Core Minutes 4/3/2007

• Collaboration meeting: (Richard) See the meeting agenda. It was a full week. Recommended (among others) for your perusal: Eduardo's overview of ISOC GRB ground processing, including a nice data flow diagram, Kevin Grady's report on mission status, David Smith's time test tale, Eric Grove's discussion of trigger and filter — and many more, too numerous to list.

  Beamtest issues include one each for CAL (energy scaling is off; can be fixed in calibration, but no one knows why it's necessary) and TKR (mc doesn't produce as many hits as it should, but it probably doesn't make any difference to the analyis).

  Launch date The official LRD (Launch Readiness Date) is now Dec. 14. This is in conflict with another mission. That one may get delayed; if sufficently so, we'd be next in line. Otherwise, could be more like February.

  Service Challenge update Obsim 2 run was a failure. There were some bright spots, however: Nicola successfully used Pipeline 2 for the processing. Some problems got fixed, but there was a serious one remaining concerning AGN. (Jim) now knows what the problem is (has to with with light curves) so help is likely on the way. [Update from Jim: the AGN problem is effectively resolved. However there is another known problem: the last three days of the year-long simulation generated zero events, even though the program ended normally, with no exceptions or other errors.]

  Finding a plug (Richard) We have a new 50 Tbytes at SLAC, of which 10 so far are installed. We need the rest! The possibility of using some space in the Kavli building on campus is very much alive. The room in question would be suitable with some work, about $200,000 worth. Such a sum might be available.

  (Toby) mentioned a couple items Richard spoke about that the meeting but left out of his own summary: 1) the ongoing discussion of CMT futures and 2) mention of simulation runs being done at U. Washington. The latter uses Condor; it is likely other sites could set up something similar without too much effort.

• ROOT meeting: (Heather) It was going on last week as well. David Chamont attended and will report next week.

• Instrument status: (Richard) ETE1A (End-to-end test 1, part A) has had more than its share of problems. It's supposed to consist of 12 hours of data-taking with the gamma filter. We (SAS) are only required to receive and archive the data, but we were planning to run it through the pipeline. The test has been delayed due to spacecraft problems, followed by difficulties commanding the GBM. The test was rescheduled for this morning; it may be in progress now [~8:30 AM PDT]. ETE1B is scheduled for this Friday. If ETE1A can't be done today it will be scrapped and we'll go straight to B.

• CMT: (Heather) Along with other long-standing complaints, we have to face up to the following:
  • There is no support for VS 2005 from CMT Central and it's not likely there will be on a useful timescale.
  • CMT makefiles can't be made to work for Intel.
  A subcommittee of those who have been wrestling with the issues for some time (Toby, Navid, Jim, Tracy, Joanne, Heather) is tasked with considering options ranging from relatively conservative (make only those improvements which can be implemented incrementally, preserving the current paradigm) to those which involve dispensing with CMT for substantial chunks of the functionality it now handles. (Richard) Toby neatly summarized a key requirement for any contemplated changes on the Burning Issues page:

  A new system must demonstrate that it works first, and must work in parallel. This would apply to each stage of a multi-stage transition.

  (Heather) All others are encouraged to contribute to the discussion to whatever degree they wish. See also earlier Confluence pages CMT Problems and SCons alternative to CMT.

  (Jim) CMT-generated makefiles are not well-behaved in that they don't always rebuild targets that need to be rebuilt and they often do unnecessary rebuilds.

  [(Toby, post-meeting) would like to remind everyone that we already have our own VS2005 support, developed precisely to avoid CMT dependence.]

  No matter what the consensus is on the major direction we take, we should work on the following which would be a help in just about any imaginable scenario:

  • Special handling for test programs, e.g. putting them in subpackages
  • Avoid unnecessary creation of large numbers of environment variables
  • More accurate specification of dependencies.
  
  
• Priorities: Moving on down Heather's list,
  • Should document build procedure for externals to the extent possible. (Navid) This is already the case for some packages. Look for a file called build.log under $GLAST_EXT at SLAC, e.g. for ROOT it's /afs/slac.stanford.edu/g/glast/ground/GLAST_EXT/rh9_gcc32/ROOT/build.log. Sometimes it's under rh9_gcc32, other times under some other tag. When available, source can also be found under $GLAST_EXT/tag/pkgname.
  • Pointer skimming (Tracy) At the Collaboration Meeting this idea was rapidly growing in scope. Can't use Eric's BaBar code as is — too many BaBar tentacles.
  • RHEL4 Goddard is already at this release; they keep some machines at RHEL3 for Glast. RM can build almost all of our code for RHEL4 now, but Navid has run into a roadblock. From his email:

    It's most of the test apps and it's visible in RM. Here's an example. Scroll down all the way to when the test app has finished successfully. Instead of finalizing correctly, it core dumps with:
    *** glibc detected *** double free or corruption (!prev): 0x0844dd48 ***
    It's in a destructor somewhere, probably in gaudi.

  • Static versus shared libraries There is a known problem with switching to shared libraries on Windows for some packages: those which need to export symbols for data as well as for functions. Tracy tracked this down to that part of CMT which attempts to discover this information by parsing the dumpbin file and has a patch.
  • System Tests Leon and Frederic Piron will be taking over the reins from Julie. (Richard) It could be that Claudia Lavalley will have some time to work on the backend as well.
  • Workbook In response to a question about saving snapshots: this is already done. Chuck makes tarballs of the Workbook when it's tagged.
• Proposed GR Release: (Toby) Would like to make a new HEAD build incorporating spacecraft geometry among other goodies. (Richard) Let's hold off on the new geometry until it's been looked at. (Tracy) Should also wait until he has finished implementing Pass 5 classification.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:46:43 -0700