Core Minutes 10/18/2011

• ScienceTools: (Jim) Likelihood-17-18-00 contains an alpha implementatoin of the energy dispersion calculation in BinnedLikelihood. Jim is running tests now. See Science Tools Development Notes for more information.

• FSSC: (Eric W.) The updated gtsrcmaps fixes are in. Dave Davis is working on a schedule to get a new ST out before the next round of proposals (sometime this fall).

• Pass7: to reprocess or not to reprocess: There has been a disqueiting internal report from Art who was looking at back section events. Art's analysis reveals strange things and Leon is hoping it is a bug or pilot error. In other news, still collecting calbrations updates, which is also useful for P8. There has been discussion of developing a schedule of updating calibrations to prevent getting behind in the future. This would include those calibrations that are deemed "fine". The thinking is that we should do updates once a year miniumum to stay fresh on the mechanics of generating the calibrations. For example, ToT gain has not been touched since LE&O but since it is elaborate and the expertise is fleeing, we should go through the motions. Otherwise, no real forward progress on how we decide whether to move forward with the P7 calibration runs, but a decision needs to be made soon.

• GR/Pass7 Reprocessing (Tom G.) Tom is recrafting the reprocessing task which will take another week or two. Richard mentioned there has been discussion with the P8 guys to reprocess a few days of data.

• GR/Pass8 (Tracy) Studies are in progress; Bill is looking at tracking details. There is a reported oddity in the truncation code that Leon is tracking down. The issue is somewhere between digitization, recon and heprep. It is complicated code and Leon is trying to insert comments as he goes.

• Other GR News (Heather) Michael Kuss and Leon had reported trouble reading back in recon ROOT files a couple of weeks ago. Heather has been looking into it. It seems the problem is specific to the TKR recon objects where we manage the memory allocation ourselves. This only affects reading, where it appears that when we over-run the default maximum size of the arrays used to store the TKR objects, we then resize (deallocate) to maintain a sensible standard amount of memory available for the next event. When reading another event that utilizes more objects than the default maximum we run into trouble since ROOT automatically handles the reading for us, and when we deallocate the memory array, the objects are still techically available since they haven't been completely eradicated. So ROOT fills those spots, but then we get an exception when trying to access those TKR objects that are sitting in deallocated memory. I think I see some ways around this and will be committed a patch soon. If this is all correct, this affects P7 as well as P8. Leon asked why we haven't seen this before. Heather replied that not many people read recon back in, and o see this, we would have to over-run the default maximum size of the arrays. In this case, there would need to be two events in one file with more than 500 clusters.

• Core talk:

  • Gaudi (Heather) To prepare for P7/L1proc's move to RHEL5, we need an version of Gaudi to built. It has been determined that compiling Gaudi v18r1 on RHEL5 is not feasible due to the age of the gaudi externals involved. The work of moving Gaudi v18r1 to more recent externals is not worth the effort. Instead the upgrades necessary for Gaudi v21r7 will be merged into the P7 branch and properly vetted using the system tests. In reality there should be no surprising differences between the two versions, so hopefully validation will go quickly and smoothly.

  • obf (Joanne) The Linux build of B3-0-0 is ok as far as Joanne knows. It will get a minor upgrade after JJ decides exactly how he wants to handle certain minor issues with include files. Next up is getting B3-0-0 to build via SCons.

  • GR and redhat 5 (Joanne) rdbGui is now working with the upgraded version of the Fox external. Meanwhile, M.E. has been moving right along with getting the other L1 processing applications to build via SCons against the SCons build of GR. She has made great process and is just about there.

  • ST and redhat 6 (Heather) Kim is reportably finished with building the externals and is in the process of reorganizating the directory structures to conform to our typical external organization. No detailed testing of the externals has been done yet. Hopefully we will be ready to try out a ST build this week.

  • SCons Windows RM (Tom S.) The windows builds are failing again, where it again appears to be an issue with the Qt configuration and LSF. Tom is looking into it. For the record, the Linux and Mac builds are fine.

• Announcements

  • SCons meeting this week at 10:30 on Wednesday.

• AOB

  • (Tom G.) Plans for sulky39 outage There is a planned four day outage intended to start this Friday which affects sulky39 (includes u35 and u52). This would mean the SCons builds would be unavailable and SCons RM would be unable to create new builds. SCCS has requested we respond by this morning to determine if they should attempt to find alternative power for sulky39 to keep us going. We agreed to make the request. Update: John Bartelt reports this should not be a problem.

previous

minutes index

next

J. Bogart, Last Modified: