Core Minutes 8/21/2007

• Core Week: (Heather) See the list of suggested topics with associated links. Out of town arrivals included Toby, Navid, Eric W. and me. [And one more very important person whom Heather unaccountably omitted.]
  • SCons development Navid walked us through his design and progress to date (note that he has gotten further since then), describing among other things contents of the main SConstruct file and other python scripts he's written to support kinds of functionality we're accustomed to. A checkout package like ScienceTools will be implemented by means of symbolic links under it (to start, under a clone of ScienceTools with a slightly different name) a within the CVS repository. Then CVS checkouts will transparently check out all the required packages along with the checkout package itself.

    Navid and others would like to change the tagging convention to something which will be sorted correctly by OS commands like ls. At a minimum this would entail using a fixed number of digits, zero-filled, for each of the three numeric fields. There is some sentiment for replacing r and p as delimiters with punctuation characters, however this has significant repercussions for the Workbook and perhaps also for MRStudio. We should weigh the benefits versus costs carefully.

  • Event collections is working, but is not yet in a GR tag. Ideally the Data Server should know about the new files; David Chamont is working on this.
  • External libraries Although we nominally support debug and optimized versions of all of them, in practice for most libraries there is only one version with a symbolic link to it for the other variant. Navid would like to officially only provide one library, usually the optimized one, for each external. There was agreement that this would be ok as long as developers are provided with either comprehensive building instructions or a debug library upon request.
  • Python libraries Eric Charles would like to add one for XML handling, needed by calibGenCAL, to Jim's list.
  • System tests RM's automatic triggering of Sys Tests builds is failing. Until this is understood, this triggering will be disabled and Leon will push the button manually. RM will still generate the configurations needed, since this is being done correctly and saves the work of doing it by hand.
  • EM retirement With some notable exceptions (new DFI library and changes it requires) future EM releases are expected to be only patches.
• Big run readiness (Richard) See also the full wish list.
  • ACD digi, recon, geometry (Richard) The goal is to get this all done and ready for testing by the beginning of next week. (Heather) ACD digi should be done by the end of the week using Eric C.'s prescription. (Richard) Eric C. will be out of town next week and needs a geometry model incorporating new features for his testing and development of Recon. (Joanne) Slanted tiles are done; progress is being made, step by tedious step, on side ribbons to match. In the fred pictures you can see (red) side tiles and the little extra pieces for bent tiles (other top tiles are now invisible since fred doesn't show envelopes), screws (blue), y-measuring ribbons (yellow - they're perpendicular to the plane of the drawing so you can't see much), x-measuring top ribbon segments (green) and bit behind bent tiles (also green). The yellow and green ribbons are unchanged from the last tag. The aqua bits are new. You can see what's left to do for X-measuring ribbons in this CAD picture in the YZ plane. She's hopeful something adequate for Eric should be ready by COB on Wednesday.
  • OBF status (Tracy) It's about done except for writing the information to ROOT trees which, he's reasonably confident, will not be a big deal.
  • GR v12 (Heather) The goal to start is to provide something Anders can use for Pipeline 2. The first attempt had some problems. She is hopeful they can be addressed today; the changes from GR v11r17 are not that great. (Leon) System Tests ran successfully. Outputs are similar but not identical to those from v11r17. He has not yet had a chance to see if the differences are simply those expected from new code. So far he has seen nothing obviously wrong.
• The new, improved 55-day run (Richard) This is essentially the 55 day run redone with the problems fixed. But it will emerge as the first two months of operations. So there will be two weeks of pointing at vela followed by 4 weeks of survey.

  Tom is starting to regenerate the Earth10 backgrounds and will just do a 1 sec/minute sampling on the two month orbit to get a day's worth. Julie added a pointing episode at the end to make sure there is enough to account for the GRB repoints that have not been included in the orbit file yet.

  This will be done with GR v11 vintage code.

• Trigger information (Richard) For real data the true trigger information comes from FSW. Some of the "same" information is calculated by OBF on the ground. We don't have a mechanism for storing both in the tuples. Furthermore not all the FSW information is accessible to Offline.

  (Bryson) The filter "result summary data" (RSD) is an array of data structures transmitted per event that contains a concise record of what each filter (also known as a "handler" in FSW parlance) had to say about the event. The shape of these structures is (or may be) unique to each type of handler, and may evolve over time. Also, the arrangement (ordering) of handlers is configurable and may also change over time. Thus designing a presentation format for delivering the contents of these structures to downstream clients is a thorny problem, since the goal is to decouple the offline client code from changes in the FSW code. JJ also contends that this content is reproducible on the ground (once, as Anders pointed out, that you've verified that the offline implementation of the handlers produces identical results). His (JJ's) priority assignment is implementing the creation and delivery of per-handler statistical summary data (including rejected events), so designing the RSD presentation format is pretty much on hold.

  There is one other piece of dynamic information we really should have which, fortunately, is much more straightforward to extract as part of DFI: we need to know which handlers are active and for which nodes. Philip Hart is taking this on, but, longer term, DFI needs a new owner. The original designer and implementor have left the project.

• AOB (Heather) Anders requests that we look at the System Tests currently run only for EM and see which should now be run for GR.
• Acknowledgements: Thank you Bryson, Richard and Heather!

previous

minutes index

next

J. Bogart, Last Modified: 28-Aug-2007 16:10:55 -0700