Core Minutes 10/26/2010

• ScienceTools: (Jim)

  • Johann made a new Likelihood tag to fix a problem with DMFIT. For more detail see Science Tools Development Notes in Confluence.
  • Jim made a pyLikelihood tag to fix a bug when plotting with a BinnedAnalysis object.

• Reprocessing: (Tom G.) Pass 6 reprocessing is complete through September. (Tom S.) A few files were transferred by hand last night and minor glitches were addressed. We're now ready to do the rest en masse. More validation will be happening at FSSC. (Tom G.) Last time we had a massive amount of reprocessing data to transfer it took a couple weeks; this time we have about twice as much. We should also see about getting the data on the astroserver. (Richard) What about data generated after September but before L1 switches over? (Tom) Expects the gap will not be terribly long so, unless there is an urgent need, will wait to reprocess the remainder in a single batch after the switchover is done.

• FSSC: (Eric W.) v9r18p6 has been tweaked to incorporate a bug fix to tip for its test program. Will start testing tomorrow.

• Documentation: (Chuck) sends the following:

  The newsletter is priority #1 for this week and is due to hit the streets on Friday. Last week, I posted major revisions to both the End-User and Instrument Analysis sections of the WB to the web, then committed all of the pages to CVS, created a new tag, and notified Berrie so he could update the mirror site. Next week, I plan to begin updating the Science Tools section of the WB.

• Pass8: (Tracy) knows of no progress on the problem with Johan's new clustering code (inability to write out the data).

  (Tracy) The long processing times with the new track finding are similar in nature to problems we've seen in the past: there are certain events, basically junk, which can take a very long time to process. Would make sense to put our efforts into recognizing and discarding these events rather than trying to reconstruct them.

  (Leon) is working on 2-pass pattern recognition: cosmic ray followed by normal. He's aiming on getting a first version into CVS today; then people like Bill and Robert will be able to exercise it. By default just the normal pattern recognition will be done. Job options allow you to ask for the extra cosmic ray step.

• More reprocessing: (Tom G.)

  • Pass 7 reprocessing is on hold, awaiting validation.
  • Missin GCR files: Gleam seems to have a memory leak, preventing runs larger than about 400 - 600k. Tom has contacted Warren about the possibility of breaking the runs into fragments. (Heather) The memory leak occurs in recon read but only in the context of Gleam. She has stripped out everything else in Gleam — still leaky. However, if she does the very same read operations apart from Gleam, all is well.

• Hardware: (Richard) Oracle relented and is now offering us our old 60% discount on servers. We'll be purchasing 5, all we can afford. Since we will have to return the space we've been borrowing, that leaves us with a net increase of about 200 Tbytes. The bad news is that these servers will be discontinued as of Nov. 19 and we won't have more money to spend on this by then. For the next round we'll have to look for other vendors.

• Core talk:

  • cfitsio (Heather) made a branch tag of ScienceTools using the new cfitsio so that Tom G. could try it out. (Tom) is almost done with testing.

  • ldfReader and Beamtest (Heather) ..issue still pending.

  • obf 64-bit support (Heather) Jana responded to her request for status update. There is progress and rumor has it that the new distribution, when there is one, will be significantly cleaner and smaller than what we have now. Will try to schedule an EVO or other get-together to find out more in November.

  • New windows box (Heather) is being loaded up with software. We're hoping VS 2003 can be included somehow, perhaps via a virtual machine.

  • Snow Leopard (Tom S.) He is able to build Qt (shared libraries only, however; not static) and Qt-based RM programs. Using these programs he was able to make a build successfully in all respects except one minor one: the stop time was not entered in the database. However it is not yet working properly in the context of RM.

  • G4 for Windows (Heather) While trying to get as much of GR going on Windows (VS 2008) as possible, Joanne quickly discovered we had no build at all of G4 for this platform and the old build we had of CLHEP was 1.9.4.2, which is not compatible with our version of G4. She managed to find sources for CLHEP 1.9.2.3 (almost certainly close enough to the version we're using for GR on other platforms, 1.9.2.2) and build it and has now also built G4 (static libraries only, which is what we have for VS 2003 as well).

  • SCons,GR and new gaudi (Joanne) has committed various changes, mostly to SConscripts, to the new-Gaudi branch of GlastRelease. GlastSvc builds and its test program runs. To get much further, CLHEP and G4 are necessary (see above). And ultimately we need a VS 2008 build of obf (see below)

    Problem with time classes in test_CalibSvc has been found and fixed. The program now runs correctly.

  • obf for VS 2008 (Tracy) requires a certain amount of hand-editing, then presumably can be built without too much difficulty using SCons. (Joanne) If we can use the same source as for VS 2003 the hand-editing step is already done. (Heather) suggests we try taking VS 2003 project files and see if the newer version of VS will successfully convert them. (Tracy) That may well be the quickest and simplest method.

• Announcements

  • SCons meeting tomorrow (Wednesday) at 10:30
  • Richard, Heather, Tom S. and Joanne will be attending ADASS in Boston, Nov. 7-11

previous

minutes index

next

J. Bogart, Last Modified: