LHCb Core Software Meeting

Date and Location

10:30 - 12:40
CERN (2-R-030)


Gloria, Hubert, Illya, Joel, Karol, Marco Ca., Marco Cl. (minutes), Philippe, Rob, Silvia



  • Marco Ca.: Marco Cl. will be the LHCb representative in the Architects Forum
  • Hubert: LCGCMT 56d has been released (for Atlas)
  • Marco Ca.: new version of ROOT due in the next weeks. We have to ask Stefan for an LCGCMT (58) base on it so that we can switch in January.

Round Table

Marco Cl.

  • Completed the port to Qt4 of the CondDBBrowser. It is a complete rewrite.
  • During the AA Meeting, Rene reported of the IO performance improvements that can be achieved tuning the ROOT files (up to a factor 10). The problem with LHCb is that we have files with ~40 TTree instances with on top-level branch in each of them, while the optimizations can only be used on files with a single TTree (even if with many top-level branches). We should address the issue (probably) in POOL.
  • The problem of the return core 134 has been solved (task #12450). Some sites still have problems, but it may be because they set by default MALLOC_CHECK_ (DIRAC should unset it)
    • Marco Ca.: we should ask linux.support to follow up the issue about thread safety of MALLOC_CHECK_
  • The segmentation fault during the finalization of Gauss (bug #60331) is understood as problem in the order of finalization (Gloria reported that it has always been the case with Geant4/GiGa). Marco Cl. will investigate how to solve the problem.
  • During a coffee discussion with Hubert and Gloria, we thought of pushing part of GaudiPolicy to a package in LCGCMT. Recently the same kind of issue arised in Atlas. In the afternoon Hubert and Marco Cl. will have a meeting with Stefan and somebody from Atlas to discuss the idea.
  • The problem with the access to Oracle (CondDB) from a laptop (ubuntu) is that libexpat.so.0 is not present (too old). Hubert will add it to the project Compat.


  • GENSER will put up a pre-release of the new generators in the nightlies


  • Linux certification meeting last week: SLC5 has been declared certified (no show-stopper)
    • one of the show-stopper was the problem with binutils we reported, for which a fixed rpm is available and will go in 5.5 (we are currently at 5.4)

Philippe pointed out that those issues should be now discussed in the IT-SRM.


  • Geant4 finally built on Windows
    • problem with a change of behavior in the way libraries are built on Windows (GaudiPolicy)
    • problem with computation of dependencies (command line too long), avoided using QUICK=1 (that doesn't work in parallel builds)
  • LbScripts v4r4 have been a long time in dev state. Can we switch to production? Wait until January (we can even jump directly to v4r5)
  • A new terminal server is available: cerntslhcb03, with independent disks for releases, nightlies and users.


  • Patch release of Gauss v38 released to fix a problem with a change in the path of GENSER.


  • XML Summary Service does not work in DaVinci (the DIRAC guys want to use it)
  • There is a segmentation fault in an event model class with a pointer that should not be dereferenced
    • Marco Cl. suggested to use an ASSERT disabled in the optimized builds.
  • For the luminosity studies we need to change the way FSRs are handled: always read and write

Philippe reminded that we discussed the usage of application specific FSRs, but nothing happened yet (lack of manpower).

Marco Ca.

  • The mechanism to check if the ONLINE partition of the CondDB is valid must be put in place ASAP.
    • Marco Cl. implemented the algorithm to validate the ONLINE partition a long time ago, but it requires that a special condition is added to from PVSS (at least) at each end of run (to be checked with Clara).
  • A technical student will come next year to work on the interfacing of the tag collector with the nightlies (in collaboration with a technical student in SFT).
  • Marco Ca. had a meeting with Thomas to decide on the organization of the Software Week
    • Philippe: we need a joint meeting between core-software and the production.
  • Marco Cl. and Hubert will start an initial migration of LHCb packages to SVN. The migration will be adiabatic, package by package, initially on a voluntary base.


  • Marco Ca. will be Software Coordinator. Philippe will be deputy for grid-related issues.
  • A reorganization took place in IT. The main issue is the change in the lead/organization of the DB group.
  • We are still running the slc4_ia32 binaries on the grid. We have to start using the "best available" for the platform the job is run on (as discussed some time ago).

-- MarcoClemencic - 09-Dec-2009

Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2009-12-09 - MarcoClemencic
    • 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