Core Minutes 26 November 2013

Core Minutes 11/26/2013

Present: Joe Asercion, Joanne Bogart, Philippe Bruel, Jim Chiang, Richard Dubois, Warren Focke, Heather Kelly, David Landriu, Leon Rochester, Tom Stephens

Announcements: (Richard) The IFC met last week. The computing budget was confirmed, a good thing.
ScienceTools: (Jim) has made updates to optimizers and Likelihood to ameliorate the effects of problems Jean has encountered with the Minuit2 optimizer. See Science Tools Development Notes for details.
Pass8 Reprocessing: (Richard for Tom G.) P8 reprocessing is done through June, 2013. L & EO will be coming soon.
Pass8 (Leon) We’ve long known that running MC using a non-ideal alignment (in both sim and recon) gives a psf that is broader than the one you get from using ideal geometry, especially at high energy. I think this is because the granularity of the detector doesn’t allow for a reversible transformation.

Up to now we’ve always used ideal geometry in simulation, and we’ve seen that the psf is broader in data than in MC. So, urged on by Philippe, over the last week or so we finally ran some MC jobs, first using the standard alignment calibration in sim and recon, and then using different calibrations for sim and recon. In each case we saw improvement, and the latter config now gives a PSF that agrees with data at high energy.

So one issue is resolved. The next one is to figure out why the ToT variables don’t agree between MC and data.
Core talk:
- GR 20-09-06 (Heather) We're not quite ready for the release; work is still ongoing, especially concerning choice of calibrations for MC. See JIRAs LPATE-168 and LPATE-171 for the full story. (Philippe) There is no need to rush to make the release because everything we're looking at can be done just with job options. We're considering using L1current for all but ACD, which would stick with ideal. (Leon) Or, maybe better, make a new flavor corresponding to L1current calibrations for a particular date so that different MC runs (with different start times) will always use the same set of calibrations.
- Sys tests (Heather) tried running them on the HEAD of GR and, up to a point, everything was fine. They ran without error. The Python script which posts the results also was happy — but didn't actually do anything! Next she will try rerunning on an old GR where everything worked before in the hope that, whichever way it turns out, it will provide some clues.
- Running on the GRID (Heather) Michael Kuss is attempting to install our software on CVMFS (Cern Virtual Machine File System). Anything intalled there has to be world-readable. The files he needs, all in afs space, are in fact readable by authorized users at SLAC because of the way acls are set, but copies are not world-readable in all cases because nfs permissions are sometimes more restrictive. He can of course copy them and then change permissions on the copies, but so far, no one has come up with a reason why they couldn't all be world-readable to begin with, which would simplify his task.
- Mountain Lion (Tom S.) has managed to build the two problem externals of last week, Xerces (turned out to be 64-bit issue addressed previously at FSSC) and CLHEP (known problem requiring small source changes). He's beginning to work on Python, the only remaining unbuilt external. It would be good to settle on a common version for as many OSes as possible (up till now Mac has lagged behind Linux). (Heather) is thinking about upgrading everywhere to 2.7.3.
  
  (Tom S.) is still unable to reach the CVS repository from our new Mac (and only from there; from all other machines access is not a problem). This issue has been passed on to unix-admin. (Heather) will send a note to Yemi since he has an interest in getting us off the old macs.
  
  Whither liblzma (Heather) Luca Latronica ran into yet another mac problem, this one on Snow Leopard. liblzma (required now because of the recent changes to ROOT compression algorithm) couldn't be found. The location of this library (and certain others) on a system depends on what repository is used. FSSC uses Macports and Luca was using something else. He got around the problem by rebuilding ROOT from scratch on his machine. (Joe) The usual recommended fix in this situation is to use sym links.

minutes index

J. Bogart, Last Modified: