Core Minutes 8/23/2011

• ScienceTools: (Jim) Activity since our last meeting on includes

  • Default irfs are now all set to P7SOURCE_V6
  • Pulsar tools updates
  • Made release ScienceTools-09-25-00, incorporating all upgrades since June 30.
  • gtselect update

  See Science Tools Development Notes for details.

• FSSC: (Eric W.) No ScienceTools news; they've been working on other things.

• CVS ChangeLog caper: (Heather) Late last week Tracy encountered a problem with CVS: the GlastSvc ChangeLog file somehow got corrupted, making it impossible to do any operations on that package. Deleting the file from the repository (with unix rm) allowed development to proceed. Later Joanne was able to recover a good file from the corrupted one and copy it back to the repository. A couple lines had been improperly repeated; the fix was to delete them. (Richard) The repository is backed up nightly and the back-ups are kept for a month so we could have recovered that way. (Joanne) Pat Nolan remarked that keeping CVS repositories in nfs is not recommended. Possibly the corruption was due to nfs flakiness. However it is apparently quite rare and, in this case at least, was not that difficult to find and fix.

• Workbook: (Heather) has successfully put in several updates from Jim. (Elizabeth) is wondering why she hasn't seen the emails she used to see when Chuck did updates. (Richard) Probably these were just the CVS update emails. (Heather) Since we're going through CVS as Chuck did, those emails should be generated. Will check.

• Pass7 reprocessing: (Tom G.) One last run was pushed through for the partial Pass7 reprocessing. No news on full reprocessing; a crude version of the full reprocessing task is ready to go. (Leon) envisions the following steps:

  1. Compare a merit-reprocessed run (P120) with the same run fully reprocessed, run with the same stale calibrations. Results should be the same within round-off. Tom is about to do this.
  2. Compare a recent run processed in the V17 pipeline with the same run processed using fresh calibrations. Check for significant changes. TKR constants are about ready, Sasha will produce the CAL constants covering the recent data.
  3. (Optional) Run a larger chunk of runs to check some physics result, Earth limb, for example.
  4. Decide if we want to reprocess everything.

• Pass8 and computing resources (Richard) Tracy has pointed out we'll need more horsepower for Pass8, but it is not forthcoming here at SLAC. We have 1600 cores; Lyon has 1200 cores and some chance of an increase. Is it feasible to do move some of the L1 work there? The main challenge would be how to get 10-15 Tbytes of digi files over there. Output would be handled as it now is for MC runs.

• New server status (Tom G.) 4 are up and running; 5 more are still in the wings.

• GR/Pass8 (Tracy) Thanks to Heather we have a new release tag, v19r3p3. The main new feature is a modification to GlastClassify allowing other components to access it via an abstract interface. This came about because Bill Atwood has been working on how to split energy between the top 2 tracks. There is no single variable which can be used to determine this; one can do much better using classification analysis. For an example of use, see TkrEnergySplitTool.

  (Tracy) Since 2001 we've been use our homegrown package, GlastMS, for multiple scattering because the G4 native simulation has been deficient, but it has been straightforward to switch to the G4 multiple scattering for testing when desired. In newer GR tags we're using Geant 9.4 and the default is now to use the native G4 multiple coulomb scattering. It is still possible to switch back and forth, but the method is different. Physics choice values of EMStandard or EMLow select our MS; anything else gets the G4 version. (Richard) Do we know that the G4 MS package is inadequate? (Tracy) Don't know. There are differences in simulations results which are not yet understood. G4 people are firmly of the opinion that they've got it right.

  (Tracy) Next week on Wednesday morning is the Pass8 Splinter meeting at the Collaboration meeting. The first hour of that meeting is devoted to a pedestrian review of all the new features of Pass8...should be a good review.

• System tests (Leon) jumped in this week-end but that should not be taken to mean that he will or should be the Sys tests person. Furthermore, it's time to review what's there and compare with what we want and need.

  (Heather) There should be at least 2 people conversant with Sys Tests. Also, now that SCons builds of GR tags are close to reality, she would like to start discussions with Tom S. about automatic triggering of Sys tests by the SCons RM. Once upon a time long ago this worked with the CMT RM.

• Core talk:

  • Windows (Heather) CMT Windows builds have been running since the known_hosts file was made read-only. Tom S. tracked down the continuing SCons Windows problems to a similar cause: disappearnce of a QT configuration file. This is complicated by the fact that we don't know where the file is supposed to be, though Tom does know how to recreate it and after he did at least one each ST and GR builds went through reasonably successfully (just mundane compile and test program errors). Meanwhile, Heather experienced a similar phenomenon first hand: she tried to read a file and that very act caused it to disappear!

  • Python 2.7.1 (Heather) The new ST tag is using python 2.7.1. We are not distributing iPython with it since, when iPython is built, it ends up with hard-coded paths which do not resolve properly when the release is installed elsewhere. It is straightforward for users to build iPython locally after installing ST.

  • GRBAnalysis (Joanne) GRB developers have now moved entirely to the new SCons checkout package, GRBAnalysis-scons. Some minor issues conscerning environment set-up have been resolved to their satisfaction.

  • Virtual machines (Toby) See the ongoing discussion. The bottom line is that Tom S. has agreed to implement the additional features Toby has requested.

    Is it time to start thinking about virtualizing GR? Richard has asked Joanne to look into it.

  • SCons odds and ends (Joanne) Bill's student Brendan has been trying to use SCons GR for development and uncovered a couple routine problems with environment variables and one not-so-routine one concerning variables needed by G4. For CMT builds they're set in a requirements file. Nothing quite comparable is available for SCons builds but I found an alternative involving small changes to the packages SConsFiles and in G4Generator.

    I'm in the midst of splitting up tagCollector as suggested by Jim so that it will submit a batch job when there is real tag collecting to be done and will email the output to relmanlist.

previous

minutes index

next

J. Bogart, Last Modified: