Core Minutes 11 December 2007

Core Minutes 12/11/2007

Present: Joanne Bogart, Anders Borgland, Philippe Bruel, Toby Burnett, Rob Cameron, Emmanuel Cephas, David Chamont, Eric Charles, Jim Chiang, Johann Cohen-Tanugi, Viktor (EVO Support), Warren Focke, Berrie Giebels, Tom Glanzman, Navid Golpayegani, Elizabeth Hays, Heather Kelly, David Landriu, Julie McEnery, Chuck Patterson, Igor Pavlin, Leon Rochester, Tracy Usher, Eric Winter

GlastRelease/Big run: Not really discussed here; see the following Calibration & Analysis meeting. The likely culprits in the Case of the 5 GeV Notch are probably in the general area of ACD veto thresholds and associated calibrations. (Toby) has also seen problems apparently associated with ACD calibrations which are preventing him from making all-gamma runs.
RHEL4 problem: (Heather) Zach has been battling a problem, believed to be in the vicinity of CalCalibSvc, which causes crashes in a noticeable percentage of runs, but only when running optimized RHEL4 32-bit code on RHEL4 64-bit machines. Because of this we've had to back off and are using rh9_gcc32opt builds rather than rhel4_gcc34opt. (Joanne) made an alternate version of the relevant shareable with symbols (the one in the release build was stripped) in the hope that Zach could use it to garner more information about the crash..if it doesn't make the crash go away altogether. No word yet from Zach about it.
Composite Event Lists and Interleave: (Tracy) Looks like CEL is not intended to do quite what Interleave needs. He has been able to incorporate it, but it's not designed for an environment in which different input files are being opened. (David C.) realizes he didn't understand well enough what the requirements are for Interleave. Once he does, he believes he can write something to support them efficiently.
OnboardFilter: (Heather) What is the status of the diagnostic filter? (Tracy) We have a build of 1-0-5, which supports running the diagnostic filter. We still need to add its outputs to our stuff, e.g. ntuple variables, and we need a way to control it from job options. In order to integrate the new build on Windows he had to cut out various connections to statistics-gathering code.
As for an anomaly concerning the 20 GeV bit in our simulations: the bit is getting set but, unlike real onboard running, we don't stop evaluating at that point. If subsequently a veto is seen the event is thrown away. This behavior needs to be changed.
SCons: (Navid) Yesterday he committed SConscript files for practically everything in ScienceTools. Still missing are one container package and one new package, created since he started in on this project. (Jim) Looked at the one for Likelihood. It seems these files were written based on an older release of the code and need some updating. (Navid) Be my guest. They really should be maintained by package owners. Next step is to make symbolic links in the CVS repository.
(Heather) As has been discussed before, we'll need a new naming convention for tags. This is further complicated by the fact that some packages are in both ScienceTools and GlastRelease. (Navid) suggests that RM could make a new-style tag when it sees an old-style one go by.
AOB: (Heather) Julie mentioned that we need to support real standalone running not requiring access to the SLAC central calibration database. (Joanne) It is possible to do this by running a MySQL server locally as well as keeping copies of calibrations locally, but it requires some work to set up. See these instructions.

minutes index

J. Bogart, Last Modified: 01-Jun-2010 15:45:44 -0700