Core Minutes 6/21/2005
Present: Joanne Bogart, Anders Borgland, Toby Burnett, Seth Digel,
Richard Dubois, Zachary Fewtrell, Dan Flath, Berrie Giebels,
Tom Glanzman, Navid Golpayegani,
Tony Johnson, Heather Kelly, Michael Kuss, Julie McEnery,
Chuck Patterson, Igor Pavlin, James Peachey, Dirk Petry, Leon Rochester,
Tom Stephens
- I & T: (Anders) [updated Tuesday PM] The
6-tower test run was taken Tuesday. Testing is currently delayed because of
several tracker problems. CA: calibrations will most probably now only happen
next week.
- EM tags: (Heather) Latest tag, released
yesterday, is v5r0608p1. New content as compared to p0 is
a new tag of CalXtalResponse from Zach. Systems tests for this tag have
several failed plots; Anders has looked at them and believes there are
no serious problems. Heather will add some words to this effect to
the CCB Confluence page.
This tag still has the shift in
simulated event time, even though Toby has tagged a fix. Heather is
hesitant to pick it up unless someone can assure her she
won't be picking up other, possibly unwanted stuff along with it. CRflux
unit test is failing. Toby
believes that problems, if any, should not affect surface muons, hence
would have no impact on EM. He recommends forging ahead.
Anders found a problem with some hot/dead strips calibration files,
now fixed.
- Data handling: (Tony)
- New CVS repository has been set up for Java projects, which
tend to have a different organization and evolution than C/C++.
CVSspam
will email nicely-formatted html message upon commit.
In 6 months
or so, may move to Subversion.
- Pipeline Dan had a problem when he last tried to use Navid's
archiver. [See Archiver item below.]
- Confluence changes Tony has revamped SLAC Confluence somewhat.
In the process some people appear to have lost access to some pages; Tony
suspects he knows what the problem is and how to fix it.
- Schedule Master timeline in progress.
- Archiver
(Navid) New version is ready to go, hence it probably is not
worth trying to debug a problem with the old one. The new version uses
the batch system rather than directing all operations to glast03.
- Windows build: (Richard, Navid) Windows
builds are still limping along on a single inadequate box because SCS
has not yet set up our new machines.
There is a problem doing builds on our CITRIX server: since it is running
Windows 2000 rather than XP there is a 2000-character limit on commands, which
we have surpassed. Navid is implementing a simplified install area.
To start with, he will handle just libraries, though
he would like to do something similar for executables and for P-files.
Implementation will just be unix or Windows shell commands to do copies
after the objects are built.
- End run evades tags: Recently Bill requested
a million all-γ events using HEADs of 8 packages along with tagged
versions of the rest. After some minor pilot error, Richard was able to
produce them, however many are unhappy with the precedent. There were
extenuating circumstances (Tracy, whose packages these were, had almost no
internet access; the events are specially for Bill, not intended for public
consumption); still, Bill did present
results
based on this run at the
Analysis Group meeting yesterday. Toby suggested we now tag the packages
we believe were used (perhaps with non-standard tags if standard ones
would somehow interfere with other work) and re-run.
- Save trees or reproduce? Julie has suggested
that, rather than save all ROOT trees from all runs, we might make use
of the reproducibility of events from the incoming particles for the times
someone wants the full tree. (Michael) there was a problem with
reproducibility in v6r8, but it now seems to be OK. (Toby) An all-γ
run is relatively simple. Even if we couldn't reproduce it exactly, it
wouldn't matter since it's easy to reproduce it in a statistical sense.
However that can't be said of background runs, which have a much larger
parameter space. Here we need per-event reproducibility. We ought to
incorporate something into System Tests which checks for reproducibility,
given input particles (and random seed) information. Michael will think
about how to do this.
- AOB: (Toby)
Software builder SCons is getting a lot
of good press.
(Richard) We may have passed the point of no return with CMT.
J. Bogart, Last Modified:
01-Jun-2010 15:47:53 -0700