Core Minutes 9/8/2009

• Collaboration meeting: (Richard) the next meeting will be in March of 2010, likely in Paris.

  Pass 8 (Leon) Almost any development in GR is going to interact with the massive restructuring associated with Pass 8. Pass 8 recon will include more than just proper handling of ghost tracks. For example, we will attempt to extend energy range on both ends and to seed tracks with ACD hits. To this end, Leon has proposed that we form a recon working group to coordinate all this and focus resources on this effort. Eric Charles' visit to Pisa could be a step in this direction.

• GR Status: (Richard) Main item is RHEL4 validation. Liz has run systests; Michael Kuss is comparing RHEL3 and RHEL4 results. Observations so far may be found in Confluence. One which could stand some publicity is the large number of warning messages in the RHEL4 builds, many having to do with type mismatch. Package owners are urged to scan RM output for a recent RHEL4 build like this one. In particular, look at the count in the Warnings column for your packages and, if it's large, follow the link to the compile output.

  (Heather) Newer fsw releases like 2-2-0 will no longer support RHEL3. To date there have been no noises about 64-bit support for either RHEL4 or RHEL5, only 32-bit.

  enums bug (Leon) In Gleam the same bit is assigned to the TEM bug as to phase error. The word containing it is not in the merit tuple. (Anders) But it is in SVAC and is used. Almost all the time the bit is set because of a TEM bug, but not necessarily always. As things stand now, we can't distinguish.

  Failure modes (Richard) There has been a recent revival of interest in this topic. Leon has verified that Tracker failure mode code is ok. Sasha is looking into current state of CAL. ACD has nothing yet; Heather will look into writing something appropriate. (Anders) would like to know, e.g., whether rates will go way up if tiles used in veto fail.

• FSSC: (Eric) No news here. (John) has been handling data server issues [apparently none of them serious. ed.] as they arise, now that the data server is seeing real use. There have been 1076 queries since the data went public about 15 days ago. There was a spike at the start and another one after the weekend. Most queries so far have been broad w.r.t. both time and location.

• Documentation: (Chuck) sends the following:

  The revised pyLikelihood Tutorial has been posted to the web and linked in from the workbook. This tutorial also documents the use of the Data Viewer when creating the xml source model and incorporates the Diffuse Group's background recommendations. The only major gap is the lack of documentation on function parameters, which Jim has agreed to provide when time allows. During the course of this project, I also updated a number of other pages, as I came across problems. These updates included the Astro Server Help page.

  During the Collaboration meeting last week, I was also finally able to incorporate Masa's latest updates to the pulsar tutorials; this went surprisingly well.

  Going forward, the major priorities on my "ToDo" list include SCons, pointlike, and PGWave.

  (Richard) There has been a new spate of problems with passwordless access to CVS and missing tokens when doing a straight ssh — the usual issues practically all developers have encountered at some time — probably because there is a new crop of iLat users. Instructions for this set-up need to be more prominent in the workbook. (Chuck) will see to it. (Toby) Perhaps iLat could be served by the Installer. (Eric) Should FSSC be distributing it? (Richard) It has a mode where it talks directly to the Astro Server, but it may also be used with purely local data so perhaps, someday.

  (Richard) Does FSSC include or have plans to include gt_apps [in sane], the generic python interface to the tools? (Eric) Will be included in a post-release release, within a month or two at most.

• Core talk:

  • New Root (Heather) There is a -gl5 which addresses an xrootd client problem.
  • Promotion procedure (Toby) How I promote a package to HEAD given the two build systems? (Heather) Procedure hasn't changed — yet. CMT is still steering. Edit the relevant container requirements file. However, RMViewer does now include a tag collector [see discussion below] so once SCons is in charge we'll have a better interface.
  • RMViewer (Navid) Version 1.0 is out and includes a tag collector, however it hasn't been exercised beyond his testing. Those who promote packages or make releases are urged to try it out, however let Navid know because for the time being these functions are nominally handled from the CMT side; tags made with RMViewer will need to be removed.
  • SCons wrapper and setup scripts (Navid) has been working on a full rewrite which will address all of the following:
    • Incomplete list in PYTHONPATH variable (found by Chuck)
    • Incorrect handling in case installation involves a supersede directory
    • Need for two flavors on Linux: one for bash, one for csh/tcsh
    • Ugly 2-step procedure on Windows
    The Linux part is all done. All that remains is to see if it's possible to build a single portable _setup for Windows instead of having to do it in two steps.
  • GR and SCons (Joanne) has been chipping away at it, making updates to SConscript, xxxLib.py and whatever else needs fixing for packages one by one. Changes have been committed for all but about 35 packages. Most have not yet been tagged. She has been stuck on G4HadronSim because it has a non-standard organization with includes that need to be public buried under src. Now that Tracy is back he can suggest a suitable reorganization.
  • ASP and SCons (Joanne) Have begun to SConsify as proposed: ST release is in base directory; ASP is in supersede. I've run into several problems so far (including need for smarter wrappers and _setups being addressed by Navid) but nothing intractable.
  • RHEL5 externals (Heather) Kim is working on it

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:48:05 -0700