Present: Alex, Heather, Joanne, Karl, Leon, Michael, Richard, Toby
Today's meeting covered a lot of ground, most of it already well documented in the web pages presented then and referenced below.
Documentation Task Force: (Heather) We had our third meeting yesterday. See the minutes here.
New meeting time arrangements. From now on we'll be meeting only every other week, still Tuesdays at 8:30 AM PST, alternating with the Data Structures Task Force (and overlapping TKR meetings). Our next meeting will be in 2 weeks (Feb. 12).
Karl solicits comments on his new page, which describes external packages and libraries we use, and also on the new look for the Latest Checkout Package Tags page.
The Task Force's focus is moving to Web documentation. Everyone agrees that GLAST pages don't belong in personal space.What do we do about the fact that we have multiple group GLAST web sites? There are good reasons for keeping both SLAC group sites (Windows-served and unix-served) as well as allowing use of remote sites, however the SLAC Windows site is to be considered primary, in the sense that index pages and organizational pages generally will be kept there, as well as most other human-edited pages. Other sites will contain primarily leaves.
We still haven't disposed of the $Header$ issue, though it has perhaps mutated to a discussion of how to keep track of package-wide release tags rather than individual file versions.
Data Structures Task Force: (Leon). Here are the complete minutes. Some highlights:
Toby reviewed McParticle and McVertex designs.
New class GlastAxis added to the idents package.
Core Issues: (Toby) See many details on many matters in his report.
See especially Heather's thorough discussion of the move to Gaudi v9 and use of various versions of the gcc compiler. Bottom line: there are still issues with compilation and also with incompatibilities between the old and new Gaudi APIs, but it's coming along.
When must a package compile on all platforms? It would be impractical to require this before CVS commit since the author would have no good way to check that code compiles on whichever platform he/she is not using for development. Toby proposed that we require a good build on all platforms before tagging; carried with no objections. The build-on-demand functionality of the Release Manager should make this practical. [Documenting differences between the compilers so that developers would be forewarned also helps. ed.]
The flux package has been merged into FluxSvc; there was no need for keeping them as separate packages.
VC++ 7.0 and friends: Toby has started to look into it. There are some incompatibilities with VC++ 6.0 which will have to be dealt with before we can think of moving.
Riccardo and Marco are about to commit a new package, HepRep, which will provide access to better graphics. In the planning stages is another package, HepRepSvc, an interface to HepRep for Gaudi algorithms. To see how the pieces will fit together, take a look at this block diagram.
Release Manager: (Karl) The spec is a day or two away from release. One of our students can perhaps work on providing a convenient Web interface for the build-package-on-demand functionality which is a part of this facility.
Database Manager Spec: (Richard) See the draft CyberDocs document, DPF Prototype Processing Database (LAT-TD-005530-D1),which he, Karl and Karen have been working on. The design is similar in many respects to the design of the analogous SLD facility, which had to handle comparable amounts of data. Len Moss, the architect of the SLD design, has reviewed this design and blessed those aspects which differ. As the next step in this project, Karen will create (in Oracle) the tables described in the document and Deepu will help her exercise them with fake data.
Sim/Recon Futures: Toby is working on a plan for what needs to be done when. This will aid in incorporating the promised help from Pisa.
J. Bogart Last Modified: 01-Jun-2010 15:48:43 -0700