Agenda

Review minutes from July 1, 2009 Software meeting
Announcements
1. New release: release-2009-07-07
Serializing ROOT objects and transferring them with cMsg: Elliott
HDGeant crashes: Beni
Group disk quota increase and usage (work and cache for that matter): all
GlueX Software Task list
Action item review
- group review of list

Minutes

Present: Pawel Ambrozewicz, Eugene Chudakov, Mark Ito (chair), David Lawrence, Dmitry Romanov, Elton Smith, Sascha Somov, Victor Tarasov, Simon Taylor, Elliott Wolin, Beni Zihlmann

Review of Minutes

We looked at the minutes of the July 1, 2009 meeting.

DRootGeom changes--multi-threading and ROOT

David reported that some of the recent changes to the DRootGeom class have been moved to the DGeometry class. Testing is in progress with low momentum protons. Large systematic shifts are still seen.

Proposed changes to mcsmear

David has made many changes. The process turned out to be a bit complicated. There is an outstanding issue on how to handle the BCAL. David will lead a discussion on this at a future meeting.

Hall D Unix groups

Elliott had Kelvin Edwards implement [the proposal]. There are still several files on the group disk whose group ownership needs to be cleaned up. Elliott will take care of this when he returns from vacation.

There was a question about how group ownership is handled for files written from Windows. We should understand this.

Tracking removed from HDGeant

Mark will ping Richard on making the change we requested: take trajectory curvature into account when calculating the point of closest approach for the CDC straws.

Not-a-number problem in the BCAL cluster code

Sascha contacted Blake who instituted a fix and checked it in.

Announcements

Mark has issued a new release. He asked whether we want to formalize the idea of a "production". We agreed that that was not necessary at this point.
Elliott reported that there are sometimes conflict in the use of voldemort.jlab.org: Primavera use here and e-cad software remotely by a student at UConn. Elliott will contact the student and ask how often he really needs to lock out other users.

Serializing ROOT objects

Elliott led us through his explanation and example of how to do it. He has collected the essential information and tricks one needs. This could have use for distributing online monitoring histograms over the network.

HDGeant crashes

Beni described the problem. The symptom is a reported FDC plane of zero, which should never happen. It seems to associated with very low energy (several keV) secondary particles and occurs in one of the "FDA" volumes. Frequency is at the level of a small fraction of a percent of events. Richard has been told about the problem and Beni has sent an example case to him for debugging.

GlueX Software Task List

Mark and David led us through the existing list. It is meant as a place to collect long-term projects, assign responsibilities, and identify broad areas where more effort is needed. Many editions were suggested, mainly in personnel, but most of these will not be detailed here. Mark was tasked with collecting the comments and making the appropriate updates. Here are some highlights of the discussion:

Pawel volunteered to look at package/release management.
Dmitry suggested that we consider having test harnesses for low-level packages, e. g., a BCAL cluster finding test, rather than depending on global statistical quantities (things like invariant mass plots) to indicate problems with the code. This is a common practice in industry. In addition, having these available can help a lot when the high-level diagnostics do indicate a problem and might be worth the (not-insignificant) effort needed to develop them. We could require them.
Mark wondered if the grid tools being developed for PWA are applicable for "standard" event simulation and reconstruction.
David suggested (and we agreed) that the calibration constants, currently checked into the Subversion repository, should be tagged and checked out along with software releases. This should be done at least until a proper database system has been developed.
Elton suggested that we discuss the idea of Monte Carlo database manager at the Physics meeting on Monday. Ryan Mitchell may have ideas based on what he has done already.
Dmitry thought that he and colleagues at MEPHI would be interested in taking on the calibration database. He will bring the idea back to his colleagues.

Organizing the Work

Mark asked how do we monitor progress on all of these tasks. We briefly discussed two tools:

Milestones: a list could be developed for each area, most likely by the responsible individuals
Action Items: these keep the effort on track on a meeting-to-meeting basis

Elton and Eugene both noted that the task list does not address the priority of the tasks and sub-tasks. Eugene mentioned:

optimization of FDC geometry
plan for (or lack of need for) mapping of the magnetic field
full reconstruction of important event topologies

These are not specifically mentioned on the task list and are of greater scope than the usual action item. And they require a significant software effort.

We started to focus on the question of mapping the field.

Mark suggested that a sub-committee be formed. It was thought that we have too many meetings already and the personnel would be the same as in the general meetings anyway.
We thought that there should be a specific sub-group of the collaboration reponsible for the issue.
We debated whether the proper home was the software meeting or the tracking meeting.
Eugene suggested that the responsibility could be split: the "software group" could come up with the tools, the "hardware group" could do the studies.
David agreed to package his software for doing studies so that others can use them.
Dmitry volunteered to work with Eugene on generating various field configurations.

Deferred Agenda Items

We had to push some items to next time:

Group disk quota increase and disk usage in general
Action item review

New Action Items

have an mcsmear discussion at a future meeting: David
group ownership clean-up on /group/halld: Elliott
understand group ownership on write from Windows: ?
contact e-cad user at UConn: Elliott
update task list: Mark

recorded by --Mark Ito 30 July 2009

July 29, 2009 Software

Contents

Location

Telephone

Video Conferencing