Present: Joanne Bogart, Toby Burnett, Richard Dubois, Riccardo Giannitrapani, Heather Kelly, Sean Robinson, Leon Rochester, Alex Schlessinger, Tracy Usher, Karl Young
Gleam today: See Toby's report for details. New features include improvements to the gui making ROOT plots easily accessible, which already has helped to uncover problems.
The plan for today is to incorporate a few enhancements and fixes into a release version and make binaries for distribution of that version. Expected modifications are
Small change to GlastDigi requirements file and change to Gleam requirements to use GlastDigi so that developers can easily (via jobOptions) switch to Bari digi algorithm if they want.
Small bug fix to xmlGeoDbs to number ACD sides correctly.
New G4Generator, incorporating some fixes Tracy has made and new code from Riccardo for heavy ions.
It has been confirmed that the so-called TkrRecon cpu/time hog problem is due to code in the gismo version of the propagator.
Release procedure, branch tags: We plan to apply a new branch tag, e.g. Gleam_v3, to all packages in the release. Then run a script which will identify every source file in the branch. Such a list is not a foolproof way to keep track of what's in there since tags can be moved but this is heavily discouraged. Only minor development will occur in the new branch. We still don't know precisely how we're going to use branches to keep straight "normal" development versus development within the branch, how to enhance VCMT and glastpack to help, etc. Alex will experiment on a copy of our repository. (Richard) Do we even want to make a branch when we don't fully understand the implications? (Toby) We know there is no harm in it.
One important component of making this work is to make it easy for developers to switch back and forth between developing release code and developing in the main branch, for example external libraries and CMT version will be different. The latter could be on stumbling block on the terminal server, where all users are pointed to a single version of CMT. Continuing development on and patches to a release branch should be carefully controlled and restricted to a small number of people.
G4Generator: Richard had some questions for Riccardo:
Riccardo mentioned that he is again accessible via email and ICQ. [Welcome back! ed.]
Conversion to new tool versions: (Toby) Output of show uses has a different format in the new CMT, requiring some changes to VCMT and perhaps also glastpack but otherwise checks out and builds our code as before. Now we can investigate making use of new features, like better support of public/private.
Most if not all of our code is ready for g++ 3.2. We're waiting for the Gaudi people to finish making Gaudi g++ 3.2 compliant; then we can move to it. [A new version was released today; see these release notes. However they include the statement "Unfortunately it does not yet run with the platform/compiler combination Redhat-7.3/gcc-3.2 due to problems with dynamic_cast<>. We are investigating it." ed.]
System tests: (Karl, Alex) They work and appear to be producing reasonable output on Linux. There have been some annoying terminal server (client side) problems interfering with Windows development, but these are about to come to an end. It should soon be possible to run the system tests on Windows and compare the output from the two platforms.
CDR stuff: (Richard) Steve Ritz has come up with several items needed for CDR which impact us:
calibration -- ability to misalign the tracker
demonstrating we understand the statistics we need for the Calibration Unit beam tests
aligning the star tracker
get JJ's trigger code into Gleam
parametrize pile-up - to estimate the deadtime effects of CNOs on the system.
not for CDR - but we need to do our annual re-evaluation of the LAT performance. [Tune has since requested that we demonstrate the background cut performance. He has also asked for major ACD failure modes and mitigation]
previous | minutes index | next |
J. Bogart Last Modified: 04-Aug-2004 15:39:23 -0700