Core Minutes 9/25/2007
Present: Joanne Bogart, Toby Burnett
David Chamont,
Warren Focke,
Tom Glanzman,
Navid Golpayegani, Heather Kelly,
David Landriu,
Bryson Lee, Chuck Patterson, Igor Pavlin,
Leon Rochester, Tracy Usher
- GR news:
- Memory leak
(Heather) Toby reported a memory leak with v12r5. With Tracy's help,
discovered evidence of the leak as far back as v11r10. It's a Windows
only problem. Conceivably it's connected with the G4 extlib, but we
don't know that yet.
- EbfWriter and GEM sequence numbers In simulation we run
EbfWriter twice. Left to itself, it assigns
GEM sequence numbers, which is fine the first time around, but, as
Anders has pointed out, inappropriate
for the second call.
- GR v12r8 includes a patch to EbfWriter
for the above situation
and new versions of AdcRecon and
CRFlux.
- Composite event lists (David Chamont)
The code compiles and runs but is still somewhat unstable;
he is now concentrating on a test suite.
- mootCore (Joanne) belatedly realized that the LATC configurations
referenced by OktoberTest data will be 'old style' (that is, built from
relatively few source files, approximately by LATC component type, rather
than from many source files respecting Precinct (application-oriented)
boundaries as well as LATC component boundaries). Most of the work to be
done in
CalibSvc had to do with supporting this finer
division, so is irrelevant to this particular exercise which is just
to make the GEM configuration available to Martin's code. For now he
can get to what he needs
by calling mootCore services which are already
in GR LATEST.
- Runs
- For OktoberTest (Tom) Over the weekend as an exercise
did a first orbit run with v12r4 [correction: it was v12r7]; will be redone with v12r8. Meanwhile
Francesco/Johan/Nicola in Italy has been getting ready for orbit 2.
For this ROOT output will
be disabled; we just want the output of EbfWriter
(before and after the OBF). [Update since meeting: orbit 1 is currently
under way with v12r8, and the first successful test of orbit 2 has completed]
- Finalizing L&EO Interleave (Tom) There had been two
sorts of problem runs: some ending in core dumps, others asking for
albedo background events
at spacecraft zenith angles greater than 90 degrees,
which were not generated in the original background runs.
The first sort of problem turned out to be transient failures that went
away when the jobs were rerun with new initial random number sequence numbers.
The second was fixed by generating new albedo backgrounds, populating
the desired zenith ranges.
Now all 7920 jobs complete
successfully. Seth tentatively says the backgrounds look ok.
[Thanks to Tom for filling in details I didn't catch the first time around.
ed.]
- UW running (Toby) By cutting down run size to 2500 events he
was able to do an all-gamma run using v12r4 which avoided serious effects
of the memory leak; however some of the tuple variables are bad. (Heather)
There were problems with some variables which are fixed by the new
tag of AcdRecon. (Toby) will try again with v12r8.
- Environment variables (Navid) The latest
facilities tag has a new class,
commonUtilities, allowing environment variables to be better
managed. See the Doxygen documentation in recent builds. Applications should switch
to using
the new routines tailored to specific situations (
getDataPath
,
getXmlPath
etc.) where appropriate.
- Travel: Tracy leaves tomorrow night.
Richard will be returning next Tuesday.
J. Bogart, Last Modified:
01-Jun-2010 15:46:51 -0700