Core Minutes 7/24/2007

• 55-day run, GlastRelease:
  • GR v11r16 GR v11r16 is ready to go. (Toby) It incorporates changes in astro and FluxSvc fixing a problem concerning orientation of GLAST. There is some awkwardness connected with it being done on a branch; in particular, there is no corresponding RM HEAD build. (Heather) We should consider this a learning experience and an opportunity to think about what RM should do in such circumstances. If it becomes too much of a problem, we can back out the stuff now in HEAD, destined for GR v12 (updates to ACD packages supporting new side tile and side ribbon geometry, which seem to be ok).
  • Overriding packages (Toby et. al.) Before the new GR release incorporating Toby's fixes was available Tom tried to run with the previous GR release + new versions of a small number of packages. Although Tom followed procedures to the letter and there is evidence that he picked up the new packages, symptoms of the old incorrect code were also present. These symptoms don't show up when the new GR release is used. It would be good to understand what went wrong.
  • Pulsar package (Jim) suggests this package, which has been the source of several problems, could use more oversight. Can some additional person with the necessary expertise take this on? (Richard) Nicola would be a good candidate.
  • GlastClassify (Richard for Tracy, who is embroiled in housing issues) Tracy's latest reverse engineering has been validated by Bill, but Tracy hasn't had a chance yet to commit to CVS.
• Beamtest crash:  (Heather) Tracy has been looking into a reported BeamtestRelease crash on Linux in the G4Generator::BeamTransform routine. He has traced this the need to update McParticle in the context of Beamtest. The Event::McParticle class only provides an initialize routine to set its data members. A side-effect of using McParticle::init is that an unnecessary daughter McParticle is created which must be removed. So in BeamTransform, a reverse_iterator was being used to search out this particle to remove it, which is the location of the crash. Leon has since updated the Event package to provide proper set routines to avoid the use of McParticle::init, which would avoid the necessity of removing the McParticle and probably take care of the crash. It is necessary to check with Tracy to see if this takes care of the problem. (Michael) There was a seg fault, now understood and fixed as described, but there is also a memory leak, not yet tracked down. (Leon) The propagator crashes under some circumstances when called outside TkrRecon because there is no exception handler. In some cases, depending on the exception, crashing may be the right response.
• New EM release: (Heather) Misidentification of thick and thin tiles has been tracked down to a bug in AnalysisNtuple, now fixed and used by the latest EM release tag, v8r1109p12.
• Where to run: (Richard) Anders would like to use faster machines for System Tests. (Leon) Currently they run on toris. (Richard) The fastest machines are boers; they're faster than toris by a factor of 2.5. Could just move System Tests there. Or could make use of a recent cross-calibration of batch machines.
• TriRowBits algorithm: (Heather, from her notes)

  There is a JIRA GRINF-34 which pertains to missing data associated with the TriRowBits calculation that we have utilized in EM. It was found that as of June 2006, this code now requires that the symbol TRIROWBITS be defined, by default it is not. TriggerAlg has now been branched for EM, so that this symbol is defined and the TriRowBits code is again included in the compilation. Toby desires that this code be moved to its own Gaudi algorithm, but thought that this code was only meant for a special study and was no longer required. So we need two questions answered:

  1. Do we expect to continue to use TriRowBits in the future?
  2. Who will move the code to its own algorithm if we do expect to continue to use it?

• Event collection update: (Heather, from her notes)

  RootIo is being updated to use the new metaRootData package from Eric Charles. David Chamont has already updated RootIo to centralize the ROOT reading. Heather has now done something similar for the writing. It is necessary to provide central access to the ROOT Trees so that the meta ROOT file can be generated when Gleam writes out its ROOT files. A new tag (or two) of RootIo will be completed by Thursday.

• Citrix: (Heather for Rodney) Our Citrix server has been updated; no adverse consequences are expected.
• SCons update: (Navid) See the SCons Demo Confluence page for complete status information. Recent progress includes
  • a new option, with-GLAST-EXT, making it possible for the user to specify a (default) location for external libraries on the command line. It is still possible to override for individual libraries.
  • automatic computation of indirect dependencies. The package maintainer will only need to list packages his/her package uses explicitly in a special python file; SCons will do the rest.
  See also Navid's updated To-do list at the bottom of the page.

  (Heather) Your work to date has been on Linux. Can we expect that, at the very least, it will be straightforward to use this facility to build on Windows via nmake? (Navid) Yes, that should be possible. (Heather) How about an ETA for a demo? (Navid) That's hard to predict but he's hoping to be ready for a demo by the time of the Core face-to-face meeting.

• GRBs here and there (Richard, from his notes)

  We made a 3 orbit-hr pointed run with full simulation, and included a GRB in the field of view. So Gregg Thayer has put it into the OBF simulator (prior to the FES) to see how the new GRB finder works. We are redoing the run containing the GRB to remove all trigger and OBF cuts for them.

  The same task can be input to L1 processing; Anders is trying to arrange that. First off is for Bryson to take the ldf files (ebf & exposure) and get them into the ingest/halpipe system. Then would fire up L1. Warren has not got anything from Bryson yet.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:48:12 -0700