Core Minutes 3/18/2008

• Launch date: (Leon) There was some news about this at the Collaboration Meeting. The missions on either side of us are higher priority than we are and one or both may slip, in which case we could slip with them. Apart from that possibility, the May 16th date appears to be firm.

• Big run: As always, follow progress on the Big Run Checklist.

  • Reprocessing morass (Leon) has dug a deep hole trying to get the Pass6 reprocessing to really work. As he tries to get more of the variables to come out right, he's had to drag in more and more code, and is discovering more TDS objects that probably should be persistent, but aren't.

    He thinks it's time to abandon this approach, and instead look at another one: to read in the existing merit file along with the regular root output files, redo the vertexing, modify only those merit variables which are supposed to change, and write out the modified ntuple.

    Tracy and Heather agree. This will require writing a meritRootReaderAlg and modifying ntupleWriterSvc so that AnalysisNtuple properly handles adding variables that already exist in the ntuple. Some of this functionality already exists in GlastClassify (and Interleave?) and in RootIo, so it shouldn't be too hard, say days, rather than weeks. Once it's done, though, it will be available to use again when similar cases arise.

  • GRBgrid woes (Heather) Recall the GRBgrid runs are interleave runs submitted as batch jobs in both Lyon and SLAC as part of the Big Run festivities. The GRBGrid runs are still on hold, pending understanding of what turned out to be two issues:

    • xrootd bugs Wilko contacted the developer and we now know that xrootd has serious issues in both 5.16 and 5.18 which can cause thrashing in some cases. The recommended fix is in ROOT 5.18b, which we are now compiling and plan to move to. Our current GRBGrid run tests are using ROOT 5.18, and Tom has moved the input Interleave ROOT files out of xrootd for now
    • bad allocation exception are thrown for a number of runs. In these cases, it is not due to an over-use of memory, hence not a real memory leak. Instead, sometimes, seemingly randomly, performing the TChain::AddFile call in Interleave throws this exception. In fact, sometimes the call will succeed for the same file where there was a previous failure. Heather is now building a debug version of ROOT 5.18. Hopefully this will allow us to run gdb and actually see what is causing the problem. Tom wonders if there is perhaps a specific issue with the ROOT files themselves, and hopefully we'll soon know the answer.
    Many thanks to Tom for his insights and running so many test jobs!
  [and to both Leon and Heather for these contributions to the minutes. ed.]

• Science Tools report: Seth went through the Science Tools Update for March 18.

• SwigPolicy: (Toby) One of the requests coming out of the Extending ScienceTools splinter session was to make public C++ classes more readily available to python users. He made some changes to SwigPolicy to build a single dll containing both the C++ classes and the python interface classes which appears to do the right thing on Windows [however he later discovered the changes cause failures on RM, so they have been backed out].

• GR news: (Heather)

  • Newest GR tag, v13r13, has gone through SysTests; she hasn't yet heard how it fared.
  • OpsSym2 used v13r11; there will be updates along that branch for use in the Pipeline
  • The new DFI is out, includes changes to LCI classes and changed one of the flags from isMissingGps() to sourceGps(), thereby changing the meaning of "true" as well. This modification needs to be reflected on the Gleam side and made accessible using a similarly named method. All this needs to be in place for ETE 5.
  • The vanilla build of ROOT 5.18 is available but of no use because of the problems with xrootd. There will be a build of 5.18B shortly.

• Moot news: (Joanne)

  Centralized support in Gleam for accessing LATC configuration

  Trigger code in Gleam is already using MOOT routines to find archived source files for GEM configuration. New tags of CalibData and CalibSvc simplify the interface by handling everything having to do with the MySQL connection and updating cached information when the LATC configuration changes.

  Still to do: Code is awaiting testing with a real client, like Trigger.

  Acquisition summary information

  A new function, updateAcqSummary, stores a limited amount of per-acquisition information in MOOT, including hardware and software key, number of events, etc. The function will be called by Bryson's half-pipe code. He has tested it and it seems to be ready to go.

  Still to do: Provide query routines so information can be accessed from Gleam.

  Filter and other software configuration

  A new function, registerConstituents, intended to be invoked after a new fsw software build has been made, stores information about each constituent (data constituents like gamma_normal, cal_db_gains and also executable code constituents like gfc). This will be an element in the grander scheme now beginning to crystalize.

  Still to do: Plenty.

• Skimmer: (David) In an attempt to avoid various problems noted by Tony, Skimmer code is now entirely compiled; it does not use the ROOT interpreter.

  Work to get the Skimmer to take CELs as input is in progress. It will require changes to the user interface.

• SCons news: (Heather) Lots of progress to report! (Navid)

  • Thanks to substantial help from Heather, ScienceTools now compiles with SCons on both Linux and Windows
  • The new RM is up and running. Karen's web interface is so far only available inside the firewall. It still needs some tweaks, but generally it's looking good.
  • Emmanaul is working on the Mac compile under SCons. Performance, already a problem on Windows, is really dreadful on the Mac. SCons wants to read every .h and .cxx. It takes about 1 minute on Linux (nfs), 5 minutes on Windows (local disk) and on a Mac (also nfs) 20 minutes! It's likely that SCons Central will have some suggestions.

• Documentation: (Chuck) There have been many updates recently to ScienceTools tutorials, thanks to Analia Cillis. ISOC web site updates are proceeding rapidly.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:47:40 -0700