Difference between revisions of "December 4, 2006 Software"

From GlueXWiki
Jump to: navigation, search
Line 13: Line 13:
 
==Minutes==
 
==Minutes==
  
Attendees:
+
Attendees: Alex D. David L.(chair) Elliott W. Matt S. Simon T. Zisis P.
 +
 
 +
===Report on Meeting with JLab IT group on OSG===
 +
Alex reported on a meeting he and Elton had earlier in the afternoon with Chip Watson, Sandy Philpott,
 +
Andy Kowalski of the JLab IT department. This was primarliy to discuss the numbers in
 +
[http://dustbunny.physics.indiana.edu/~dzierba/osg Alex's presentation]
 +
to the [http://www.opensciencegrid.org OSG group] in 2 days. In short, the participants were
 +
all on board with the material in the presentation.
 +
 
 +
Alex also noted that they are still awaiting feedback on the
 +
[http://www.nsf.gov/funding/pgm_summ.jsp?pims_id=6681&org=PHY&from=home PIF] proposal
 +
submitted last September which relied heavily on the OSG. They don't expect to hear anything until
 +
February.
 +
 
 +
===Tagged versions of the source code===
 +
This was the main discussion topic of the meeting. At issue was that we are now in a mode where several
 +
people are contributing to the source code repository and we are nearing the start of several
 +
studies that will be done in support of CD-2. We should tag versions of the code so that we can
 +
reproduce any results obtained from these studies.
 +
 
 +
David noted that subversion is different from CVS in that tags are not "frozen" in the same way. Essentially
 +
the code is copied to the "tags" directory where it can be checked out, modified, and committed. The
 +
concern is that would be a little too easy for someone to accidentally check changes in to a tagged
 +
version that they did not want changes to go in to. David noted that it might be possible to restrict
 +
checkouts to exports only or to limit check-ins using the hook scripts. He will investigate this.
 +
 
 +
Currently, there is an old tag in the repository called "STABLE". David exaplained that this was an
 +
attempt to make a "floating tag" that allowed persons to always check out the most recent version
 +
of the code that was blessed as being stable. This would allow developers to commit broken
 +
or half-developed code in order to backup work that is significant, but not yet complete.
 +
 
 +
Matt pointed out that CLEO uses a tagging scheme where the date of the tag is used. If someone
 +
wants the most recent tagged version, they just list them and take the latest one. This avoids
 +
the problem of STABLE not being stable. It was agreed that we would adopt a similar policy in
 +
principle. David would present a specific proposal next week.
 +
 
 +
Matt suggested that we also put tarballs of the external packages(such as xerces) required to build
 +
the Hall-D source code on the web somewhere. These would be version known to work with
 +
the Hall-D souce code and would save folks time in trying to identify and find the correct version
 +
of a package. David noted that these were on the hall-d group disk at JLab already and he
 +
would somehow link to them from the Wiki.
  
  
 
==Action Items==
 
==Action Items==
 +
 +
#Freezing tags in subversion through hooks (David)
 +
#Periodic release tagging (David)

Revision as of 18:13, 4 December 2006

Agenda

  1. Tagged versions of the code (releases)(halld subversion repository)
  2. Near future policy decisions
    • Tagged versions of the code (releases)
    • Make system behaviors
    • Repository structure
    • Documentation (in and out of code, placement on web, organization ,...)
    • Event generators (standard set of inputs to genr8, other generators, ...)
  3. Status of CD-2 milestones


Minutes

Attendees: Alex D. David L.(chair) Elliott W. Matt S. Simon T. Zisis P.

Report on Meeting with JLab IT group on OSG

Alex reported on a meeting he and Elton had earlier in the afternoon with Chip Watson, Sandy Philpott, Andy Kowalski of the JLab IT department. This was primarliy to discuss the numbers in Alex's presentation to the OSG group in 2 days. In short, the participants were all on board with the material in the presentation.

Alex also noted that they are still awaiting feedback on the PIF proposal submitted last September which relied heavily on the OSG. They don't expect to hear anything until February.

Tagged versions of the source code

This was the main discussion topic of the meeting. At issue was that we are now in a mode where several people are contributing to the source code repository and we are nearing the start of several studies that will be done in support of CD-2. We should tag versions of the code so that we can reproduce any results obtained from these studies.

David noted that subversion is different from CVS in that tags are not "frozen" in the same way. Essentially the code is copied to the "tags" directory where it can be checked out, modified, and committed. The concern is that would be a little too easy for someone to accidentally check changes in to a tagged version that they did not want changes to go in to. David noted that it might be possible to restrict checkouts to exports only or to limit check-ins using the hook scripts. He will investigate this.

Currently, there is an old tag in the repository called "STABLE". David exaplained that this was an attempt to make a "floating tag" that allowed persons to always check out the most recent version of the code that was blessed as being stable. This would allow developers to commit broken or half-developed code in order to backup work that is significant, but not yet complete.

Matt pointed out that CLEO uses a tagging scheme where the date of the tag is used. If someone wants the most recent tagged version, they just list them and take the latest one. This avoids the problem of STABLE not being stable. It was agreed that we would adopt a similar policy in principle. David would present a specific proposal next week.

Matt suggested that we also put tarballs of the external packages(such as xerces) required to build the Hall-D source code on the web somewhere. These would be version known to work with the Hall-D souce code and would save folks time in trying to identify and find the correct version of a package. David noted that these were on the hall-d group disk at JLab already and he would somehow link to them from the Wiki.


Action Items

  1. Freezing tags in subversion through hooks (David)
  2. Periodic release tagging (David)