Core Minutes 7/12/2005
Present: Joanne Bogart, Philippe Bruel, Toby Burnett,
David Chamont, Dan Flath,
Richard Dubois, Warren Focke, Riccardo Giannitrapani, Berrie Giebels,
Tom Glanzman, Navid Golpayegani,
Tony Johnson, Heather Kelly, Matt Langston,
Igor Pavlin, James Peachey, Leon Rochester, Robert Schaeffer,
Tom Stephens, Tracy Usher
Topics concerning DC2 readiness came first:
- Data handling: (Tony)
- Java CVS repository Some projects
have been moved there.
Cruise
Control, which does automatic builds upon CVS commit, is enabled
for these projects.
- Data access
The web portal talks to the Pruner and Peeler as well as the
(Pipeline?) web server. Working on access to the metadata catalog,
including the ability to add entries to this catalog (for example,
for DC2 data generated remotely). (Richard) At least need a way
to switch to the test database for Pruner and Peeler.
- Indexing data on a sphere Jean Paul is looking into using
HTM or
HEALPix.
(Toby) believes HEALPix is superior; we're already using it elsewhere in
our software. He would like to be kept in the loop.
- Peeler status (Tom) It's in test now. (Richard)
Typically the reason people want to pick out individual events with the
Peeler is to
bring them up in the Event Display, but the ROOT tree doesn't contain
all the information one might want to display (in particular, all the
MC information, too much to save for production running). 2-stage event
generation could take care of this. The first stage—just the MC source
particles—could be imported by the user, who would then run Recon
locally to reproduce the full event.
- Pipeline (Richard) have been doing some stress testing to see
how it behaves with a large number of concurrent runs. It (the stress
testing) did its job: some problem areas have been identified.
- Flaky nfs. After n successful accesses by n jobs of a particular
file,
it can happen that the the n+1st will fail.
- Errors in user script
- May be some difficulties to be ironed out with Navid's batch
interface.
- CalRecon: (Tracy) It's already improved
since Bill showed results at Goddard last week. Main thing we're waiting
for now is refinements in MIP finder, expected in a day or two. We should
have a stable release no later than Friday. Need to make sure that
n-tuple doc is up to date. (Toby) Can we fix
use statements in requirements file? (Tracy) Already in progress.
- GR HEAD: (Toby) Since Julie can now
readily run System Tests on HEAD builds and there hasn't been one
for a while, he made one. RootConvert failed
to compile on Windows though it had compiled successfully on Linux [albeit with
warnings. ed.]. The complaint was a serious one: a utility called
by the package should have been declared
const
but hadn't
been. Should we be thinking about moving to a newer version of gcc?
(Richard) Gaudi doesn't compile under newer versions of gcc. (Toby) but
perhaps it doesn't need to, since we just use it as an external library.
- Dead channel list (Leon) He has a suitable list
for DC2. He will give it a special flavor to distinguish it from "normal"
calibrations. He will deal with partially-dead strips
by picking a subset to be always dead.
- RM:
RM is up and running again on Windows.
- New Windows hardware: (Matt) Machines
were not configured quite as we had asked—need to get some more
things off the CD and so forth—he expects to get this done
today.
- Common lib directories: (Navid)
He made some changes to GlastPolicy to create a lib directory per item
in CMT path, now on Windows as well as Linux. He will commit it so
alpha users can try it out. Once it is tagged, RM will be using it
as well. Then Installer builds will also have the lib directories.
(Richard) Since neither developers nor end-users typically want to
rebuild packages from an Installer build, how about cleaning up intermediate
files before packaging up into tar.gz or .zip and making the whole thing
read-only? (Navid) Shouldn't be difficult.
- MRvcmt: (Riccardo) He has changed it to use
CMT setup files directly. Also the ability to set CMTconfig from
MRvcmt on Windows has been disabled since it's unnecessary. The correct
value can always be set automatically. He's working on a way to use
Toby's script to clean up intermediate files. This would be an option
only, not something that always happens.
- FRED: (Riccardo) He will close
FRED-40.
The new HepRepCorba tag appears to have fixed
it (too bad we don't know exactly why). He will investigate moving to
a newer version of OmniOrb.
J. Bogart, Last Modified:
01-Jun-2010 15:46:11 -0700