Core Minutes 1/11/2011

• ScienceTools: (Jim) Since the last meeting before the holiday break Johann has made several updates required for compilation with the gcc 4.4. There is a new tag of pyLikelihood which has nothing new in it except a name change, for clarity. Finally, there is new tag of Likelihood with an enhancement to gtexpcube2. See entries (starting with the one dated Dec. 20) in the Science Tools Development Notes for details. [Note also a new tag of ScienceTools, ScienceTools-09-20-00, was created the day after our last meeting before break, as promised at that meeting. ed.]

• cfitsio: (Heather) James Peachey sent an update stating it would be another couple of weeks to get the tip patches in place to handle the cfitsio upgrade.

• FSSC: (Heather) No report today (also no Goddard people present and no access to public data) because the HEASARCH site has been hacked. They hope to be back in a day or so.

• Documentation: (Chuck) sends the following:

  If all goes well, the revised version of the Binned Likelihood Analysis will be posted to the web Wed. afternoon, ready for review and comment by the Subject Matter Experts. Plans then call for work to resume on updates to the remaining Science Tools, beginning with gtexpcube2, which is the first of the Science Tools to require a name change. The next newsletter is scheduled to hit the streets on Friday, February 4th; deadline for submissions is Tuesday, February 1st.

• Pass8: (Tracy) A new GR tag, v18r8p1, was made last week, the first to include all the new Cal and Tracker recon code. There were naturally some problems which are in the process of being addressed. Will commit patches today and make another tag soon after. We hope to use it to create real datasets, to be available in a few days and suitable for studies before the next Pass8 Face-to-Face about a month from now.

  The Pass8 code takes longer, mostly because of the new tree-based tracker reconstruction. This was expected. It's actually faster for many events, but there is a long tail of events which can take minutes, even tens of minutes, to reconstruct. The goal is to identify such events and cut off reconstruction on them. If this is done, Tracy is confident his goal (no more than a factor of two increase over the old pattern recognition) will be achieved.

• Core talk:

  • Passwords (Heather) It's time to change various system passwords, such as those for glast and glastrm. We're hoping we can get by with passwordless access from now on for those who occasionally need to log in to these accounts for various maintenance and development activities. That will minimize disruption caused by the requirement to change passwords every six months.

  • Missing SCons log files (Heather) Tom S. found out from SCS how to increase the MySQL table size and this has taken care of the logging problems.

  • New Windows box (Heather) It has Windows Server 2003 installed; we're waiting on LSF.

  • Snow leopard (Heather) SCS is eager to pull the plug on our Tiger machine due to security problems with this OS. John Bartelt made a special request to Security on our behalf to keep it going for another 90 days (but that was on Dec. 16; only 64 left!) so need to get the Snow Leopard machine fully functional pronto.

  • CMT ScienceTools builds (Heather) Jim has proposed we shut them down as soon as can be managed. Some issues must be addressed first:

    • There is no SCons equivalent of the CMT Red Dot package summary web page.
    • SCons builds do not generate Doxygen documentation.
    • There are no clean RM builds of ScienceTools on Windows. Jim sent Heather a list of packages with problems.
    • Joanne's container tagging script needs to be dusted off to be used to trigger HEAD and tagged builds since the current system piggy-backs on CMT builds.

    We will announce the intention of shutting off CMT builds with the expectation that it will take a week or two to accomplish.

  • SCons and GR (Leon) has been working on fixing SCons failures (on Linux!). All packages now build, but there are still several test programs which fail, probably because some form of set up is not being done (e.g., missing values for environment variables which, for CMT builds, come from requirements files).

  • SCons and Windows miscellany (Joanne) has been tangling with issues arising from building GR with VS 2008. Here they are in order of increasing intractability:

    • test_Tables app in Event wouldn't compile. This turned out to be a case of newer compiler fussiness. Tracy made a suggestion which took care of it.
    • Apps dependent on XGetopt (Windows-only substitute for standard unix function getopt) would not link. This is because XGetopt is currently part of the facilities shared library and needs to export data as well as a function. Our current method of building dll's won't handle this. However, if the XGetopt module is removed from the facilities shared library and linked into the executable directly, all is well. Alternatively, one could put XGetopt in its own tiny package which would build a static, rather than shared, library.
    • The TMine library still won't build. An alternate spawn technique on our 3-step process to create a dll (make .def file with bindexplib, perform the link, insert manifest file) got most of the job done, but the insert-manifest command is failing. I'm now thinking of dropping this approach in favor of splitting up the TMine library.
    • The G4Generator library fails to link due to unresolved references coming from modules in G4HadronSim. We have two versions of G4 built with VS 2008, one debug (built shared, which it probably shouldn't have been) and one not (built static). Surprisingly, linking against the non-debug, static G4 build produces more unresolved references. It may be that neither library is built correctly for our circumstances. Tracy suggests consulting Norm Graf, the official Windows G4 expert.

  • SCons meeting Wednesday at 10:30, as usual

previous

minutes index

next

J. Bogart, Last Modified: