Core Minutes 11/30/2004
Present: Joanne Bogart, Toby Burnett,
David Chamont, James Chiang, Johann Cohen-Tanugi, Seth Digel, Richard Dubois,
Warren Focke, Navid Golpayegani, Heather Kelly, Matt Langston,
Francesco Longo, Julie McEnery, Chuck Patterson, Leon Rochester,
Tom Stephens, Tracy Usher
-
RM news Thanks to a timely patch by Navid
LATEST builds will now accept tags similar to v1r7p3gr1. This has eliminated
most but not all compile errors we've been seeing in LATEST.
Navid has determined that it would take considerable effort to reinstitute
automatic optimized builds in the old RM. Instead, he will make them
manually for tagged releases (he recently made one for GlastRelease v4r6),
meanwhile working on the new RM. He knows of no other pending requests for
old RM.
-
System tests (Julie) She will be running
tests on v4r7. In an attempt run them on v4r2 (at the time the latest version
with an optimized build) she saw the memory usage go up to 2 Gig, followed by a crash in TkrDigiAlg.
- GlastRelease, G4 and TkrRecon:
Richard proposes we make a first v5 GlastRelease tag including new G4
but not new TkrRecon, to be followed shortly thereafter by v5rX including
the new TkrRecon. Francesco has been comparing histograms from old and
new G4.
So far, so good. Look for a final report at next week's Analysis Group
meeting.
- Nightmare
on Root Street: (© Heather) Alas, although the build of Root v4.01.04 (version recommended
for Windows by Root forum) kindly supplied by Navid works fine on Windows,
test_rootIo produces the same old core dump on Linux seen with 4.00.08, but
not with
4.00.08F. It seems to have something to do with our use of Gaudi.
Still searching for the magic version that works everywhere.
- Pipeline:
Planning on an MC run this week.
- Matt needs to make some fixes to front-end:
eliminate restriction on number of runs displayed; make links to log files.
Goal is to have everything ready by tomorrow.
- Dan is putting in his LSF log file fix; it was more complicated than
expected.
- Still need a solution to the limit (200) on concurrent Oracle connections.
Haven't been able to get around it with the program Alex found to do
database pooling, nor with a Java tool which was supposed to handle it.
- Doubles or floats
(Leon) Using all that space in our ROOT files for doubles rather than
floats seems pointless. He proposes we make the code changes to write
out floats instead. (Richard) This is to save space in files. What
about TDS representation? We could leave that alone. (Leon) We might
want to use floats in the TDS as well for reproducibility, so that
event data is the same whether it's generated in your process or read in.
Toby is also in favor of switching to floats. Discussion to be
continued next week.
Coming Soon
Navid, Heather and David will all be visiting SLAC next week, providing
good opportunities for discussion of CalRecon and RM issues.
J. Bogart Last Modified:
01-Jun-2010 15:46:32 -0700