Core Minutes 6/23/2009

• ScienceTools: Jim gave this week's report.

• FSSC: (Richard) There has been an email thread about how FSSC will handle releases, versioning, etc. once data goes public. How often do you think you'll be making releases? (Eric) A typical cycle would involve releases approximately every few months; would expect them to be more frequent initially. (Richard) It could operate somthing like the Pipeline for GR; usually is not using most current GR release. Updates just for Pipeline are made along a branch. (Eric) That would probably suit our situation. (Richard) What about a CCB or similar process? (Eric) Changes we need to make (e.g. to support a platform not supported at SLAC) are usually minor; CCB seems like overkill. However, there should be some form of documentation; e.g., a list of what was changed for any given release. (Jim) requests that all changes, even those to FSSC-specific branch, be first cleared with package owners. (Eric) Agreed.

• France trip report: (Richard) Memo of Understanding was one topic of discussion. We're asking Lyon to continue to supply computing, in particular for processing of high-energy electrons. They have informally promised us the use of 400 cores and have kept to that promise even though they are suffering from cooling problems and have had to cut back other programs. They will get a new chiller in the fall, at which time the issue will be re-opened for discussion.

  (Richard) The Galactic meeting in Bordeaux took place in a decommissioned church. In celebration of Fermi's birthday, everyone at the meeting got a bottle of specially-labeled bordeaux.

  He talked with David Chamont about skimmer. [See a more on skimmer in Core Talk below.]

• Pass7 reprocessing: (Richard) We don't understand Pass 7, so Pass 6 will be the vehicle for the 1-year release. This takes some of the pressure off. (Toby) is not feeling very relaxed about the high-energy recon problem (for Pass 6 as well as Pass 7). See his C & A presentation for details.

  (Tom) reports:

  The handful of problems experienced in the first cycle of the P100 reprocessing (and worked around) are being addressed. A couple of outstanding issues include still being worked on:

  1. leap second correction for data prior to 12 Dec 2008
  2. ROOT I/O error detection and exception throwing in the ScienceTools package tip

• Core talk:

  • GlastRelease (Heather) The latest GR tag, v17r32, differs from previous only in that it used the latest ROOT tag, 5.20.00-gl4. She can make a simlar update to an earlier GR if anyone so requests. (Tracy) Is there a debug version of the new ROOT? (Heather ) will make one.

  • skimmer (Tracy) managed to skim digi and recon files with the command line skimmer (all that's available in such a case) to match a skimmed merit file, but it was not easy or quick.

    (Heather) She, Tom and Tony can submit issues to David's bug tracker. Alternatively, we can make changes to our local copy of the skimmer source and propose that David merge them in. She will try to arrange a virtual meeting with David during Core Week.

  • Dumping RHEL3 (Heather) Latest production ROOT release does not run on RHEL3, yet another reason to dump it. (Elizabeth) has started on revisions needed for Sys Tests but still has a ways to go.

  • SCons RM (Navid) ST v9r15p2 now builds on the Mac (other than expected failure in st_facilities test program). The installer will allow him to download it but, it seems, no one else. He is adding diagnostics so that he'll get some help from logs of other people's failed attempts.

    There is still a problem on RHEL5: a compile failure, complaining about a missing #endif which is actually there. (John) believes Eric encountered a similar problem and found a way around it. Navid will drop him a note.

    (Heather) noticed that a Windows build started yesterday morning appears still to be in progress. (Navid) will check to see if it's hung or just very, very slow.

  • RMviewer (Navid) has come out with a new release, 0.6, which features syntax highlighting for errors and support in its installer for zip as well as tar.gz files. However this last feature is untested and, at the moment, untestable since there are no zip files to download and install. He'll be coming out with 0.7 soon.

  • Externals (Heather, Navid) are getting us moved to Xerces 2.8.0. For Macs we need a patched version, 2.8.0-gl1, to handle the byte order problem. We'll use 2.8.0-gl1 for Linux as well. Windows doesn't need the patch and we typically use binary distributions so there we'll probably stick with the official 2.8.0 release.

    (Emmanuel) is working on python 2.5.1-gl1 fro RHEL4 64-bit and is nearly done. Recall the -gl1 release entails several third-party packages. See this Confluence page for details.

  • Nagio (Emmanuel) is writing up some documentation in Confluence to help us out when he departs on July 10.

  • SCons and Windows (Joanne) The last couple features on the GoGui-for-Windows list — bringing up terminal window with proper environment variable settings and bringing up Visual Studio with one of the GoGui-created solution files — have been implemented. See screen shots by clicking on the links in the June 18th entry of the Windows log. These features are implemented in a new GoGui release, 0.9.5, available via ftp. (Heather) Is it possible to try this out? (Joanne) Possible, but not easy. Volunteers will need

    1. GoGui 0.9.5
    2. Visual Studio 2008
    3. SCons 1.2.0.d20090223
    4. A somewhat-modified SConstruct and several .py files to be added to the site_scons/site_tools directory (or replace production files already there)
    5. Mods to SConscript for any package for which project and solution files are to be generated.
    6. Some hints concerning a couple undocumented GoGui features of particular interest to Windows users.

    Some of the code in #4 could and probaby should go into production now. [Just did.] Other parts should not because they will only work with the latest SCons tag. The news on that front is not good. I can find no ETA for SCons 1.3.0. And current plans from SCons Central for 1.3.0 do not include a fix for the bug I submitted, concerning use of scons --interactive on Windows. We can survive without it, but its lack makes GoGui on Windows slower and clumsier than it would otherwise be.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:45:49 -0700