VDT-Environment sets environment variables so that just sourcing the
VDT's setup.SHELL will let you use Pacman too. Unfortunately, it sets
PYTHONPATH and this causes a problem for LIGO. With recent versions of
Pacman, we no longer need to set PYTHONPATH, so we should do the minimal
set of environment variables to get what we need.
I suspect we also don't set the PATH correctly.
> From: Gregory Mendell <gmendell@ligo-wa.caltech.edu>
> Date: November 1, 2007 5:27:48 PM CDT
> To: John Brunelle <brunejo@bu.edu>, Alain Roy <roy@cs.wisc.edu>
> Cc: Saul Youssef <youssef@bu.edu>, Stuart Anderson
<anderson@ligo.caltech.edu>
> Subject: Re: pacman path added to PYTHONPATH
>
> Hi John,
>
> Thanks for tracking this down!
>
> Hi Alain,
>
> I have an LDG user that has reported a problem because the PYTHONPATH
> includes "pacman-3.21/src" when the user sources the setup.(c)sh file
> after installing LDG. (Note that pacman-3.21/src also gets added to the
> users PATH as well.)
>
> In particular, the user finds the module called "path" under
> pacman-3.21/src differs and causes conflicts with the standard python
> version of this module.
>
> John found these paths get added by lines 26-28 of
>
> http://vdt.cs.wisc.edu/vdt_181_cache/VDT-Environment.pacman.
>
> Do you know whether these lines are needed?
>
> In the short term, I think I can give users instructions on how to
> remove the lines from their setup.(c)sh files, or give them a post
> install script they can run to remove them, if necessary.
>
> Let me know however, if those lines could be removed, or if a VDT
> question could be set to skip those lines in the future.
> From: Alain Roy <roy@cs.wisc.edu>
> Date: November 1, 2007 4:41:18 PM CDT
> To: Gregory Mendell <gmendell@ligo-wa.caltech.edu>
> Cc: John Brunelle <brunejo@bu.edu>, Saul Youssef <youssef@bu.edu>,
Stuart Anderson <anderson@ligo.calte\
ch.edu>
> Subject: Re: pacman path added to PYTHONPATH
>
> On Nov 1, 2007, at 5:27 PM, Gregory Mendell wrote:
>
> > John found these paths get added by lines 26-28 of
> >
> > http://vdt.cs.wisc.edu/vdt_181_cache/VDT-Environment.pacman.
> >
> > Do you know whether these lines are needed?
>
> I don't have a strong opinion on them. I did some archaeology, and
> as best I can tell, this was added to the environment in January of
> 2006 based on some request from OSG. I don't know who,
> unfortunately.
>
> I think the goal was that when users source the VDT's setup file,
> they can use Pacman without needing to source Pacman's setup
> file. We set the PATH and PYTHONPATH because Pacman 3.16 (which was
> the current version at the time) did that in its setup file. It
> appears that more recent versions of Pacman only set PATH and not
> PYTHONPATH.
>
> So I have two questions:
>
> 1) Is this whole idea flawed? Should we not make the VDT environment
find Pacman?
> 2) If we do want to find Pacman, what is the minimum that I should
set? I'm thinking I should set the P\
ATH to <pacman>/bin and set PACMAN_LOCATION, and do nothing else. Is
that right?
>
> If we come to an understanding on what we should do, I'm happy to fix
this in an update to VDT 1.8.1.
>
> -alain
RT for crt.cs.wisc.edu
