LHCb Core Software Meeting

Date and Location

May, 28 2008
10:30 - 12:30
CERN (2-R-030)

Attendees

Gerhard (EVO), Hubert, Joel, Karol, Marco Ca., Marco Cl. (minutes), Markus, Patrick (EVO), Philippe, Stefan, Thomas

Apologies

Gloria

Subjects

Specific Topics

LHCb Nightly Builds

Karol presented the status of his work on the nightly build system for LHCb.
  • The system is still in development, with some problems, but basically works.
  • Based on LCG AA nightly build system, with improvements (that will be integrated in the LCG AA version).
  • The probably biggest problem is the checkout of the project (getpack instead of cvs and overrides on both a per package and per project level).
  • The configuration of the slots is based on XML (easily editable and extensible).

Questions and Comments

Marco Ca.
  • Can it be used? (modified the XML files, etc.)
    There is not yet a "common" instance (so far it has been run by Karol).
  • Currently running without the "keep going" option of cmt broadcast ("cmt broadcast- make"). It must be changed.
  • It should be useful to have automatic mails sent to the contact persons for a package if it does not compile.
Marco Cl.
We should split the development version from the production version of the system (as it was in LCG some time ago).
Patrick
  • Can we use the result of the builds? (e.g. changing CMTPROJECTPATH)
    We need the copies on AFS. The procedure is currently being tested.
  • What if the the version directory that getpack is creating for the head version is not consistent with the one in the Sys package?
    No problem, the Sys package is modified on the fly by the nightly system, so that it requires the version called head, and symbolic link is created.

Software Releases

Gaudi (Marco Cl.)

  • New state machine
    • Being tested with Boole and Brunel.
    • Beat and Markus say that it is mandatory to make Online work in production.
    • The changes will be committed to CVS as soon as we know more about Boole and Brunel.
    • Online people should test the changes to see if they fit their needs.
  • The release plan for Gaudi has to be modified:
    • Within 2-3 weeks we will have Gaudi v20r0
      • new state machine and de-SEALed LCG stack
    • The new developments will go in Gaudi v21r0
  • Recent commits by Charles (Leggett) in Gaudi HEAD to change the name of the libraries on OSX from .dylib to .so. It is not clear why it is needed by Atlas (no comment in the release notes), but the LCG policy it to always use .dylib, so we will go back.

Hubert:

  • Fixed a problem with the MessageSvc finalization (it was finalized after before the other services instead of after).
  • Introduced the template doc fragment to be able to process files through "cmt filter" to expand CMT macros in files. It is needed to be able to produce doxygen documentation on a single package.
  • Added a pattern for generation of binary python modules (.pyd files). Since they are "standard" libraries and Beat asked to be able to link against them, they are built as .so and a symbolic link is created (copy on Win32).
  • Implemented a path-cleaner script that removes invalid entries (empty/non-existent directories) from pathes (PATH, LD_LIBRARY_PATH, PYTHONPATH,...). It is sourced automatically by CMT after a setup.

LHCb, Boole, Brunel (Marco Ca.)

  • Preparing a new release for next week
    • Fixes for DaVinci.
    • Hoping for OT decoding (but Jan is away until Tuesday).
  • Review of the tracking code ongoing
    • Discovered and fixed unwanted conversions between double and float (causing discrepancies between Windows and Linux).

Marco Cl.:

  • Just received the new version of the IT geometry (for 2008), but it will need some work for the integration.
  • Magnetic Field conditions should be available in the next release of SQLDDDB (together with the IT changes).

Marco Ca. mentioned that we still have problems with overlaps in the beam pipe (nobody have the time to fix it). Thomas agrees that we need to find somebody to work on the parts of the geometry that are not included in a sub-detector.

HLT/DaVinci (Patrick)

  • Lot of changes in HLT. Waiting for next version of LHCb for a release.
  • Bug-fixes in DaVinci.

Philippe: Has it been understood the problem in CCRC appeared after the fix discussed last week? (retention dropped)
The problem is due to the fact that the option file that Patrick provided has been appended to the options in DIRAC instead of pre-pended.

Marco Ca.: Do we introduce Python options in the tutorials of next week?
The best solution seemed to have examples comparing old options and Python options in the Gaudi tutorial.

Online (Markus)

  • Online v4r9 has been released

Panoramix (Thomas)

  • Guy is working on a new version of OpenScientist.
  • Following the changes in the software stack.

Release and Deployment (Hubert)

  • Released online v4r9
  • Problems releasing Alignment on Win32 (everything OK on Linux). The maintainers should be fixing it.

Round Table

Marco Cl.
Wee need to decide when/how to move to Qt4.
The decision has already been made. Currently, the only "active" application that is based on Qt3 is the CondDB browser, which Marco Cl. will migrate as soon as possible (not this month). After the migration, we can ask LCG to provide interfaces for Qt4.
Phillippe
GRID middle-ware is being migrated to SLC5. gLite will be built on LCG externals too.
Stefan said that he is preparing a native SL5 (SLC5 is not yet available) build of the software stack (on a virtual machine).
Thomas
recently had problem with the usage of the EndEvent incident in Muon code. In Python, the control is given back to the user after the EndEvent incident has been triggered, which leads to some corruption if the user wants to interact with the transient store.
The actions (clean up) that are taken in EndEvent should probably be taken in BeginEvent (even the transient store is cleared at the beginning of the event). Marco Ca. will ask the developer to move the code from EndEvent ot BeginEvent. Marco Cl. can modify the way EndEvent is triggered to give the control to the user between the end of the TopAlgs and the EndEvent incident.

-- MarcoClemencic - 28 May 2008

Edit | Attach | Watch | Print version | History: r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r2 - 2008-05-28 - 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