Core Minutes 5/15/2012

• SeeVogh First trial today was not a success. We quickly retreated to EVO.

• ScienceTools: (Jim) A new package, SolarSystemTools, has been added to ScienceTools. It will be promoted to HEAD soon, once it builds on all platforms of interest.

• FSSC: (Tom S.) No news. It's a week since Eric left; Alex has been picking up his responsibilities.

• Pass7 development: (Leon) No news this week.

• Pass7 reprocessing: (Tom G.) Now about 78% complete. If all goes well this step will be done by about June 2nd or 3rd; then FITS; then catch up with current data.

  (Leon) We're missing a couple CAL calib files for this year's data. Sasha is working on generating them.

• New servers: (Tom G.) There is a plan for new disk servers and how they are to be allocated. Replacement of compute servers is not so far along

• Pass8: (Tracy) We're hung up on getting a merged GR tag that works. Things seem to work on Windows; there are problems on Linux. We're a couple weeks behind where we had planned to be. (Heather) New tags of Overlay and Interleave should do the trick. She had to rearrange CMT requirements files somewhat.

• OBF filter bits: (Joanne) After two long debugging sessions, JJ and I believe we understand the discrepancies in HIP and MIP filter bits reported when we run against B1-1-3 and B3-0-0. It's a conjunction of a couple things:

  1. B1-1-3 is old enough that, when there is an unexpected diagnostic contribution, it's called an error. The build now running on the instrument does not treat this occurrence as an error.
  2. In case certain kinds of errors occur, including the unexpected-contribution error, there is a bug in the HIP/MIP status reporting which results in the loss of bits we've observed with B1-1-3.

  So the problem is not in our code, but running against such an old build leaves us open to various kinds of discrepanies. This particular one probably has no serious consequences — anyone who cares about these bits would use the ones reported in the original data, not the ones from our OnboardFilter package. Still, it seems like we should have caught this earlier from monitoring or from systests, but currently there is nothing in systests using real data. This incident demonstrates that we ought to have such tests. (Heather) agrees we should but doesn't know where the resources to do it will come from.

• Other GR validatation: (Heather) For rhel5 validation she needs to find a way to get random seeds lined up.

• Core talk:

  • RM news (Tom S.) Everything is working! He fixed the deleteBuild and eraseBuild programs so that they will now delete Windows zip files belonging to the build which should keep u35 from getting cluttered.

  • GR vc90 (Joanne) fixed the rootcint problem of last week, then attempted to build everything in the allGleam solution file in Visual Studio (that is, all libraries, Gleam executable and test_Gleam executable). Of the 90 or so projects, about 70 projects built successfully and the rest failed. There were 6 or 8 independent kinds of failure. Some of these (the easy ones) have now been fixed. In a second build trial there were only 10 failures coming from 2 or 3 root causes. There is at least one show-stopper among the remaining types of failure: test_Gleam and Gleam fail to build because the project files do not currently emulate the fancy footwork in the SCons implementation for Gaudi programs. Project files for any other Gaudi test programs would be similarly deficient.

    (Toby) notes that python wrapper libraries are failing to build on Windows for ScienceTools. (Joanne) also noticed this a while ago, but it's not relevant for GlastRelease and she wasn't aware anyone was using the SCons Windows build of ST. Other work has taken priority. (Toby) might be able to look into it.

• Announcements

  • NO MEETING next week due to CHEP. Richard, Heather, Anders and Tony will be going.

previous

minutes index

next

J. Bogart, Last Modified: