Core Minutes 1/25/2011

• News: (Richard) Leon has agreed to be co-leader of C & A.

• ScienceTools: (Jim) will be making a new tag soon. Several packages updates from Johann for the gcc 4.4 compiler. The P6_V11_DIFFUSE irfs will go in. A couple other projects are in progress:

  1. Jean found an instance where the convolution produces an incorrect result and sent a test case. The fix for this is in the works and should be done soon.
  2. Johann is working on adding priors on fit parameters to the Likelihood calculations to address a request from Jan Conrad for the dwarf stacking paper.

  The first of these is likely to make it into the new tag; the second probably won't.

• FSSC: (Eric) is in the process of porting Science Tools v9r20. Otherwise, the main news is the continued fall-out from the security incident. (Tom S.) Everything is still offline except for static pages while it undergoes scans (which will not turn up anything in the Data Server since it's written in Perl but the scans are designed for php) and is modified to meet new requirements. This could take a couple weeks unless Tom can convince the powers that be that the changes they're requesting are not likely to decrease security risk.

  (Richard) It would be well if those trying to access the Data Server were informed that weekly data is still available. (Tom S.) will see if this can be done. He doesn't have control over the main Fermi page (which is where data server pages are currently being directed).

• Storage: (Richard) The switch-over to the new user disk was smooth; no complaints to date.

  Things are less rosy elsewhere. A couple strategies which might have increased the disk available on our existing data servers have proved infeasible. The Computer Center has located about 100 Tbytes which will keep us going for a while. A new vendor, DDN, is being investigated.

• Pass8: (Tracy) The latest GR tag, v18r8p5, is being used to generate all requested dataset varieties. There are still a couple problems with this release:

  • Cosmic ray tracking does not play nicely with the "regular" tracking, hence is turned off for the time being. The difficulty is understood; a fix will be incorporated into a future release. However, it is perhaps just as well to make these first runs without cosmic ray tracking so that the effects of the new tree-based patrec are easier to see.
  • There is a bug (or perhaps a couple bugs) in AcdRecon2. It crashes on Linux and doesn't run at all on Windows. Tracy has a possible work-around for the Windows problem; Eric C. is making progress on Linux. It seems likely this will be sorted out soon. In the meantime, we're running with just the original AcdRecon.

  Reconstruction time is longer than with the old reconstruction, not a surprise. It can probably be reduced significantly by concentrating on finding especially anomalous, long-running events at an early stage and cutting off processing on them.

  (Leon) has been looking into truncations in the tracker. No breakthroughs yet, but he believes it is a worthwhile place to expend some effort. We might want to consider changing the configuration settings to better maximize useful hit information.

• Pass 7.4 (Tom G.) A new round of reprocessing is just starting: rebuilding FITS files using the Pass 7.4 classification. The first goal is to reprocess the first two years of science data (Aug 2008 through July 2010) and will likely be followed by a 'catch up' round after evaluation. The first 2 years consists of nearly 11,000 runs.

• Lots of cores (Richard) Our allocation at Lyon somewhat mysteriously doubled to 1200. And we've been getting up to 2500 at SLAC, well above our official allocation. [Whee! ed.]

• Documentation (Heather) Some of the pages available from the pre-SCons workbook, such as one describing how to run Doxygen, have been dropped from the current version and need to be reinstated. Chuck knows about this but has work to finish up on ScienceTools first.

• Core talk:

  • u15 (Heather) Kim has cleaned it up somewhat, but it would still be a good idea to direct temp files currently being written there elsewhere; Tom G. suggests u14. This is most likely just a matter of finding the right script or database entry to change.

  • No Windows builds (Heather) It is still the case that neither RM is able to successfully start up a Windows build.

  • New Windows box (Heather) Last lsf issue has been resolved: the box is specifiable to lfs via a named resource. it's ready for us to try out.

  • Snow Leopard Tom S. has tried various approaches to deal with the optimizers/f2c issue, such as attempting to use FSSC-built externals, to no avail. Jim will see about working around the trouble spots for this OS only with #ifdefs, just so we can get the RM Snow Leopard builds going and start distributing them.

  • Cut-off (Heather) received no angry responses — or responses of any kind! — to her announcement that CMT ScienceTools builds would end on Feb. 1, so we will proceed as planned. Status of issues mentioned previously as necessary or desirable to resolve is

    • Doxygen generation. This is now being done. It's not yet accessible from the RMII web pages, however. Karen will take care of that when she has a chance.
    • Red-dot display. Tom S. outlined for Kim where the relevant information is to be found in the database.
    • SCons tag collector. Joanne updated the file ScienceTools-scons/packageList.txt, not essential but nice to have since it provides an easy to access record of which tags are in which HEAD or tagged build. Jim is ready to use the SCons tagCollector script at the next opportunity.

  • SCons and TMine (Joanne) The new TMine external, provisionally to be called TMineExt, is close to ready for its unveiling. It builds on both Linux and Windows and TMineApp appears to run properly. Starting with a modified GR LATEST, with the rootExtensions/TMine package removed, she was able to build GlastClassify by adding TMineExt to the list of externals.

    Along the way she made minor enhancements in a couple of related areas:

    • Added special interpretation of --variant SCons option when value is NONE. In this case libraries are installed directly in lib rather than in a subdirectory with a name like redhat4-i686-32bit-gcc34-Debug, executables go directly in exe, etc.
    • Added a new item, extra SCons options to GoGui's Options menu. It gives the user a chance to specify arbitray additional options to SCons in addition to those already settable from GoGui by other means.

  • Passwordless login was supposed to make life easier, but that's not evident so far. It works — that is, those in a special new afs group Tom G. has created can log into the glastrm account from their personal unix account without supplying a password — but, since such a log-in keeps the original username's afs token rther than getting one for glastrm, those logged in this way don't necessarily have the same access glastrm has to directories in afs space. In particular, those logged in this way cannot write to the GLAST_EXT area or to the glastrm .ssh subdirectory. Work-arounds, probably to involve extra maintainer userid's for those concerned, are in the works.

  • SCons and Windows miscellany (Heather) has not been able to make a rootmap for the G4Generator when built with the VS 2008 compiler on her Windows 7 box, nor has Joanne using the same compiler on XP. Heather can make the rootmap when VS 2003 does the building.

  • SCons meeting Back to the usual time: this week Wednesday at 10:30.

previous

minutes index

next

J. Bogart, Last Modified: