Core Minutes 7/20/2004  

Present:    Joanne Bogart,  Toby Burnett, Jim Chiang, Johann Cohen-Tanugi, Seth Digel, Richard Dubois, Marco Frailis, Riccardo Giannitrapani, Berrie Giebels, Navid Golpayegani, Traudl Hansl-Kozanecka, Heather Kelly, Michael Kuss, Matt Langston, Julie McEnery, Pat Nolan, Chuck Patterson, James Peachey, Sean Robinson, Leon Rochester, Robert Schaefer, Tom Stephens, Tracy Usher

• Upcoming events: (Richard)
  • There will be an SVAC review 10-12 on Wednesday here at SLAC.  It may also be possible to dial in.
  • There will be a Science Operations Center review on Aug. 4.
  • Barbecue next week!
• Package dependencies: (Traudl) From this index page you can find links to many graphs like this one displaying dependencies among packages and to her presentation (pdf, ppt). The presentation describes the tools used to show dependencies and contains suggestions for reorganization.  There was time for only a few brief comments:
  • (Toby) Proposes that the tool used to create the diagrams be invoked by Release Manager routinely for each new release.
  • (James) Asks that the reorganization effort be split into a GlastRelease part and a ScienceTools part rather than attempting to tackle everything at once. Traudl did not mean to propose that ScienceTools undergo reorganization as it is in good shape already.
  • (Joanne) Contrary to what is shown on slide 7, the concept of validity interval has no relevance for geometry as provided by GlastSvc.  GlastSvc is concerned only with ideal geometry. (Traudl) the chief point of the slide is the proposal to separate GlastSvc into a service for event data and another for geometry.
• More Core issues: (Toby) See a collection of other items of current interest on his summary page, such as an updated list of largest files and a curious compile error spotted in recent Release Manager output. (Leon) Most unit test programs could probably get by with much smaller data files, including only a few events.

• st_graph: (James) He's working on a plotting/graphing package for ScienceTools to be called st_graph.  Initial implementation uses ROOT for its backend, but it will not use features peculiar to ROOT. Minimalist interface will make it easy to avoid dependence on any particular backend and should simplify maintenance.

• Installer: (Navid) His write-up describes the new installer: some background on how it works and examples of how to use it.  It will install external libraries as needed (in particular, it will not redownload libraries if they're already installed) for the combination of [top package, version, OS/compiler CMT tag] requested by the caller.  Currently it is implemented as a Perl script and requires the Perl DBI package. He plans to rewrite it in C++/QT to make it more portable. A later version will handle all binaries (e.g., libraries for Glast-written packages), not just external libraries. Since the Installer is dependent on output from the Release Manager, it currently can only serve Linux users.  It doesn't have a comparable record of libraries for Windows (Toby and Richard suspect that, even without RM builds of Windows, there ought to be a way to handle requests from Windows users as well.  To be settled offline.)  For anyone (on either platform) needing Perl, Bob suggests the free ActivePerl distribution.  You can find out about DBI here. (James) Is the Installer primarily for end-users or for developers?  (Richard) Primarily end-users.  Developers might be interested in the output of another of Navid's projects which produces source tarballs for each OS as part of the RM output, available from a summary page like this one.

Flight integration news for the week:

• EngineeringModel news: (Heather)
  • Just made new tag v3r0402p19 to pick up fix to em2 geometry.
  • I&T delivery date is August 13th.  Would like to aim for a final-ish version by August 4th.
  • New LDF library is available.  Has been installed on unix; still need to make a Windows build.
  • LDF System/Unit test is about done. Still have to deal with GEM data and  finish ACD There is a bit of a snag in that the Online c test program does not handle FITSified files, only the naked LDF format. May need to strip off the FITS.
• Calibration news: (Joanne)
  • She made a tag late last week of CalibSvc which uses timestamp from LDF data to select appropriate calibration. Thanks to help and information from Toby and Sean, she ought now to be able to do the same for MC data.
  • ROOT calibration files: some progress.  Will be touch and go to get this in place by the Aug. 4th deadline.
  • Movable components: during integration it is possible that a particular calorimeter or tracker will be moved from one tower position to another.  In order to keep a history of hardware performance, I&T needs to be able to associate serial numbers with calibration data. A couple different approaches have been proposed; no resolution yet.
• Tracker: (Leon)
  • improved bad strips calibration
  • ToT on hold, awaiting definition of format, ability to read the file into Gleam,...
• Údine work: (Riccardo)
  • Random event access from FRED is working.
  • He and Marco have made a new release of MRvcmt, v0r7.  Among other things it incorporates a fix for recent problems in generating Solutions files and the ability to build all binaries (libraries and executables) belonging to a package by selecting the new target any. [The new version has been installed on SLAC Linux and may be accessed with either of the comands $GROUPSCRIPTS/MRvcmt-HEAD or $GROUPSCRIPTS/MRvcmt-v0r7.]

• Pipeline: (Richard for Dan) At a meeting last Monday the decision was made to reorganize Pipeline directory structure for I&T.  This has been done.

And finally..

• System tests, unit tests: (Leon) He once had a test for TkrDigi code which does a diff on output versus expected output. It doesn't fit the standard form for either system test or unit test, however when this kind of test, whose output is a very clear "pass" or "fail", is available, it's usually preferable to a statistical approach. (Joanne) If it were possible to invoke a script for a unit test (which in turn could invoke one or more test program and also shell commands like diff) Leon's test could be a unit test.

previous

minutes index

next

J. Bogart Last Modified: 01-Jun-2010 15:47:18 -0700