LHCb Core Software Meeting

Date and Location

17 January 2007
10:30 - 11:45
CERN (2-R-030)

Attendees

Andres (phone), Florence, Hubert, Joel, Juan, Marco Ca., Marco Cl. (minutes), Markus, Matt, Olivier, Pere, Philippe, Thomas, Wouter Hulsbergen

Subjects

News

• Wouter joined LHCb for his 2nd year of fellowship.
• Bug-fix version of ROOT waiting to be tagged
• LCG 49 is almost ready, just waiting for ROOT
• LCG 50 will be released soon afterwards with the new version of COOL (2.0.0)

Software Releases

Gaudi (Hubert)

• v19r1 is ready, just waiting for LCG_49 and a working version of CMT (v1r18 cannot be used because the dependencies are not computed correctly, v1r19 does not work on Win32, v1r20 is not build for Win32)
• StatusCode check disabled by default
• Few changes in DataStreamTool
• Atlas made ApplicationMgr inherits from Service, which does not work. Fixed by restoring the old inheritance tree and implementing the functions needed by Atlas
• Vanya agreed to split the parser in smaller cpp files, so it doesn't need anymore 1GB of memory to compile.
• Few patches by Atlas on Handle, Auditor and MessageSvc

Marco Ca.:

• Testing using the nightly build of Tuesday.
• We should not wait for CMT for this first release. We need at least the linux build.

Florence: we can build with CMT v1r18, the dependencies will not work correctly, but it is not an issue when building from scratch.

We agreed to have a build with the most recent working version of CMT (v1r18 or any newer if it comes in time).

LHCb, Boole, Brunel (Marco Ca.)

• Brunel and Bool works with the Install Area (outputs are OK)
• The tests are extremely slow: we need an optimized build in the nightlies.

Pere: Once the machinery is working, we can add new platforms (it's just a matter of adding PCs). The priority, now, is to get Windows in the system.

Gauss (Florence for Gloria)

• Gauss not yet ported to new Gaudi (problems with plugins in Giga)
• Somebody has to validate the new generators (we are still using old versions and they are not mantained)
  It should be addressed to the physics planning group.

DaVinci (Juan)

• Released a couple o versions of DaVinci
  • Checking efficiency and performance of pre-selections
  • Finale release by the end of January
• DaVinci can be built with the new Gaudi. It runs, but it has to be tested.

Round Table

Juan

By mistake, archived a version of Phys used by a non-archived version of Panoramix (to be archived).

Marco Ca.

• On the 31st of January lxplus will switch to SLC4
• He is supposed to present LHCb plan on Tuesday.

Summary (quote from Marco's mail to David Quarrie):

• For development on slc4 against binaries existing in slc3 only, we will provide a script that:
  • Sets the CMTCONFIG to slc3_ia32_gcc323
  • Sets the PATH and LD_LIBRARY_PATH so as to pick up the gcc32 compiler with -m32 option
  • Builds slc3 executable cross-linked
    [sharing the Atlas slc3compat external area]
• For development against new releases, for which slc4 libraries exist, we will work in a pure slc4/gcc34 environment
• For the GRID
  • Ganga/Dirac submission works on SLC4
  • For productions, Dirac will be able to select the right binary depending on the node OS
  • For user analysis, we cannot select on which OS to run, so we should stick to SLC3 (users will have to compile their binaries for SLC3)

---

-- MarcoClemencic - 21 Jan 2007