Core Minutes 4/16/2013

• ScienceTools: (Jim) Just one item has been added to the Science Tools Development Notes page concerning handling of diffuse sources. Details may be found in JIRA LK-86

• FSSC: (Alex) They're hung up with a run-time problem (memory leak) for one of the test programs on Fedora 18 32-bit systems. It seems to be associated with a particular version of the c standard library. The release has been validated for all other OSes. (Tom S.) This is one of the same programs that fails on Windows.

• Pass 7 Reprocessing (Tracy) Philippe had discovered a discrepancy in the way the cal centroid position correction was applied — or not. As of yesterday runs were started with a version which, we believe, has fixed the problem.

  (Richard) What about sys tests? (Heather) We're trying to compare behavior of overlays with a release back when it was all working: v17r35p23. Comparison of CAL x,y,z position was failing because those values are no longer saved. This points up a class of perennial problems with sys tests. The responsible person has changed several times. Knowledge of sys test behavior and quirks is typically only partially transmitted from one maintainer to the next.

  It seems a fire has been lit under Johann; he is looking into the current set of problems.

  (Heather) It might be better not to use cint to run the sys tests. It's lenient about various kinds of errors and warnings. (Tracy) Could fill output histogrmas from userAlg. (Heather) would prefer to work directly with the root files and not introduce gaudi into the procedure.

  (Richard) any news on ACD calibrations and whether we have to reprocess everything since September? (Leon, Tracy) Not yet known; people are still working on it.

• Hardware (Richard) is hoping to purchase a high density server for $90k if we can scrape together the cash. If it worked out well we would get more later.

  Money for new Macs is difficult to find, especially since we will need new Oracle servers.

• Pass 8 (Tracy) Progress has been made on hit errors, but it's clear this is not going to get fully sorted out any time soon; probably will take at least a month. And the gains (in resolution at high energy) are not that significant since resolution there is already very good. So the new plan is to let this go for now and hopefully be ready for reprocessing in about 3 weeks. Still to be addressed before then:

  • Alignment in pattern recognition (Leon) has learned a lot but has yet to achieve what he set out to do. Putting alignment into vector links is apparently not sufficient. Fallback is to introduce alignment into MC.
  • Performance (timing) (Tracy) Changes were put in to eliminate tails. They had the desired effect but also hurt efficiency at high energies. The problem may have been uncovered

  (Tracy) We hope to have resolved the issues above and to be ready for reprocessing in about 3 weks. (Tom G.) will need a few dedicated days minimum (assuming everything goes smoothly) to get ready to pull the switch.

• Core talk:

  • fermi-ts; Windows and SCons (Heather) Leon, Tracy and Bill are all up and running there. She hasn't heard anything recently from Bill and is hopeful that means things are going smoothly. (Tracy) Bill ran into the usual kind of problems associated with unfamiliarity with a new system, but got past them. (Joanne) He likes to work with userApp. Up till now two special solution files can be created by SCons: one called all.sln, including projects for all libraries and executables, another called allGleam.sln, including Gleam executable and all libraries. To make things convenient for Bill there really should be an alluserAlg.sln but for the time being he's getting along with all.sln. I have since made a new tag of SConsFiles (SConsFiles-01-10-05) which will make alluserAlg.sln. (Heather) will pick it up for next release. (Joanne) the other thing Bill ran into had to do with installed files, something everyone has to get used to when moving from CMT to SCons.

    (Leon) had been somewhat impatient with the overhead of any SCons operation from GoGui (about a minute and a half on his machine) but this decreases to a much more tolerable 10 seconds when using a supersede directory.

  • Proposed new externals containers (Joanne) spent some time last week mapping out how this might be done. See notes in Confluence, especially the section labeled A Snag. After taking a look at this page, Johann (a key player in that it was his wish to get the psf into GR, which is what set off this whole line of inquiry) is of the opinion we should hold off:

    Guys, are you sure this is worth the trouble? I needed the PSF (and I do not think we will ever need anything else than the psf) in part because Bill needed to simplify the covariance and hit error studies, and dividing the cov values by the psf from the irfs removes a great deal of the cosTheta dependency for instance.

    We still need to prove that his effort bears fruits (the hit errors are the main issue to solve in terms of coding in the current P8 iterations, together with running time and recon sizes).... so while it is a good idea to brainstorm on possible modifications to our codebase architecture, I would not rush to concrete implementations at all costs.

    (Joanne) agrees. The task looks doable without an excessive amount of coding and reorganization, but would take lots of careful testing to make sure none of the many modes of operation gets broken. (Heather) suspects we will have to do something along these lines at some point — if not now, then later.

  • RM (Tom S.) Windows: no information yet from Tony concerning the Jenkins problem (build runs correctly when invoked interactively but not from Jenkins). Tom has sample code which will eliminate the error assertion dialogs. He needs to integrate it into the test programs, possibly one at a time in some cases.

    Mac: Jenkins LATEST builds have been fine. It's time to move HEAD and tagged release over there as well.

previous

minutes index

next

J. Bogart, Last Modified: