Core Minutes 2/7/2012

• ScienceTools: (Jim)

  • Binned versus unbinned testing completed
  • Working on pyLikelihood issues
  • Met with Steve to discuss projects for Emilia. First one will be to make diffuse response more efficient.

  See Science Tools Development Notes for details.

• FSSC: (Tom S.) No big news this week. Work on new version of ST in progress. Ingesting the reprocessed data will provide an opportunity for Alex (new hire) to start and stop server, etc.

• Pass7 reprocessing (Heather) Tom G. sent this report:

  The FT1 reprocessing is complete through 26 January 2012. The rest will be done once Level 1 switches over to the new code. At some point we transfer the kit and caboodle to the FSSC. In parallel to this, Warren/Maria Elena will also reprocess the FT2 files. The full Pass7 reprocessing will be ready to go once the final calibrations are loaded into the appropriate database.

  (Leon) has cleared the deck of other high-priority items. He plans to get together with Joanne today to register new calibrations in the dbs.

• Batch farm news (Richard) Remember the two "black boxes" (which are actually white)? They provide about 5000 cores, and Oracle suddenly decided not to support them any more: end-of-life is coming up in a month.

• GR/Pass8 (Tracy) Gearing up for Pisa. The new GR tag which we had hoped to have last week didn't quite happen — too many new things to be integrated all at once. But it should happen later today

  Concerning truncation, (Tracy) has had no news on fsw software work needed for testbed run. He will poke again. (Leon) has written code to implement Bill's idea (reducing data from new-style truncation scheme to look like old-style). It's in place in P8 branch and also a Pass7 branch. It seems ok in testing so far (which has not been exhaustive).

• Core talk:

  • GR tags and branches (Heather) In addition to the new Pass8 tag (GR V19r4p1gr0) she will soon be making a tag on Leon's Pass7 truncation branch. And there is talk of yet another Pass7 branch so that Markus and Keith can use an updated cosmic-ray flux model to explore extending the measurement of the extragalactic diffuse gamma-ray background below 200 MeV. Heather would like to know what the expected lifetime of this branch is.

  • GR and end of rhel4 (Joanne) has been stymied by test_Gleam bug on the Pass7 branch incorporating new Gaudi. It crashes in GlastClassify with a bad allocation exception while processing the first event. Configuration and initialization stages appear to be ok. AnalysNtuple::execute, which comes just before GlastClassify::execute seems rather slow, but maybe that's normal. The crash in GlastClassify happens on an innocuous line (allocating a local std::string variable) which has already executed successfully twice before. Possible future gambits include

    • turn valgrind loose
    • exclude GlastClassify via job options change; see if there is still a crash somewhere
    • ponder where there might be some sort of mismatch arising from the process of merging new gaudi into Pass7 and affecting memory usage

    (Tracy) There was an agreement for Pass8 to replace GlastClassify with TMine but it hasn't happened yet.

    (Heather) finished upgrades to event display to use gaudi observers, as is done for random numbers, but it's still not working. In fact, even the random number stuff stopped working. She has tracked the problem to boost. Up till now, she has used boost binaries directly from gaudi central, but she is now in the process of making her own debug build.

    Under the circumstances (rhel4 end-of-life only 3 weeks away; no usable Pass7 Gleam yet) we need to pursue schemes to keep a rhel4 box going beyond EOL. (Richard) just sent another inquiry to Teri about extended support.

  • u30 disk (Heather) It's still pretty full, at 95% last time she checked. Michael has been working on archiving old builds. The following is extracted and summarized from his email:

    So far some of the smaller chunks on u09 and u30 have been archived, but the bulk of the u30 stuff (over 400 Gbytes) remains. After getting some information concerning maximum archive sizes he will continue archiving.

  • G4 and vc90 (Heather, Joanne) As of 9.5 Geant support has discontinued building via cygwin in favor of cmake. Heather discovered that, unofficially, cmake files are also available for 9.4, p03. There, as for the cygwin build, the options were not quite right for a debug build, but she was able to change them. Joanne used the resulting libraries to build G4Generator and G4Propagator and everything looks fine: libraries load and test programs run for both.

  • SCons RM (Tom S.) It's generally working well. Some minor problems surfaced and were resolved after Jim attempted to trigger an optimized HEAD build. Windows builds had stopped working; turns out that was because the password was close to expiring. A warning dialog box would pop up and no one was there to click "ok", just another example of Windows/batch incompatibility.

previous

minutes index

next

J. Bogart, Last Modified: