Core Minutes 3/29/2011

• Where to keep what (Richard, Jim, et. al.) Various options for storing diffuse models and catalogs are under discussion. Maintainers must have the ability to update files with little or no intervention. Another requirement is that it must be possible to disallow public read access. CVS does not meet the privacy criterion without extra work, possibly substantial extra work, of one sort or another, so [at this meeting and shortly after] we decided to go with external libraries, modeled on but separate from the existing external library extFiles. Eric W. will need to know, for each ScienceTools release, which data files should accompany it in the public release.

  One option considered above was to disallow public read access on the entire CVS repository. Several (Richard, Heather, Julie, Joanne at least) were not enthused about this idea. Jim reports that Gino thinks that the user area at least should be more private.

• ScienceTools: (Jim) ST 09-22-00 was tagged on March 19th. Since then there is a new tag of caldb. See Science Tools Development Notes for details.

  The FSSC wants the IRFs to be automatically specified based on the data selection performed by gtselect. Downstream tools would determine the IRFs to use based on the DSS keyword information inserted into the file headers by gtselect.

• FSSC: (Eric W.) Topics of interest have already been mentioned.

• Documentation: (Chuck) sends the following:

  So far, the KIPAC FTOOLS packages are working fine in a bash shell, and I'm standardizing the setup for the Science Tools tutorials accordingly.

  Currently, I'm working on the pyLikelihood: Binned tutorial and hit a glitch late yesterday a.m. This tutorial is still a good introduction to pyLikelihood analysis, but I think we need a new, expanded version as well. However, it's probably more important to finish updating the remainder of the ScienceTools tutorials first to bring them into line with the new setup procedure, and to make sure they are working as intended.

  (Richard) Gino and Eric have called a meeting tomorrow at 3 to discuss our somewhat fragmented web presence.

• Pass8 and related matters: (Tracy) (starting with a related matter) For some time JJ & co. have been working (on and off) on upgrading obf to be rhel5-compatible and also to disentangle the code so that we would not be dependent on so many flight packages. This work is nearing fruition. Two weeks ago JJ gave Tracy a pile of stuff. He will try to make it go first on Windows but will backtrack to Linux if need be. Guess is this will take a small number of weeks.

  Moving on to Pass8 — the Face-to-face meetings have been quite effective so another one has been scheduled for July in Santa Cruz.

  Recent tracking work has been aimed at minimizing dependence on obtaining axis from CAL. He's implemented a filter which, more often than not, does a good job of finding the axis on its own, but fails 20-30% of the time. However, it generally can determine when it's failing. The best strategy may be to go back to using information from the CAL just for those failures.

  (Leon) The TkrReasonsTool is ready to go — in the head of GR, but not yet incorporated into Tree-based pattern recognition. It's a set of methods to help determine whether a cluster is missing for a good reason, that is, being close to an insensitive region: failed plane, gap, bad strip, truncation.

  Next on the agenda is to finish up truncations; that is, to make the best possible use of the knowledge that the hit buffer was full.

• Core talk:

  • GR/L1 (Heather) New tag on the L1 branch has ability to skip events by event index. If there are problems processing certain events, the error messages generated will contain the event index. One may then make another pass, skipping the bad events.

  • Windows binaries (Heather) Windows user distributions are not quite functional. Even when the build creates everything that should be created, the zip file doesn't have everything it should.

  • New windows box (glast-win04) (Heather) SCons builds are using it now; when Kim returns we could try CMT builds as well. Queue parameters have been adjusted appropriately so that there is no longer an artificially low limit on number of concurrent jobs.

  • Windows SCons builds (Heather) For the last ST the vc90 builds have gone through, but not vc71. (Tom S.) The vc90 builds are only there because he made them by hand — which works fine. But the lsf jobs are not running properly. He hasn't had a chance to look at it recently because he's been busy with an ingest at FSSC.

  • RMViewer 1.0 RC3 (Joanne) This new release of RMViewer has some modest improvements. Binaries for rhel4, rhel5 and Windows may be downloaded from ftp space. The code has been ready for some time; the release was held up because I had no suitable Qt installation on Windows to build against. New features in this release are:

    • "Quit" menu item (or use control-Q) in addition to old method (deleting main window).
    • More compact (not so wide) single-build display
    • Floating checkout, compile and test output windows may be searched and their contents may be saved to a file.

  • Include files on Windows (Joanne) Heather noticed that, while builds with SCons on Windows would search the local package directory for includes, when the same package was built using the SCons-generated project files, only the installed include directories would be searched. This bug is fixed in the latest SConsFiles tag. However there is another difference in behavior between building with VS and with SCons which will likely be very annoying to Windows developers: the SCons-generated projects don't know how to install include files (or xml files, python files, etc.) as part of a build.

  • G4 (Heather) In the never-ending quest to have genmap run on the G4Generator library produced using vc90, we built G4 9.4.p01 and CLHEP 2.1.0.1 debug. After obtaining Francesco's updates to our code to handle those upgrades, we can now build G4Generator in SCons. Unfortunately genmap fails in exactly the same place it did when we were working with G4 8.0.p01. Running genmap through the vc90 debugger reveals where the failure occurs. Next step is to try out the shared version of G4 9.4.p01 and hope that this problem disappears.

    It should be noted that using vc71 and G4 8.0.p01 "works" in that genmap runs successfully on the resulting G4Generator library. Unfortuately vc71 is no longer officially supported by G4 but they said we could try it out at our own risk. Heather has since attempted a vc71 build of G4 9.4.p01 and it fails out of the box with about 4 errors, that we may be able to work around. [but the consensus is that our time is better spent on vc90 rather than trying to make G4 9.4.p01 work on vc71]

  • SCons meeting 10:30 AM PDT on Wednesday.

previous

minutes index

next

J. Bogart, Last Modified: