Difference: CoreSoftwareMeetingMinutes20080312 (1 vs. 2)

Revision 22010-06-11 - PeterJones

Line: 1 to 1
 
META TOPICPARENT name="CoreSoftwareMeetingMinutes"

LHCb Core Software Meeting

Revision 12008-03-12 - MarcoClemencic

Line: 1 to 1
Added:
>
>
META TOPICPARENT name="CoreSoftwareMeetingMinutes"

LHCb Core Software Meeting

Date and Location

March 12, 2008
10:40 - 12:00
CERN (2-R-030)

Attendees

Gloria, Hubert, Joel, Juan, Marco Ca., Marco Cl. (minutes), Markus, Philippe

Apologies

Patrick, Thomas

Subjects

News

  • POOL and CORAL are migrating away from SEAL, COOL will follow (CORAL is currently broken in the nightlies).
  • Application Area meeting in the afternoon. Main topic: SPI build system.

Software Releases

Gaudi (Marco Cl., Hubert)

  • Marco Cl. working on a rewriting of the auditor interface (triggered by patch #1725).
  • Marco Ca.: on the gaudi-talk mailing list, Paolo Calafiura posted a message to ask to move some functionalities from GaudiCommon upstream to Algorithm.
  • The freeze for next version of Gaudi will be on the 3rd of April, and the release on the 10th.
  • Hubert committed the changes to genconf to allow the preload of an helper library (fixes a problem with Gauss and some special Geant4 classes). The changes consist of:
    • extension to genconf.cpp
    • one pattern to generate the helper library
    • one pattern to use the helper library (could come from a different package)

Marco Cl. suggested that we should have a way of generating the doxygen documentation of a single package, to be able to validate the doxygen comments before the release. Marco Ca. added that the generation of the whole documentation should be generated using CMT (action) instead of a stand-alone script.

Marco Cl. mentioned that GCC 4.3 has been released last week. The changes include some performance enhancements. Since we are not yet at 4.2, we could jump directly on 4.3.
Marco Ca. will mention it at the Architects Forum.

LHCb, Boole, Brunel (Marco Ca.)

  • The complete software stack based on Gaudi v19r8 has been released.
  • Brunel need to be re-released because of a job option file that was missing. The missing options are needed to avoid a link between the DST and the RDST.
  • Bug/feature in MDF selector: if the file cannot be opened, it puts an empty RAWEvent in the transient store instead of nothing. Markus will try to fix it.

Gauss (Gloria)

  • Gauss will be released, but needs the special database tag head-20080225-no-TT to produce a reasonable number of hits in all the detectors. The TT team is preparing a long term solution.
  • GENSER Meeting last week * Main topic was HepMC. The backward compatibility is non-trivial and there will be the need of ROOT developers. * Problem with the system of units. The system of units of HepMC is not guaranteed to be the same as ours (copy of CLHEP). A proposal is being drafted and will be circulated.

OnLine (Markus)

  • Implementing all the wanted features

Marco Ca. reported a problem with the override of the python version in Online.
It was needed because of a bug in LCG 54, but not it is fixed and the hack has to be removed.

DaVinci (Juan)

  • Phys and Analysis released
  • Doxygen documentation is being built
  • Sill problems with fastjet on Win32.
  • Problem in the persistency of Linkers in uDST. Olivier is helping. Anyway uDST are usable.

Joel could not find an option file that was present in the previous releases and is needed in production. Under investigation.

Panoramix (Thomas)

  • Panoramix v15r11 released (based on Gaudi v19r8).

Release and Deployment (Hubert)

  • Full stack built.

Round Table

Marco Ca.
  • Next week we will have a bunch of tutorials.
  • Problem with tbroadcast. It spawned too many processes, which causes too much swapping (then a reboot, by CERN watchdogs). Probably it happened because of the concurrent running of Doxygen. We need some automatic limit of the number of processes (e.g. number of cores). Marco Cl. will take a look.
Philippe
  • We need to define how the CondDB replica is chosen.
    Marco Cl. said that it is implemented and uses the environment variable LHCBPRODSITE. That environment variable will disappear in DIRAC3, so the issue have to be discussed with the DIRAC team.
  • Which is the status of the handling of CondDB tags in DIRAC?
    Marco Cl. discussed with Stuart and Joel. The baseline is to have a specific optimizer for the LFNs used for the tags. For offline partitions (DDDB, LHCBCOND, SIMCOND), the LFNs will be requested only if the tags are explicitly requested. For ONLINE, we can use the run number as fake tag, but is has to be automatically discovered from the input file LFN (possibly using the bookkeeping, but temporarily we can rely on the file-name convention).

-- MarcoClemencic - 12 Mar 2008

 
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