Present: Joanne Bogart, Monica Brigida, Toby Burnett, Sasha Chekhtman, Xen Chin, Richard Dubois, Berrie Giebels, Heather Kelly, Michael Kuss, Francesco Longo, Christophe Perrot, Leon Rochester, Alex Schlessinger, Karl Young
GlastRelease nightly builds: (Alex) Basically in good shape. Gaudi v11r4, when compiled with the new compiler, appears to have problems, but Alex thinks a later version will be OK.
Occasionally, as happened last night, the unit test summary page is not generated automatically, as it should be. In this case Alex has been generating it "by hand".
What's new in GlastRelease: (Toby) He presented a document describing in some detail all the significant new features in GlastRelease as compared to the Gleam era, a far-ranging collection:
GlastRelease is a pure container package, referencing packages needed not only by the current Gleam executable, but also packages (e.g. GRB) needed by other applications or by anticipated future releases of Gleam.
The nightly build system builds three kinds of objects:
Package collection as specified by a tagged version of GlastRelease requirements file
Package collection as specified by GlastRelease HEAD requirements file
Latest tags of all packages mentioned by GlastRelease requirements
The last is particularly useful in determining when it is appropriate to promote a tag to the GlastRelease requirements file (done by sending a message to the Infrasoft list with "GlastRelease tag request" in the subject line).
Use of Gaudi tools has increased considerably. Note tools should defer all serious initialization to the initialize() method.
We have a new container package, IExternal, to keep track of interface packages for all external libraries.
Gaudi is now an external, but still organized using CMT conventions.
Geant4 changes. We are now at version 5.0. It can be built on Windows using Tracy's Geant4build package.
ROOT changes. We're at version v3.04.02. There is (still) a problem with ROOT and vc7.
Gleam is no longer a container package. It builds the Gleam executable and contains the system tests, however Karl is planning to move the system tests to a new package of their own. The natural schedule for tagging system tests doesn't bear any close relation to the tagging schedule for the Gleam executable.
CMT wrappings. There have been some enhancements to vcmt (Windows and Linux versions) and glastpack. Linux vcmt is not yet available from the CVS repository but is installed at SLAC.
Documenting debugging: Heather has uncovered some information, in need of a little freshening up, in the Gaudi User Guide concerning debugging Gaudi applications under Windows. Joanne has written most of a Linux debugging guide which describes how to get started with gdb, particularly from within an emacs sessions, and also information specific to debugging Gaudi applications.
Calibration infrastructure: (Joanne) Calorimeter continues to be the focus. She is trying a new templateless approach; so far the amount of boilerplate code which needs to be duplicated is small and manageable. Data classes fall into two categories: data-per-range classes (see, e.g. the one for pedestals) and full-collection classes such as CalCalibPed.The latter sort all inherit from a base class which handles the common organizational issues. Classes (data classes in CalibData and converters in CalibSvc) needed for pedestal and gain datasets exist but await debugging.
G4 news: (Francesco) See his report for a complete rundown on G4Generator upgrades. To summarize:
now have a choice of two supported physics lists: only_em and full
discovered a facility in G4 allowing one to save a table of various physics properties (e.g. cuts), then in another run retrieve it, thereby saving the time it would take to rebuild the table. He has added properties to G4Generator to allow access to this facility via job options. (Richard) Should we save standard cuts in CVS? (Francesco) Difficult to manage well since cut values in general depend on the detector description, e.g. the materials used.
So far these upgrades have only been tested on Linux; Windows testing is about to commence.
EM data format: (Heather) She just got a current version of Ric Claus's parser, and is now creating a standard SAS CMT package so that we may compile and use it for our own purposes in EM data processing.
System tests: Hung up just now on technical problems having to do with Carrot Server.
Handling major version tags: A new major version tag for FluxSvc, v6r0, necessitates changes to requirements files of several other packages which use FluxSvc. Toby suggested that the nightly build output could include the results of cmt show uses. Alex says this output exists in the file compile_packages.txt but is not currently easily accessible from the nightly builds output page. He asked if we should continue to include use statements of the form use myPack v3r* or whether we should depend on GlastRelease to keep track of all versioning information. Discussion to be continued offline.
previous | minutes index | next |
J. Bogart Last Modified: 01-Jun-2010 15:46:19 -0700