GlueX Software Meeting, September 27, 2021

From GlueXWiki
Revision as of 19:23, 10 October 2021 by Marki (Talk | contribs) (Report from the September 2nd SciComp Meeting)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

GlueX Software Meeting
Monday, September 27, 2021
3:00 pm EDT
BlueJeans: 968 592 007

Agenda

  1. Announcements
    1. /work/halld3: transition to new server
    2. new work disk has snapshots
    3. New recon-launch-compatible version sets with tagger energy fixes
    4. New version set: 4.46.0
  2. Review of Minutes from the Last Software Meeting (all)
  3. Review of Minutes from the Last HDGeant4 Meeting (all)
  4. Report from the September 2nd SciComp Meeting (all)
  5. FAQ of the Fortnight: How do I check on the status of my JLab-mandated training?
  6. Documentation Writing Festival (Sean)
  7. Review of recent issues and pull requests:
    1. halld_recon
    2. halld_sim
    3. CCDB
    4. RCDB
    5. MCwrapper
    6. gluex_root_analysis
  8. Review of recent discussion on the GlueX Software Help List (all)
  9. Meeting time change? (all)
  10. Action Item Review (all)

Minutes

Present: Alex Austregesilo, Thomas Britton, Sean Dobbs, Mark Ito (chair), Richard Jones, Simon Taylor, Beni Zihlmann

There is a recording of this meeting. Log into the BlueJeans site first to gain access (use your JLab credentials).

Announcements

  1. /work/halld3: transition to new server. The transition happened as scheduled on September 2nd. Mark asked the group whether we want to maintain a set of links to preserve the viability paths like "/work/halld3" even though /work/halld3 no longer exists as a physical partition. Recall we have some links like this to /work/halld2 from many years ago. No one seemed bother by the current situation, so we will likely leave well enough alone.
  2. new work disk has snapshots. Mark pointed out the special directory one must use to access snapshots on the new work disk: /work/halld/.zfs/snapshot .
  3. New recon-launch-compatible version sets with tagger energy fixes. Sean explained that these changes have no effect until the photon end-point energy table has been filled in for the relevant runs. Once those values are in the CCDB, the latest scheme for energy assignment to tagger counters will kick in.
  4. New version set: 4.46.0. Mark released this version set yesterday. He is preparing another version set that has the same versions of the GlueX-written software, but updates our support packages, e.g., ROOT and Geant4. He mentioned that the version of Geant4 in this upcoming release looks like it solves HDGeant4 Issue #195: Crash adding daughters on RHEL/CentOS 8.

Review of Minutes from the Last Software Meeting

We went over the Minutes from the meeting on August 17th without significant comment. Unfortunately the last software meeting was held on August 31st. This may explain the lack of significant comment.

Review of Minutes from the Last HDGeant4 Meeting

We went over the minutes from August 24th. Again the chair directed us to the wrong meeting. The last meeting was held on September 7th. Not a lot of comment here either.

Report from the September 2nd SciComp Meeting

Mark went through the slides that Bryan Hess showed at the meeting. Please take a look for all of the details. One item: statistics on files permanently lost during the tape library incident last year.

Thomas filled in details of the discussion on OSG-related topics. scosg20.jlab.org is now in production, replacing scosg16.jlab.org. This new Submit Host is "on and island;" it does not mount group, cache, etc. to prevent abuse by users. No Hall D users are part of the history leading to this decision, AFAIK.

Richard raise the question of access to files from tape via the cache disk. Thomas pointed to use of dtn1902.jlab.org (the Data Transfer Node) as a possible solution. In the interest of time we were not able to bring the discussion to any conclusion. We need to take up the topic in the near future in some forum or another.

Mark mentioned that there is a new mailing list at the Lab dedicated to discussion of OSG-related issues for the JLab community. Go here to subscribe.

FAQ of the Fortnight: How do I check on the status of my JLab-mandated training?

Mark pointed out that the link he posted in the chat at the Collaboration Meeting is recorded in an FAQ. This also a FAQ on how JLab staff can get a list of their sponsorees.

Documentation Writing Festival

We had a discussion on how we might improve documentation for our software.

  • Beni proposed the idea of a Documentation Writing Festival.
  • Sean: documentation exists but is scattered and hard to find.
  • Beni: a Doxygen tutorial session would be useful.
  • Beni: we also need a PDF-file-like description of software systems
  • Beni: code added since Paul Mattione was here has been added to and the additions are not usually documented.
  • Beni: variables in ROOT trees are poorly documented. This leads to use of variables that are not fully understood.
  • Mark has a back-burner project to revive our Doxygen site. It was broken with the sim-recon to halld_recon/halld_sim transition.
  • Sean pointed us to the link for the sim-recon Doxygen page. He thinks the organization could be improved.
  • Mark thought that we should first decide on the structure and practices with which documentation should be developed and maintained.
  • Beni: the terminology of software objects should be standard or we run the risk of talking past each other.
  • Thomas used Doxygen when he was a grad student on LHCb. He liked it.
  • Richard asked about Doxygen and FORTRAN. Apparently this is not impossible.
  • Thomas told us that software packages for LHCb often came with a nice, thick, comprehensive manual, with explanations and examples.
  • Thomas and Beni: Multiple levels of documentation needed. One level does not do the job.
  • Thomas: software documentation is a cultural problem.
  • Beni: approval of pull requests could depend on having an acceptable level of documentation in the code.
  • Alex and Beni: the DSelector and the GlueX ROOT Analysis package might be the most important to document since they have the greatest number of users.
  • Richard: the analysis library in halld_recon is very complicated. It would be hard to document without Paul around.
  • Richard remarked that the trend is software these days is toward Python. We should keep that in mind.
  • After Sean had to leave, Beni expressed the hope that Sean would take the lead on this effort.
  • Mark will also think about how we organize the discussion going forward.