LHCb Core Software Meeting

Date and Location

1 August 2007
10:30 - 12:00
CERN (2-R-030)


Gloria, Hubert, Marco Cl. (minutes), Patrick, Philippe, Thomas



  • LCG 53a has been released (not announced yet)
  • No progress about the proposal of splitting LGC externals in 2, a group with fixed versions and one with always the latest one. We shall have to implement the separation on our side.

Software Releases

Gaudi (Hubert, Marco Cl.)

  • Hubert is working on the installation of Python scripts and modules in the InstallArea.
  • Marco Cl., following the proposal to have a cbrt (cubic root) function taken from LHCbMath in Gaudi, introduced a small header file providing only the platform independent definition of that function (GaudiKernel/cbrt.h) without the namespace that is used in LHCbMath/Power.h. The LHCb code should be adapted accordingly.

Gauss (Gloria)

  • Testing Sajan's patch to take into account the alignment conditions in simulation (using a few lines of options to open the Velo provided by Marco Cl. and available on the CondDB How-To).
  • Working on a patch to skip the simulation of empty events (see previous meeting minutes).

HLT/DaVinci (Patrick)

  • New version of PHYS is ready , HLT (v1r8) is in the pipe-line. Juan will prepare DaVinci v19r4 as soon as the others are ready (should be already tagged).


  • Progress on the crash of stripping jobs?
  • Status of archival of DC04 versions of DaVinci?
    v12r14 has to be kept a little longer, v12r16 and v12r17 will be archived, users must move to v12r18.

We had some discussion about the problems with the documentation. The current LHCb Computing home-page is ugly, obsolete and not helpful. Patrick suggests that we migrate as much as possible our pages to TWiki. Gloria thinks that the main page must be kept as HTML.

Panoramix (Thomas)

  • Waiting for the upstream projects.

Release and Deployment (Hubert)

  • Problems building PHYS on Win32.
    Patrick implemented a trick to try to compile and link the code for the solutions of the tutorial with every release of PHYS. It works on Linux, but on Win32 it seems that the default link options do not allow to link. We agree that a proper library should be built, but should not affect the installation. The best of these possibilities have to be identified and applied:
    • link library, but not used in link statements
    • component library, without rootmap files
    • private (link or component) library (not exposed)

Round Table

Marco Cl.
  • Found a bug in GaudiPython: the simple instruction print gaudimodule.gbl.DataObject() throws an exception because the class stringstream is not known (probably just due to missing dictionary)
  • Found a bug in DetCond: the new feature that allows to avoid to connect to the database until it is used, was not introduced correctly, so the thread that tries to disconnect does not check if the connection was ever made, so if you do not connect to the DB or you do not disable the extra threads, you will get an abort after 10min.
    (Note: a work-around has been just sent to lhcb-core-soft)
  • Before the introduction of UpdateManagerSvc and GeometryInfoPlus, it was possible to clean up the Detector Transient Store to re-load it (feature mainly used in Panoramix). To re-enable that functionality, the UMS has been instrumented with a purge member function that reset the pointers and the dependencies for objects that can be re-loaded. GeometryInfoPlus have to be modified in order to unregister when it is deleted (the UMS cannot know that it can be regenerated).
Tried to run the scripts to collect statistics on the production, but the query to find the production id of a Gauss v25r10 does not work. Asked Carmine and Marianne.
Received 4 files with one event each from Mitesh (B inclusive)
  • How to find the original file?
    No way if they do not point to the original file, we should use ETCs or DSTs. It would be quite useful to have a tool that is able to find the files required by an ETC and produce a DST from the ETC.
  • When going from one of this files to the following, the event read is always the same. They contain different events, but only the one from the first file is visible.
    After the meeting Marco Cl. ran DaVinci through the debugger and discovered that those files have all the same UUID, so POOL keeps using the first one. The reason for it is that the files were generated with the same name and POOL "remembers" in test_catalog.xml which was the GUID associated with a filename (thanks to Patrick for the suggestion). To avoid the problem, it is enough to change the name or remove test_catalog.xml between two executions.
Is the conversion between enum and string already available from G.O.D.? Marco Cl. will check (the patch should be already applied).
Since the migration to InstallArea/CMTPROJECTPATH, it is not possible to setup the environment for an application using the setup.(c)sh script that is in the project directory from another directory: CMT needs to be called from inside the project. A bug report has been sumbitted to CMT.

-- MarcoClemencic - 01 Aug 2007

Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2007-08-01 - 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