Present: Joanne Bogart, Toby Burnett, Richard Dubois, Marco Frailis, Riccardo Giannitrapani, Traudl Hansl-Kozanecka, Heather Kelly, Michael Kuss/Johann Cohen-Tanugi, Sean Robinson, Leon Rochester, Alex Schlessinger, Tracy Usher, Karl Young
See also Toby's Core Issues/Status document for some comments on several of the items below.
Releases: v3r1 will be the next-to-last release of Gleam. We'll put in the new ACD geometry today and call it v3r2. At that point it will be good exercise to go through the motions, e.g. making binaries for distribution.
We need to review package tests: make sure they exist and are meaningful.
Subsequent code releases will be under the new scheme, in which the release has a container package (GlastRelease) at the top with no c++ source or resultant executables. It will contain all packages of interest (e.g., things like userAlg, ntupleWriterSvc) whether or not they're needed to build a particular favored executable. To start with packages will be referenced directly by the top container, but this should soon evolve to a structure in which the top container contains other containers. Plan on discussing the scheme next week; see Traudl's documents: releaseTagManagement.ps or .rtf and the accompanying diagrams in GlastPkgDependencies. We also need a plan for implementing it and person to do it, keep track of new tags, etc. Alex was volunteered.
Tools status: (mostly Toby) Gaudi v11r1 has been released and (thanks in large part to our Pisa folks) we can expect gcc 3.2 to compile it without difficulty. Alex confirmed that gcc 3.2 is already available on SLAC Linux and it is a simple matter to point to it. The new Gaudi requires new versions of some external libraries, among them CLHEP and Xerces. We should check that the new Gaudi and new G4 can get along with the same version of CLHEP. [I built the "new" (newer, but still old) version of Xerces, 1.6, some time ago and used it successfully with our code, as did Michael and Johann in their gcc 3.2 work, so it is unlikely to cause any problems. ed.]
The new CMT, v1r12p20020606, has a somewhat different format for the output of the show uses command, requiring adjustments in vmct. Toby has a new Windows vcmt which will work correctly for either version of CMT. Marco has made a new Linux vcmt which will work correctly with the new CMT, but is not backward compatible with the old. He will see about enhancing it further to make it backward compatible. Meanwhile, Alex has successfully used the new version of CMT on Linux with our current Gleam package.
With our current style of CMT use, files which should be rebuilt don't always get rebuilt, at least on Linux. The cure for this is a proper use of stamps. We need to develop some expertise on this subject.
The format of project files has changed for vc++ 7.0, so vcmt has had to adjust. It all appears to be working; a new version of vcmt, compatible with both compilers, will be released soon. The developer environment of vc++ 7.0 is significantly improved and pleasant to use.
CalDigi: (Richard) Expects to add a new bit field status word to keep track of hardware status.There is no rush to get it in since so far no existing code is expecting it to be there; code is ready to go on the HEAD which makes use of the status field in conjunction with the failure mode service.
CHEP '03: (Richard) We have done a lot, much of it not published anywhere. See this list of possible subjects.
Analysis tools: Bill has found what appears to be a wonderful tool, Insightful Miner. Wonderful, perhaps, but definitely expensive for non-academic users (among which Insightful chooses to include SLAC, for their purposes a DOE lab rather than part of Stanford). Most likely we will never be able to make this tool widely available to the Collaboration. Tracy warned against putting ourselves in a situation where important analyses cannot be reproduced. No one was inclined to disagree; this was already a problem during AO response preparation.
We need to find a way to make the essential functionality of Insightful Miner available to everyone. A favorite feature is "recursive cuts". The underlying ideas are not new and were developed by a SLAC guy, Jerome Friedman. He is co-creator of another data mining tool, CART.
Dan Flath, our old co-op student, has been working on a ROOT gui (histogram browser, ntuple analyser, etc.).
Nightly builds request: We are about to embark on a period of enhanced instability and inconsistency of software in the HEAD. We will temporarily turn off the sending of emails during nightly builds of the HEAD since many if not most of them will be about transitory or irrelevant problems. There is no harm in continuing to do the builds, however, so that anyone who cares will still be able to look up the results.
previous | minutes index | next |
J. Bogart Last Modified: 01-Jun-2010 15:46:36 -0700