Present: Brian Baughman, Joanne Bogart, Toby Burnett, James Chiang, Johann Cohen-Tanugi, Richard Dubois, Berrie Giebels, Navid Golpayegani, Traudl Hansl-Kozanecka, Heather Kelly, Michael Kuss, Matt Langston, Julie McEnery, Leon Rochester, Alex Schlessinger, Tracy Usher
DC1 status:
(Toby) This run of background events and another of all gammas demonstrate that we are not yet replicating Bill's work in practice DC1 runs. See the section starting with TriggerAlg output, about 40% of the way through the file. Of 633 triggers, 46 survived the Onboard Filter, of which 3 were accepted after analysis cuts. However in run of all gammas only 1 event made it past analysis cuts, which doesn't make sense.
Plea to algorithm writers: output a meaningful sentence or two in finalize. Even just a count of events processed would be much better than "finalized called".
Another show stopper at the moment is a crash in AnalysisNtuple under some circumstances. Toby and Leon hope to find and fix this today [as of 10 AM PDT, looks like they may have already], but even so GlastRelease v3r3p3 is not going to be the final word. It will contain the latest XML geometry description, which adjusts the vertical position of the CAL slightly.
Johann is working on creating the level 1 FITS file.To start he's making a root file containing the relevant data, with the algorithm in RootIo, but this is not its final resting place: it will move to merit.
(Heather) She will start over with simulation+OnboardFilter runs (5 million AllGamma, 50 million Background) when v3r3p3 comes out. She and Toby will discuss post-meeting how to rationalize the creation of Job Options files for the individual jobs from a template. [done.]
EM software upgrade: (Heather) New versions of ebfReader and EbfConverter which can handle CAL data, using new EBF library release, are tagged and believed to be fine. No complaints from users so far.
(Toby) Perhaps we should institute another package, like the DataChallenge package, as a place fto keep job options suitable for EM data processing.
Windows Release Manager: (Alex) He is about done with the stuff needed to invoke Navid's Soap service. The (web) output currently is parallel to the Linux Release Manager output. Next step is to provide something supporting easy comparison between the two platforms.
Opus (Alex) Nothing for this week; plan is for some test runs early next week.
gcc3.2 (Alex) He has built a complete set of external libraries with gcc3.2. There was a problem with Xerces that necessitated going to version 2.2 (our standard is still back at 1.7). This entailed an extra change to XMLEXT for different include path, but that was all.
Changes are in the works to support multiple CMTCONFIG values gracefully, so that users may investigate different configurations.
There are enough external libraries that it is getting to be quite a burden to maintain them all for different configurations. Alex requests that whoever is responsible for such a package provide a straightforward method of building it on both platforms, or complete and accurate documentation [better yet, both. ed]
Xerces notes (Joanne) Xerces 2.2 is substantially different from 1.7 internally and the routines we're using are deprecated and undocumented in 2.2, but Alex's experience indicates this is a change we could make if we have to without too much disruption. The very latest Xerces release, 2.3, is reputed to have some problems building on Windows.
System tests: (Matt) See some sample plots (1 GeV run and 100MeV run) from the new system, comparing CAL cuts of 100 microns and 700 microns. A quick look turns up no obvious differences in the plots, but subsystem experts need to examine their favorite plots more closely. Matt will dig up timing statistics for the two configurations. A newer version with nicer output runs on Matt's laptop; he is awaiting SCS's blessing so that the same code can be run by the Windows server.
Three of the tests are failing (thanks to Julie for help in understanding this). One, gamma 10 GeV, is just timing out. It could be split into several runs of fewer events. ACDTop and ACDDigi fail because they have not kept pace with ACD code changes. Matt will work with Heather on bringing them back up to speed.
Hearing no objections, Matt will remove the legacy CVS systests subdirectory of Gleam/src. The code under active development is in its own systests package, where it more logically belongs.
G4: (Richard, Toby) G4 electromagnetic physics guys have been visiting at SLAC. Richard and Tracy spoke with them last week. They are at least somewhat responsive to our problems and concerns. They will integrate our tests into their standard ones and will give us access to "Developer Releases" which come out once a month, but not to their CVS repository. We will try for a VRVS meeting tomorrow AM, ideally including Francesco and Riccardo.
Toby reported on the G4 meeting he attended in Vancouver a while back, in which he gave one of the User talks about our not entirely happy experiences. He was the only one to publicly voice any complaints, and there was essentially no response to what he had to say. Someone else at the meeting gave a talk pointing out large differences in behavior of muon decay between G3 and G4 which also met with no response. In a separate discussion, Toby's proposal that the project move to an open source-like model of development (e.g., allow at least read access to the project's CVS repository to anyone) was not met with open arms.
Travel: There will be a lot in the next month or so, quite a bit of it vacation after the Collaboration meeting.. Check the travel schedule for details. Looking ahead to mid-October, Heather will be making the first of what should be a series of monthly visits to SLAC. Navid will also come to SLAC for about a week at that time.
previous | minutes index | next |
J. Bogart Last Modified: 04-Aug-2004 15:41:53 -0700