ScienceTools: Jim presented
the
report.
FSSC Report: (Eric W.) Some python/ROOT
problems are being resolved. Eventually will be distributing most of the
ScienceTools python interface.
Recent network problems at Goddard were not detected by the
automated checking since it only pings the main server, which was not
affected. (Eric W.) It's not possible to ping most Goddard nodes because of
security restrictions. He suggests periodically initiating an ssh session
to a Fermi machine as a check on connectivity.
Pass7 reprocessing:
(Richard) It will take the rest of this week for Bill to finish up.
Other issues:
- How to deal with changes in number of merit columns. Proper procedure
is understood.
- Getting FT1 event class in shape. (Leon) Several issues still need
to be discussed.
- (Tom) Removal of obsolete merit columns.
- GR crash which happens with the new ROOT (5.20). (Heather) It's
specific to getting data from xrootd. And the problem can be reproduced
independent of Gleam, just by using interactive ROOT. Wilko has
relevant materials and will look into it.
(Warren) FT2 is ready to go.
Tapes: (Richard)
The Computer Center at slac is putting up a new tape library. We will
be among the first to make use of it. Tape is cheap! The price has gone
down by a factor of four. We might consider changing our strategy to
archive to tape rather than delete old recon files. (Disk is still too
expensive to consider keeping everything there.)
Pass 8: (Tracy) See
notes from a meeting with Robert Johnson, Bill, Leon. One decision taken
was to get together regularly, perhaps every couple weeks.
Documentation: Chuck provided this
report:
I'm currently fine tuning the Help files for the pulsar analysis
tutorials, and will be cutting another tag for the workbook so Berrie
can update the mirror site (again). Later today, I plan to update the
obssim/orbsim tutorials and reference files, and then I'll be resuming
work on pyLikelihood.
Core talk:
- GR v17r18 (Heather) has the rebuild of G4 with the new compile
options detailed last week. Johann will redo runs in Lyon which were
crashing with previous RHEL4 G4 build.
- ROOT build (Heather) Pat has pointed out a 1-line bug fix to
Minuit2, needed to avoid problems in the optimizers
package. It will be in future ROOT versions, but we might not want to wait
for that. Instead, she plans to make a new local version of 5.20 incorporating
the patch.
- RHEL3 (Richard) Are we ready to abandon it? (Heather) We have
to try out the new G4 on the runs that failed. (Jim) Maria Elena had some
problems running jobs at SLAC with the RHEL4 build. (Heather) would like
to know the details. (Richard) believes the Computer Center will be
converting the RHEL3 machines over the summer [so we had better get
cracking. ed.]
- ACD (Heather) Eric Charles has a list of tags for the new
ACD recon which he would like to be incorporated along a branch. The
ACD updates are not going to make it into this round of reprocessing.
- skimmer, merit (Heather) Needed to massage it a bit to get it to
handle the new merit header. There are some repercussions. She would
like to add a flag to the metadata to indicate whether or not the
data is a simple tree.
In order to handle merit files with different numbers of columns in the
same job one needs to skim in "slow" mode. Tony J. fixed the web front-end
use use checksums to discover if the number of columns differs.
- SCons RM (Navid) The Computer Center has not yet found a way to get
an afs token to macs for batch jobs, so RM for macs has to stay off.
He's been working with James Peachey on ScienceTools unit tests.
An additional environment variable needs to be defined. He would like a
general way to handle this kind of thing, probably an expanded version
of the scheme Bryson put into place for CHS.
Speaking of CHS, (Navid) has had no further input from concerned
parties about fixing the packages which RM does not currently build.
If need be, he can do it himself; it should be straightforward.
- SCons and Windows (Joanne) The one missing piece in the
project file/solution file arena is handling for rootcint. A nano-GR
consisting only of
commonRootData makes a good testbed since
it requires no extra external libraries. Heather and Navid quickly
unraveled the first problem she ran into: missing compile-time define,
not needed before ROOT 5.20). There is still a lot to do, however.
- SCons, GoGui and RHEL5 (Joanne) has a new laptop running
RHEL5 64-bit. Many things work — no problems building GoGui, for
example — but 64-bit python is not one of them: a module required
by SCons works correctly only for 32-bit. Navid came up with a work-around,
which is to use a 32-bit build of python (and all other externals) and
add an option to SCons to allow users to explicitly ask for 32-bit or 64-bit
builds. It's also been added to the GoGui interface.
- Fred and FOX (Emmanuel) Tried rebuilding FXRUBY as well as
Fred and FOX but still has the same problem. He's not positive the new
FXRUBY is the one being used, though.
- ScienceTools build (Emmanuel) is working on getting a recent tag of
ScienceTools, LATEST-1-2602, to build under SCons. He says:
Over a week ago, I discovered that scons could not find
celestialSourcesLib.py. I checked and it is clearly there so I began
debugging the SConstruct file to find out why SCons cannot find that file.
I identified the problematic code in the file SConstruct. After providing
a fix, I was able to build celestialSources and
other packages. Thus far I have built 12 out of 39 packages. My progress
is documented
in Confluence
J. Bogart, Last Modified:
01-Jun-2010 15:45:29 -0700