Present: Joanne Bogart, Toby Burnett, Claudia Cecchi, Marco Frailis, Heather Kelly, Sean Robinson, Leon Rochester, Alex Schlessinger, Karl Young
Gleam release: (Toby) See his Gleam notes (new features, to-do) for details. Items discussed include:
Failure of test program on Linux.
Need for a proper scheme to limit MC particles written to ROOT file. The files as currently written are unpleasantly large. A first attempt (just a simple cut-off) caused problems elsewhere when position hits pointed to non-existent particles.
Finding eof in ROOT files. Has been handled. Current solution is not perfectly graceful, but does allow finalize( ) to run as it should.
Writing ROOT output using global rather than local coordinates. This is straightforward to do and probably should be done so that ROOT analysis programs don't have to have access to full geometry services. ROOT output would then presumably differ from the form of the information in the TDS, but there is no reason they have to be identical.
Status of Tracker code.
New code fixes some bugs (maybe relevant to Linux failure) but is not yet tagged. This will happen soon.
There has been some discussion of what information belongs in the TDS/PDS as data and what can be relegated to computations within methods. This is of course not just a Tracker issue. The Atwood criterion is to save as data everything required for the next stage of processing. [Clarification: in this context information which may be calculated from saved data without recourse to geometry or other extraneous services is not considered to be required.]
Random seeds, a long-standing if not yet high-priority item. We need someone to look into controlling random number generation via the Gaudi service; we must be able to reliably reproduce production runs. Toby suggested using the same sequence throughout the code (for G4, FluxSvc, digi noise generation,...). However Leon proposed a requirement that we be able to generate the "same" set of events with and without the noise, which would imply using at least two sequences.
Old data and new code. Can we expect to be able to analyze, e.g., testbeam data with the new code? One significant hurdle, though not the only one, will be geometry description. There is nothing in principle preventing us from writing a suitable description in the new format, but it will take some effort.
Relational tables for particle/hit relations (Marco). The code is done; he and Riccardo will be writing something to test it tomorrow. One worry is that, since maps can't be used for this kind of application with Visual Studio 6, the implementation is relatively inefficient.
Releases and builds: (Alex) The Test Release facility is working; follow links to see the logged output. When it is properly integrated with the Nightly Builds its output will go to a more suitable public place. The (binary) results of the build could readily be written to a release area, but currently we don't have the space; this is being looked into. More problematical, and probably of higher priority, is to create a binary distribution, usable apart from CMT.
Worth looking into:
There is a new tool available for Windows machines, TortoiseCVS, providing a much nicer gui interface to CVS than WinCVS.
Heather has provided and documented a new interface package to the standard cfitsio library.
J. Bogart Last Modified: 04-Aug-2004 15:39:50 -0700