Core Minutes 11/16/2004
Present: Joanne Bogart, Toby Burnett, David Chamont,
James Chiang, Johann Cohen-Tanugi, Seth Digel, Richard Dubois, Zach Fewtrell,
Warren Focke, Riccardo Giannitrapani, Navid
Golpayegani, Theodore Hierath, Heather Kelly, Francesco Longo, Julie McEnery, Chuck Patterson, Leon
Rochester, Tracy Usher
-
Farewell, Alex: (Richard) Alex has decided
to move on to other things, effective at the end of this month. Navid will be
picking up his RM work and acting as Sys Admin for servers for the time being.
Richard has been assured we can hire a replacement; a job req. is in the
works.
-
G4 move? (Tracy) G4 6.2 has required some
changes to G4Generator and G4Propagator, currently committed but not tagged.
This also affects TkrRecon. Meanwhile, other changes in Trackerland have
been proceeding. None of the new stuff can go in until we go ahead with 6.2,
or decide to back out. We've already branched for EM; managing yet
another branch is a most unattractive prospect. (Richard) Do we want
6.2? Is it stable? Have heard rumors of crash upon exit. What
about validation? (Francesco) The crashes may be due to an improperly built
installation. There are improvements in hadronic interaction models in newer
versions of G4. The Geant collaboration is about to come out with version 7,
so if we upgrade to 6.2 we'll still be behind the curve. Could we instead wait
for 7? (Toby) ..has heard that version 7 requires a newer version of
CLHEP which uses namespaces, hence would require widespread coding changes on
our side, whereas the changes we need for 6.2 are restricted to a small set of
packages.
(Richard) Proposal is that we move to 6.2, subject to approval at next
Analysis Meeting. (Francesco)..confirms it is straightforward to run his
validation suite. (Toby) would like to tag now, in anticipation of approval.
(Julie) suggests we first make a GlastRelease tag including at least a couple
very desirable bug fixes: one to fix a Tracker crash, another to fix a memory
leak in OnboardFilter. (Leon) There is one other bug fix, currently available
only in Anders' development area, which should maybe also go into this
release.
- System tests:
(Julie) See her
presentation for a concise description of the architecture and status.
Problems surfaced when we moved from Redhat 7.2 to 9 which were addressed by
making database connections serially rather than simultaneously. There has
been a stream of lesser bug fixes, perhaps nearly at an end, needed before the
procedure can be automated and run from the Release Manager.
We need an improved front-end before the effort needed to run Systests on
pipeline data (changing focus somewhat to validating data rather than
validating a code release) would pay off. Do we want to devote resources to
this?
Next project is to implement systests for head releases. Do we
want to compare two most recent heads, or most recent head to most recent tag?
Overall mood tended towards the latter.
- External data and interface package:
(Theodore) He presented a
proposal to create a new interface package for external data, analogous to
interface packages for external code. There are examples of such data, most
notably the JPL Ephemeris file, which are large, very rarely change, and may
be used by more than one standard code package, similar to external libraries.
Since standard conventions will be followed for the interface package, support
from the Installer should be more or less automatic. [Well, maybe. Navid
will look into it.]
There was general agreement that this is a good direction. (Francesco)
Would be useful for G4 test files. (James) It's probably not appropriate for
test data, which tends to change with code changes, but is a good idea for
user data. Consider subdividing into ST-only data, GR-only data, and
Common, similar to the way external libraries are handled.
- ROOT upgrade: (Heather)
Work is in
progress on moving us to ROOT 4.00.08. Earlier stumbling block was removed
after Navid provided her with a build of 4.00.08F. Still need to do a bit more
testing.
- Xerces upgrade: (Joanne) Went in last
week with no apparent problems. Waiting for resolution of other LATEST build
issues (such as G4 move) to be absolutely sure everything is OK.
- RM tidbits: (Richard) Navid has
reinstated email notification of failures and expects to get optimized builds
running again soon. Request is in for one last small enhancement to the old
system: a script to do a particular build or clean up a particular build on
request.
- Pipeline: (Richard) In the wee hours
this morning Dan committed his work-around for the batch job completion
problem which has been interfering with pipeline operation. No fix has been
forthcoming to date from the other side (SCS). The appearance of the problem a
couple weeks ago more or less coincides with the use of new file servers;
perhaps incorrect configuration is responsible.
- J2EE front: (Richard) Matt is chasing
a last problem with XML file upload, specific to the Oracle database.
- ScienceTools: (James) Not much to
report. (Julie) Does it get tested regularly on Windows? (Toby) It all builds,
but he hasn't looked at it in detail for a while. (Jim) regularly does test on
Windows..yes, it's ok.
∫ Flight(w, ..) dw
- EM status:
(Heather) She just fixed a
problem in RootIo
detected by Anders. The Cal UDF (U is for User) contribution will be
available in LDF files. Also it will be available as a string in the ROOT
header David Chamont is working on. (Richard) Can this be extended to
other subsystems? (Heather) More or less, yes.
Coming Soon
J. Bogart Last Modified:
01-Jun-2010 15:46:14 -0700