AF meeting 26 July 2007

Oliver Keeble, Dirk Duellmann, Pere Mato, Marco Cattaneo, Vincenzo Innocente, Stefan Roiser, David Quarrie, Fons Rademakers

version 1 (27/07/07 @ 16:00)

The selection board for the IT post in the area of POOL/CORAL is scheduled for end of August, before CHEP.

Follow-up MySQL client installation issue reported by ATLAS concerning a hardwired AFS address in the installation (David, Stefan)
Nothing new on this action.
Questionnaire about external packages in use (Experiments)

CMS is working on it, they will be comeback soon with the list of packages.
LHCb has not provided yet the final version. The final answer will also come soon.

MacOS (intel). Stefan reported that this platform is being build in the "dev1" slot. It looks in pretty good shape. LHCb has noticed that GSL is not being built for this platform. It is rather strange that ROOT did not find this problem first, because GSL is needed to building the library mathmore.
Dirk said that we should look at the compilation/link flags very carefully and eventually standardize for all AA projects (set them in LCG_Policy). Dirk claims that ROOT needs to build different the Python extension module (PYROOT) as a bundle instead of dylib otherwise it does not work for the intel platform. This needs to be clarify between ROOT and COOL. There is a long savannah report on this issue.
Dirk reported on the ongoing campaign to make the nightlies greener. This is progressing slowly. In same special cases to do with inconsistencies of the C++ standard library he only managed to get rid of the warnings by adding flags to suppress then in the compilation.
Windows. The builds for RELAX has developed some problems.
Gcc 4.1. There is currently a problem with GCCXML that does not detect the proper version of gcc4.1 and therefore the proper headers files are not located. Pere asked if this could be forced with a flag somehow.

Fons said that there are a number of problems with 5.16 and perhaps a new development version will be done end of August that should solve most of them. This is fine for the experiments and they can wait until the end of August to start integrating this version. The main issue from their point of view is that this development version would require to be patched in case of errors discovered during the integration.
The conclusion was to wait for the end-august development release 5.17 and build the AA software stack on this one. Eventually, a CVS branch for patches will have to be introduced as if it was a production version.

In parallel we need to make sure that the next version of ROOT is build with the new version of GCCXML.

LCG_53 was finally released, on request of ATLAS, with the version of POOL reverted to 2.5.4 (identical to 2.5.3). A new configuration LCG_53a was created with the most recent version of POOL 2.6.0. The difference is in the event collection area, which should not be affecting other experiments. A new patch in the collections is now available (tagged as preview) and being used in the nightlies. David thanked everybody for the good response for this late change in the plans.
Preparation of LCG_53b:

Stefan suggested that we need to use the new version of CLHEP 1.9.3.1 that removes the HepMC and HepPDT packages. We need to double check with ATLAS from where they get HepPDT.
A new patched version of GCCXML will be needed to cope with gcc 4.1
New version of RELAX will be needed to fix problems with the generated rootmap files.
It was agreed to build completely LCG_53b with the gcc4.1 and provide only the MacOSX(intel) binaries and not the PowerPC ones. In addition, the SLC3 platform will be provided only with the optimized version and remove the debug one.

David mentioned that the CMT version should be 20070720. The old version has a problem with calculating packages dependencies. In addition, the binaries for all platforms, in particular windows and MacOS(intel) need to be produced. Grigori Rybkinis currently taken care of the CMT maintenance. Announcements of CMT new releases is an issue.

They have reported a problem LCG_53 with a segmentation fault somewhere in Python. Investigating.
For the Online project they require the installation of PyQt for Qt3 and Qt4. The Qt3 version is mainly needed for PVSS compatibility.

They are releasing CMSSW 1.6. The next version 1.7 is scheduled by September. They would like to start including ROOT 5.16 (5.17) if the September one. No special request.

Reported that Geant4 8.x is generating ions as particles with special PDT ID numbers, which are not handled by HepPDT. In order to fix the problem they had to use their own version of HepPDT to be able to handle these PIDs. This needs to be followed and make sure that the changes are put back in the official version.
In the context of the plugin service of ROOT/Reflex, they would like to have the utility 'genconf' that generated the rootmaps for the plugins modified to avoid duplications of entries in the rootmap files. The improved version should be part of ROOT/Reflex and not private in Gaudi or Atlas in order to ease later maintenance and evolution. Axel is currently looking into it.