[lsst-dm-stack-users] A test development version of Winter'14 stack

[Mario Juric]

On 2/19/14, 10:33 , Chris Walter wrote:
> Hi Mario,
> 
> OK, I also built with no problems on 10.9.1.  Nice job!
> 
> I have a few questions on how to use this version of the system:
> 
> - now loadLSST.sh etc are gone. So, I take it we should just 'source 
> eups/bin/setups.csh' etc now. Is that right? It does everything we need?
> 

Correct. It will set up EUPS, and then EUPS can be used to set
everything else up.

> - Then we do a setup.  Can I do the same setup I did before?  
> 
> say: 'setup pipe_tasks'?  Or do I need to 'setup lsst_distrib' etc?
> 

Doing 'setup pipe_tasks' should "just work" (if it doesn't, it's a bug).
You should never have to setup lsst_distrib (see below).

> - What is the difference between 
> 
> eups distrib install lsst

This is a package that sets up some often-used convenience environment
variables. From its table file:

envSet(GITROOT, "git at git.lsstcorp.org")
envSet(LSST_GIT, "git at git.lsstcorp.org")
envSet(LSST_DMS, "git at git.lsstcorp.org:LSST/DMS")
envPrepend(PATH, ${PRODUCT_DIR}/bin)
envPrepend(EUPS_PKGROOT, http://lsst-web.ncsa.illinois.edu/~mjuric/pkgs, |)

The only one you may care about is EUPS_PKGROOT, which tells EUPS where
to look when you wish to install (or upgrade) packages with 'eups
distrib install'. But otherwise, you do not need it to run any of the
code (bottom line: you don't need to set it up).

> eups distrib install lsst_distrib
> eups distrib install lsst_apps ?
> 

These are meta-packages that make it easier to install all of LSST code
(incl. the middleware for running on clusters), and just the science
pipelines (the "apps"). There should be no need to ever setup these
directly.

> 
> - How can I tell what is a stub or not (either before or after
> installation)?  'eups list' no longer shows 'system' for matplotlib
> etc.  I found a note from Robert that said 'eups startup' would show
> the manifest but for me that doesn't return anything.
> 

There's really no good way to disambiguate the stubs right now -- I'm
treating it as a bug that I plan to fix by renaming them to
"<name>_check" (e.g., python_check, matplotlib_check, etc...).

Hope this helps,
-- 
Mario Juric,
Data Mgmt. Project Scientist, Large Synoptic Survey Telescope
Web : http://research.majuric.org     Phone : +1 617 744 9003