Software Coordinating Committee Conference Call
		       September 9, 2004 3:00 PM EDT
			  Recorder: C. DeTar

Present:

  DeTar, Francesca Maresca, Holmgren, Brower, Osborn, Dru Renner, 
  Edwards, Watson, Mawhinney, 

Absent:

  Neff, Jung, 
  Simone, Pochinsky, Petrov, Mendes

** action items
=====================================================================

Showing off new web page 

   Brower:

      See http://super.bu.edu/~brower/lqcd/
      There will be links to individual group pages linked to this site.
      Please comment on the name of the collaboration

        QCDSP

Code release procedure

   DeTar: Who is the librarian?
   Edwards: Put pages under CVS.  Then the developers are their own
     librarians.
   Watson: Pacman is software that tracks dependencies and is able to access
     multiple repositories.
   Brower: Do we need to use something this sophisticated?  We have
     only four packages.
   DeTar: It would be best to have a single tar ball and single
     build process.
   DeTar: 
     Two obvious models of testing code:
     Developer model: A developer or tester is given an account
     at each center and tests each release on all.
     Center model: We need a testing regimen at each center to test candidate
     release code at the local installation.
   Holmgren: I prefer the center model.
   Brower: I agree with the facility testing model.
   Brower: Bob, is this OK with you?
   Mawhinney: It would be difficult for us to test code that we
     didn't write.
   DeTar: The idea is to have a "make check" procedure. 
   Mawhinney: That would be fine.
   Osborn: Not enough just to have a "make check".  We need a suite
     of high level code that builds on a parallel machine.

     [ There was a consensus that we use CVS for the release web page
       with links to the packages and mirrors at each of the labs ]

   DeTar: How does a developer post a new tarball?
   Watson: Check in the tarball under CVS and the updated web page
     that links to it and the same cron job can simultaneously copy
     the tarball and the web page.
   Brower: Can we have someone start the web page?
   Edwards: I'll get a web page started and a new module under CVS.
   Osborn: We need bug tracking and mailing lists.
      The mailing list would be for discussing the best compiler
      flags, etc.
   Edwards: savane does that sort of thing, but not trivial.
      There is a bugzilla.org that looks like it is hosted on
      sourceforge.
   DeTar: Maybe a simple listserv is the way to start.      
   Brower: Don and Carleton should put together regression
      testing for FNAL.
   Edwards: We need scripts and some diffs for comparing results.
   Brower: Status of QMP testing?  Celso Mendes is leaving
      the project.  He is willing to finish the test suite.
      Bob Sugar and I are discussing a possible replacement.
      It might be good to bring in someone whose responsibility
      is more strongly testing.

   
=====================================================================
Committee conference concluded at 4:20 PM EDT.  
Next plenary call September 16 at 3 PM EDT.