20060808 - GLAST Core Software Meeting
Attendees (may be incomplete, apologies to those left out):
Richard, Toby, Tom, Heather, Jim, Navid, Chuck, Tracy, Francesco
==> Beam Test Status(Heather/Francesco)
* Beam is back.
* Some problems with BTtuple generation. Also, merging of ntuples has exposed a problem in ROOT v5.10 - which seems fixed in v5.12
* BT pipeline now using latest BT release (v4r0909p7) containing much of Heather's recent ROOT I/O work
* Gearing up for new BT release today (will include ancillary data) which is hoped to be the final BT release for a significant block of BT data.
* A socket connection between the data collection software and Gleam may be completed for Linux today. This will be used for online data monitoring.
==> Announcements (Richard)
* ISOC review next Thursday, Aug 17th, at SLAC, Kavli Bldg. Practice talks this Thursday.
* Persis Drell requests closeout of last ISOC review this past Spring
==> CVS history file corruption (Tom)
An email circulated to the core software list described the corruption present in the GLAST CVS code repository's history file [ref:
http://www-glast.stanford.edu/protected/mail/infrasoft/1593.html\].
This file contains a log of cvs actions (tag, commit, etc.) for the entire repository and can be a valuable tool for understanding the sequence of changes that have been applied to a given code package.
Terry Hung, CVS maintainer for BABAR, does have some tools to deal with this problem and has offered to provide advice to anyone within GLAST who is interested.
Navid has offered to look into this problem.
==> 08:15 Another unscheduled Evo (partial) outage. Many people drop out of RAT and Koala windows and must reconnect. Note there is a "Monitoring" facility for EVO, accessibly from the EVO start page (http://nextgen-caltech.cern.ch/evoGate/). By clicking on the "GMap"
icon on the left side, one brings up a springy real-time network diagram of the EVO universe. Lots of options for viewing audio connections, total traffic, etc. Useful if one can avoid getting lost in the details.
==> Library file generation, .a and .so files (Tom)
An email circulated to the core software list described a problem wherein during the operation of the Pipeline for which the GlastClassify package overrode the default version in GR-v9r9, it was not picked up and used as expected [ref:
http://www-glast.stanford.edu/protected/mail/infrasoft/1589.html\].
This was due to a linked-in static dependency on GlastClassify by the "merit" package.
There was discussion about the pros and cons of producing sharables, the details of which include: differences (conflicts?) between the Windows and Linux development environments; Gaudi vs. non-Gaudi packages; special consideration for packages in both GlastRelease and ScienceTools. Several proposals were put on the table.
1. Create .so (dynamically loadable) libraries for all packages, Gaudi or not.
2. Add code to Gaudi to check that the code being used is consistent with the dependencies within the official release.
3. Create .so libs for a selected packages that have generated problems in the past, e.g., GlastClassify
4. Merge the merit and GlastClassify packages
5. Adjust the requirements file for GlastClassify so as to create a .so library and to properly provide linkage (directory path) information.
Toby noted that we currently lack people with the appropriate expertise and time to fix this problem completely at this time.
He also raised the question of whether the pipeline should really be running code consisting of official releases with overrides. The reality is that while these overrides are not desirable and should be minimized, they are inevitable due to the need for flexibility in debugging a particular pipeline task in a timely manner.
At the end of the discussion, it seemed agreed that we go ahead with some combination of proposals 3., 4. and/or 5.
==> migration to ROOT v5.12? (Francesco, Heather, Richard, Toby, Tracy)
There are now two identifiable problems with the current GLAST version of ROOT v5.10. First is the problem associated with ROOT v5.10 and ntuple merging seen by BT group. There is also a problem associated with the use of xrootd as seen by Heather and Wilko Kroeger. Both problems seem fixed by ROOT v5.12. Heather pointed out that v5.12 was released on 3 August 2006 and has already been patched. She expressed some concern about jumping to 5.12 prematurely, and offered to contact Rene Brun for an official reading of the tea leaves.
It does seem likely that a migration to a new version of ROOT will happen in the not too distant future.
==> Visual Studio 2005 (Toby, Tracy)
Toby and Tracy discussed another aspect of the new ROOT version associated with a new version of Visual Studio 2005 (Windows development environment). The new VS2005 is more C++ standards compliant and comes with new runtime libraries that are incompatible with the old ones to the extent that all code must be rebuilt. For the various GLAST_EXT packages this can be a problem. ROOT v5.12 claims to support with the new VS2005 so is a step in the right direction. But there are other packages, such as OmniORB which either require substantial Windows expertise and time to rebuild, or support from the developers. Work continues...
[Afterword from Toby:"I took a quick look at omniORB, since the minutes feature it. There is a "VC8" version for the latest release, 4.0.7. (We are using 4.0.3.) So this does not seem to be an issue either, except that isn't it only Riccardo that has any expertise?
This is only for FRED, right?
I think that we are approaching a situation where the "big" external libraries are available, or implemented as sharables. That includes xerces and mysql."]