Librarian and Integrators meeting

Tuesday 1 Sept 2009, 11:00 → 12:00 Europe/Zurich

TBA

Stefan Roiser (CERN)

The bi-weekly meeting of the LHC librarians and LCG/AA projects

Attendees: Marco Clemencic, Bertrand Bellenot, Grigori Rybkin, Andreas Pfeiffer, Stefan Roiser, Andrea Valassi Excused: Emil Obreshkov News === - Stefan has been preparing instances of the opengrok source code cross referencing tool in a test instance http://cern.ch/test-lcg-opengrok and in a production instance http://cern.ch/project-lcg-source. Opengrok is a java application and the websites are hosted via the IT/JPS tomcat service connecting to an afs volume which is the data container. Both deployed instances contain the source code of all AA projects for the configurations LCG 56-56c. The proper scripts to deploy the data automatically have been prepared. LHC experiments showed interest in this tool, new further instances can be easily created. - Stefan has been working on the preparation of mathlib and roofit from the latest ROOT 5.24 patches branch against the current 5.22 production version. The build instructions are part of LCGCMT and the build and deployment process works. We are waiting for confirmation of one last bug fix in roostats before producing the installations in afs. - Discussion about the literal naming of the compiler on slc5 and gcc 4.3. The last release of LCGCMT (56c) does not contain any changes in this respect. The HEAD of LCGCMT and also 56-patches branch rename the compiler to e.g. "lcg-gcc-4.3.2" (same for "g++" and "gfortran") in the case of the CMT tags "CERNDISTCC&target-slc5&target-gcc43". Marco would propose to use the "lcg-gcc-4.3.2" name also for compilations outside distcc. Advantages would be additional safety to not mix up compiler environments. The followup on this discussion will be to - Check with Atlas if the tag CERNDISTCC is really used or whether it is "use-distcc" (part at the moment of GaudiPolicy) - For what use cases we want to rename the compiler. The three possible use cases are "usage on the CERN distcc cluster", "usage of distcc in general", "usage of gcc 4.3.2 on slc5" Stefan will contact Emil to get some more information and then go with the info from Marco and Emil to re-discuss it in this weeks AF meeting. Round Table === LHCb --- - Planning a new release of Gaudi with next LCG release (changes of compiler name) - SVN service was broken because https was dropped and introduced http without notice. The question about the status of this was asked to svn.support (also by SPI) - no reply so far. ROOT --- - gccxml on snow leopard fails. Some non standard defines "DEFINE_BLOCKS" which cannot be parsed by gccxml, with gcc 4.2 with snow leopard. A bug report has been submitted. CMS --- - Problem with HepMC, occasionally gccxml dictionary generation for a CMS package fails. It seems to have a problem with the order of a template constructor. Persistency --- - Oracle, devdb10 will be migrated to the "light character set" so we can move to the "light instant client". Andrea will provide a new installation of the client libraries of version 11. - Last week first tests of CORALSERVER at Atlas point 1, successfully completed. - There will be a POOL meeting today. - In the Atlas control room, processes are failing with Python error in os.listdir which is fixed only after 2.5.4 - Might be useful to move to rebuild 2.5.4 with this patch (one .cc file) - Andrea will provide more info on this. Aob === - Snow Leopard: We will add 64 bit compilations, 32 bit to be turned off? To be decided in the next AF. - CMT : Grigori is preparing a new version of CMT and attended to get feedback about fixes that should be put in. - An email was sent by Stefan to Grigori with problems that were encountered in the context of SPI - In case of "make <target>" on Windows if the target group is not defined there is a warning spit out by CMT a la "NMAKE : fatal error U1073: don't know how to make 'tests'" , which is not the case on linux/darwin. It would be good to have the same behavior on all platforms. - Generation of MS Visual Studio VC9 solution files - Generate a CMT tag which denotes the exact compiler version on Windows (e.g. vc71, vc9) - Generate a generic tag for the slc version (e.g. slc5 instead of slc52), similarly for the compiler name "gcc34" instead of "gcc346" as we are only depending on the "short versions" of platforms and compilers on these os. - More issues that came up in the discussion - CMT locking has a problem. "lockfile" is not guaranteed to be installed (this is part of the "procmail" rpm). Enabling disabling the locking in CMT would be useful. Disabling is already possible by emptying a CMT macro. Marco prefers to have it disabled by default and enable it with a CMT tag. - make -j all_groups doesn't work because if a group depends on another group (tests->default), CMT tries to build the default group twice. - CMTUSERCONTEXT with 'make -C' does not work. You need to go to CMTUSERCONTEXT and then do "cmt make" - CMTUSERCONTEXT and cmt broadcast - this is fixed?

11:00 → 11:10 News

11:10 → 11:40 Round table

11:40 → 12:00 Aob