LHCb Core Software Meeting

Date and Location

30-Mar-2011

Attendees

Subjects

News

• From the LIM meeting
  • Atlas wants to test a new version of gcc and have the possibility to use their own private patches.
    It is probably fine for testing only, until they want to have the SFT support for their patches (we want well documented and proved arguments). (more infos)
  • Discussion between Atlas and SFT about improvements to Savannah.
    It's a pity that we were not involved in the discussion, but we can give feedback too. A poll for requirements for a tracking system to replace Savannah will be circulated around Easter with the aim to get a proposal and a prototype for June (more infos, SPI-33 on JIRA)
  • Some obsolete/unused interface packages will be removed (more infos)
    • vdt
    • EDGRLS
    • slc4_compat
    • MathLib
    • cgsigsoap moreover, roofit will be fixed to point to the current version of ROOT.

Round Table

Marco Cl.

• I've been working on CMake to clean up the configuration proposed by Pere in Gaudi. I'm more and more convinced that it will be much better than CMT even from the point of view of the readability of the configuration.
• Created a branch for Gaudi v22. The only differences between the trunk and the branch are two changes: auditors bug-fix and VC9 fixes.
• It has been suggested again to have the -g option in the optimized build. I had a look to its effects on Gaudi: the average size of the libraries is increased by a factor 5 (or 3 in Hubert's tests). The runtime memory is not affected by the presence of the symbols. In any case, the change in the size will cause many problems on the grid (size of shared areas), cvmfs (larger transfer) and even on the Online (NFS overload) where the feature was needed to be able to have some meaningful traceback in the case of failures in the farm.
  We agreed to keep the current configuration, but have a simple way to turn on the -g flag and build locally to analyze a core dump (we need to ensure that we have one in case of a crash in the farm).
  Markus found on the web that it is possible to have the debug symbols in separate (optional) files (more infos on stackoverflow and on strip manpage). With that we can package the debug symbols independently. To be investigated further.

Paloma

Karol

Hubert

Rob

Marco Ca.

Markus

Gloria

-- MarcoClemencic - 30-Mar-2011