LHCb Core Software Meeting

Date and Location

18 April 2007
10:30 - 12:30
CERN (2-R-030)

Attendees

Florence, Gloria, Marco Ca., Marco Cl. (minutes), Markus, Matt, Philippe, Thomas

Apologies

Andres

Subjects

Specific Topics

A TWiki-based Structured FAQ System

Ben Lloyd presented the new FAQ system that will be used for GENSER. It is a well structured set of TWiki pages that allow easy access to questions and answers and it is generic enough that can be adapted to other projects.

The top page features a Google custom search text box. The reason for it is that the default TWiki search result is not good enough (the Google one is showing the hierarchy of the pages).

The FAQs are divided by groups (they use generators while we can use projects), and for each group you can have categories and sub categories.

The system looks nice and we agreed that we should try to use it. Ben volunteered to set up the initial structure (with the Google search box) that we will customize. (See LHCb.FAQ)

News

  • LCG 51 has been release on Monday
    No need for an urgent build of Gaudi (Marco Cl. will prepare and test the code to use the LFC capability of CORAL before we move).
  • (Philippe) No Architects Forum tomorrow, next one in two weeks (urgent matters can be arranged via email).

Thomas would like to have a version of Panoramix using all the fixes in ROOT. Marco Cl. suggested to use the nightlies and will help Thomas to use them.

Software Releases

LHCb, Boole, Brunel (Marco Ca.)

  • Marco Ca. experienced a problem with the SEAL plug-in manager on his Windows box. The problem is that, during summer (summer time), the time of a file on AFS is different between Linux and windows if you use a recent version of the AFS client.
    A solution is to copy the files locally or use an old version of the windows AFS client.
  • A bug in the stripping has been discovered: when reading an ETC, the tool used by EventClockSvc to generate the ODIN structure checks if the entry exists already, and actually it exists in the rDST, which is then opened.
    We already had a similar problem with "/Event/Rec", which was solved adding as first algorithm in Brunel the NodeKiller. Unfortunately it cannot work for ODIN because the EventClockSvc is called before the first algorithm.
    After some discussion we found 2 solutions that should work without touching the C++ (pure job options):
    • call the NodeKiller on "Rec" and "DAQ/ODIN" just before writing out the ETC event
    • write DAQ/ODIN in the ETC Marco Ca. will try them.
  • Adlene implemented the AnalyticFieldSvc, which should provide the value of the magnetic field from an analytic parametrization. Unfortunately it was not implemented as an IMagneticFieldSvc, but now it has been modified to follow LHCb policies. The only draw back is that the name "MagneticFieldSvc" is hard-coded in many places.
    Markus suggested to use the option
       ApplicationMgr.SvcOptMapping +=  { "AnalyticFieldSvc/MagneticFieldSvc" };
       
    that will allow to ask for "MagneticFieldSvc" and get and instance of "AnalyticFieldSvc".

Gauss (Gloria)

  • Waiting for feedback about the changed at the generator level to speed up generation (DC04)
  • Few 1000s events generated with the latest version to check the physics and a problem that seems to affect the new GEANT4 when using default e.m. physics (much more stuck particles, but it doesn't seem to affect us)
  • Ported Gauss to latest Gaudi. Long debug session with Marco Cl. to sort out the problems:
    • a factory was in a link library, thus creating conflicts
    • the ToolsSvc was modified to use the INamedInterface class which was broken (see the post on savannah), the result was a strange segmentation faults

Round Table

Florence
  • Tomorrow there is a LIM meeting (librarians), do we need anything to be added to LCG?
    • Markus: C/C++ version of xmlrpc (+package in LCG_Interfaces) is needed for OnLine
  • CMT v1r20p20070208 has a problem that prevents to build Panoramix, so the default has to be changed back to v1r19 (in the mean time Florence will look into Panoramix requirements to see what's wrong)
Thomas
  • What will happen to the CondDB disconnection message? removed, changed?
    It has already been changed, but it didn't end up in the latest release. It will go in the next one.
  • Ulrich provided a first version of microDST for testing and it seems to work well (~ a factor 25 of reduction in size)
    We talked a bit about the way the option RootOnTES is used (or not used) in our algorithms, and it seems reasonable that it should be handled at the level of GaudiAlg (see the patch posted by Chris in the afternoon).
  • Working some timing measurements
Marco Cl.
prepared an extension to CondDBAccessSvc that copies the SQLite files to a local directory (the main use case is the event filter farm, where the SQLite files need to be copied from NFS to a local ram disk)

-- MarcoClemencic - 18 Apr 2007

Edit | Attach | Watch | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r4 - 2010-06-11 - PeterJones
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    LHCb All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback