version of Python, which has been 2.2.2 thus far. Since every linux
distribution (and indeed, almost every major OS in fact) comes with a
different version of Python, linking against the system-provided Python
interpreter won't work for VMD.

Where this leaves those of us wanting to nontrivial stuff that requires
the full Python library in VMD:
   1) If you want to use the same Python setup that's on your system
already,
      it's currently much easier to compile VMD from source code and
link it
      against your local Python than to try and get around the Python
library
      version issues. If people need help with this, contact me and I
can
      help you get past any sticking points in your builds. You can
either
      build from one of the source distributions, or you can get the
latest
      VMD source code from CVS.

   2) If you want to use John Mongan's IED plugin, or other things
that need
      Tkinter, you're probably also headed for 1) above.
      (tkinter is usually dynamically linked against whatever Tcl/Tk
       version is favored by a particular Linux distribution, so
unless you
       go through great pains in the Python build and are also lucky,
the
       Python tkinter module built on one system may not work on
another)

Regarding upgrading VMD to one of the more recent Python revs:
   Several times I've tried to upgrade to Python 2.3 and 2.4 only to
find
serious problems compiling the newer versions on several of the
64-bit platforms that VMD supports, and I won't even mention the fun
involved in building Python from source on Windows if you want to
add Numeric and Tkinter support, which is something most VMD users
want so they can run IED and similar tools.
I've reported these problems in the Python bug tracking system, and
I've provided the Python developers fixes for the ones I could solve
for myself, but the problems still continue and I haven't had time
to go through another attempt since the last one a few months back.
In the mean time it would seem that I'm going to have to stick with
2.2.2
for the 1.8.4 release, though it pains me greatly.

What I may do to ease the pain for Python hackers is to try and release
Python 2.4 based VMD 1.8.4 binaries in addition to the normal ones, for
the few platforms that they'll work fine for. (Linux for starters)

   John

Geoffrey Wood
Ecole Polytechnique Fédérale de Lausanne
SB - ISIC - LCBC
BCH 4108
CH - 1015 Lausanne