Core Minutes 6/26/2007
Present: Joanne Bogart, Anders Borgland, Toby Burnett, David Chamont,
Jim Chiang, Richard Dubois, Warren Focke,
Tom Glanzman,
Heather Kelly,
Michael Kuss,
Julie McEnery, Chuck Patterson, Igor Pavlin,
Leon Rochester, Tracy Usher, Eric Winter
- Absences: Navid is out this week (getting
certified for diving). Toby will be away for 2 weeks, starting right
after this meeting. See
the SLAC Glast vacation calendar for more
upcoming absences.
- Service Challenge issues:
- Zero energy events (Richard, Toby) Five background events
with zero energy triggered out of about 200k total. There is not enough
information in the tuple to figure out why; someone needs to look at
the digis. (Toby) Is there a reconstructed track? (Richard) No.
(Leon) There is about a 2*10**(-5) probability of tracker
noise hits, so typically only a very few per tower in a single event.
This is unlikely to account for five 6-in-a-row events.
(Richard) Since the number of aberrant events
is so small, they could just be discarded.
[See Leon's
Thursday presentation
for more on this topic.]
- Crash in termination (Richard) Our batch jobs now tend to run on
64-bit machines in compatibility mode (it's possible with some effort to
force them to run on 32-bit machines); the resulting behavior is not quite
identical to that on a 32-bit machine. In particular, double-delete of
memory is flagged as an error.
Michael has suggested this can be avoided by setting
a certain environment variable to a particular value. We'll try this.
(Heather) Do we know where the double-delete is coming from? Is it the
old CLHEP/Gaudi one or something new?
- Source id (Richard) Interleave events have one million added
to source id. We have decided to use the DC2 scheme of negating source
ids when background events are interleaved, but only values in the merit
tuple have been updated this way; source ids appearing elsewhere in the
data have not been changed.
- Spacecraft pointing angle (Richard, Toby) Julie reported
at last week's C & A meeting that she saw huge excursions in
the spacecraft pointing angle. Toby confirmed that the quaternion
interpolation code (interpolating from values taken at 30-second intervals)
appears to be making on average one bad calculation per orbit. For most
of the orbit, the 30-second sampling interval provides more information than is
needed for a good interpolation, but during slew maneuvers, when the
angle is changing rapidly, it may not be providing enough. In the short term
Toby will replace the existing interpolation scheme with a simple linear
interpolation.
- Albedo events (Tom) They need to be binned.
How many more do we need?
- GR: Heather
will be managing GlastRelease releases during Toby's absence. In
fact, Toby recommends that Heather become the new GR Head Honcho permanently.
The last GR tag is broken for Windows because of a mix-up concerning
the new G4 shareable library. Tracy has fixed this with a new tag of
Iexternal/Geant4Runtime. A new GR release including
this tag is imminent. (Toby, Heather) This will probably be the last tag in the
v11 series. Many upgraded packages are waiting in the wings; those
which are well-behaved should
be incorporated into an early v12 release.
- EM: (Heather) There is a new EM release
based on v11r9. There is an outstanding issue with job options, but more
worrisome is Anders' experience with memory leaks. When processing
real data he sees a leak in Recon of about 250K/event. There is no such
behavior when the same code is used on MC data. Heather will be poring over
ValGrind output kindly produced by Anders.
- MRStudio update: (Joanne)
Since there have been problems in the past with library incompatibilities
on Linux, a bit over a week ago
Riccardo packaged up sources for needed graphics libraries and a script to
build them with. With some adjustments I was able to build usable
libraries on a noric (RHEL3) but not on either of two RHEL4 systems.
Using the noric version I found some bugs and reported them to Riccardo.
He was able to reproduce them and also found others...work is in progress.
- CMT Action Committee: (Heather)
NPMS has been withdrawn from consideration. If we move to anything,
it will be SCons-Enhanced — enhanced to handle package management
functions in a manner similar to what we have now. When he returns Navid
will be looking into what's needed for RM.
- Core Face-to-face meeting: (Richard)
It's been a while; it would be useful. Proposed date is second week of
August. Who is/isn't available then?
- AOB:
- (Richard) Note new support email
address for EVO: evosupport@vrvs.org.
- (Jim) Is there any reason not to go ahead and move to Python 2.5?
[Apparently not.]
J. Bogart, Last Modified:
01-Jun-2010 15:45:00 -0700