Present: (Alex), (Claudia), Heather, Joanne, Karl, Leon, (Pasquale), Richard, Sasha, Sean, Toby, Traudl, Tracy [( ) indicates seen but not heard]
Gleam/reconstruction: (Toby) Recent release contains lots of fixes; still a problem with rotations in silicon. One big memory leak has been found and fixed. We still have at least one lesser leak. We need to resuscitate leak-finding tools.
We also have to discover what might be causing a crash on events like this one. Likely it is a combination of a bug on our part and bad error handling by G4. Heather suggests we try a more recent version of G4 to address the second issue.
A nice debugging feature of the flexible geometry is the ability to specify a "top volume" other than the default (full detector). One can then focus on any problem area. As in this example from the Recon package, a new top volume can be specified in the job options file. (Scroll down a bit to see the relevant section.) Any box, stack or composition appearing in the XML description may be used as a top volume. Refer to it by using the value of its name attribute.
Claudia et. al. have run into problems building Gleam on Linux (cf. Nightly builds). They have built the system on Windows in order to continue their work, but this should not be necessary.
CalRecon is definitely ready to be re-integrated into Gleam; TkrRecon is as well, or soon will be (depending on definition of "ready").
(Tracy) The G4-using version of RcParticle is nearly ready to be committed, now that changes have been made to cooperating packages to make sure that there is only one instance of RunManager.
Nightly builds: (Karl) Most recently all were successful except for TkrRecon. Karl had to make some changes to code to get all the successful builds; he needs to confirm that the changes made didn't affect the intended operation and then commit them.
We still don't have a good solution to knowing what to build against. As Traudl mentioned (already largely instituted by Alex), intelligent use of versioning to indicate interface changes and the partial wildcard form of use statements, e.g.,
use GlastSvc v5*
helps in many situations. However we have another class of problems where one package depends on incompatible but not-yet-tagged features of another. Conclusion: we need to tag more often [..meanwhile respecting the rule that packages be known to build on both platforms before tagging. ed.]
ROOT news: (Heather) ROOT versions of MC classes are in CVS; still a bit of clean-up left to get rid of warnings on Linux. Near future: make an algorithm to write a (ROOT) file from the TDS classes, then make digi ROOT classes (ACD - Heather, tracker - Leon, Richard has already started work on CAL).
Calibration infrastructure: (Joanne & Madhup) Two new nearly empty packages have been added to the repository. MYSQLEXT is an interface package for MySQL. We have an unofficial installation of MySQL client software available on Linux at SLAC (really it belongs with the other external libraries) which MYSQLEXT points to. The other package, calibUtil, will ultimately contain classes for accessing both the metadata in the new database and the actual data, probably in the form of XML and ROOT files. We are working on the XML description for bad strips.
Travel schedule: (Richard) It should be a useful reference. Please keep it up to date.
J. Bogart Last Modified: 04-Aug-2004 15:40:30 -0700
