Core Minutes 6/30/2009

• ScienceTools: Jim gave this week's report.

• FSSC: (Richard) Confirming that v9r15p2 is the target release to be distributed when data goes public in August. (Eric) Correct. FSSC is currently working on a bug affecting 64-bit. We have a "fix", but it doesn't work on 32-bit platforms. The real fix, when found, will go in CVS on a branch of v9r15p2. (Richard) Commits to this branch will only be made by FSSC folk, who will also be responsible for merging fixes into the main branch. There will be no formal CCB-like procedure on the LAT side overseeing work on the branch.

• Documentation: (Chuck) sends the following:

  Having just returned from vacation, I'm working on getting the newsletter out before the holiday weekend, then it's back to pyLikelihood. Masa has also updated the Pulsar tutorials, and I'll begin incorporating those updates as soon as pyLikelihood is completed.

• GlastRelease:

  • (Richard) believes Eric Charles is working on getting TMine to the point where it can generate classification trees. There is a worry that at some point Insightful Miner may fall off a cliff.
  • (Heather) has made a tag [v17r31p2] along a branch for Eric's new ACD code. It builds all right; needs to be checked for validity.
  • (Tracy) is looking at high-energy PSF issue Toby raised a week or so ago. Of the 12 events he has looked at in detail, 3 are clearly mistracked. The CAL axis is pointing the wrong way and the search region around it is too small to find the correct tracker hits. Instead other hits are found and so tracks are bogus. If the search region is enlarged, the tracking works correctly. No idea yet why the CAL axis is wrong. Position is pegged at one end or the other of the crystal, already a problem.
  • (Richard) Since Pass7 will not be recommended at time of 1-year release, focus returns to Pass6. Some reprocessing is needed. Tom will work on it when he returns.

• Core talk:

  • Dumping RHEL3 (Elizabeth) has seen differences in Cal digi behavior which are too significant to be readily explained by differences in random number generation.

  • Disk space (Leon) suggests backing up all the Override files and copying newer ROOT files to xrootd. Then it should be safe to delete.

  • SCons and Mac (Heather)The Mac user build of v9r15p2 can now be readily downloaded using RMviewer. (Richard) Download works and tools can be run with wrapper scripts. However there is a problem accessing the tools from python. He's in touch with Navid about it. (Navid) Most likely Linux has the same problem. [Update: There are a couple issues involved. One concerns both Mac and Linux and is readily fixed by installing a couple files in their rightful places and adding a directory to PYTHONPATH. The other is Mac-specific and has to do with the distinction on Mac between "regular" shareable library and loadable library (which is what the wrapper libraries generated by Swig are, or should be). Navid is on the trail of a fix for this one as well.]

  • SCons conversion — user issues (Heather) A couple came up at the lengthy meeting last weeks:

    1. Developer release What would you like to see in it? In the past some have wanted to minimize release size; others care more about avoiding rebuilding. (Joanne) proposes the developer release include everything but the build directories. In particular this includes the full source check-out and all build products which get installed (libraries, executables, wrapper scripts). It does not include object files; it's her understanding that, once such a distribution is unpacked somewhere else, SCons would try to rebuild those even if already present. (Navid) This is the way SCons behaves by default; it might be possible to override this behavior. (Jim) It is. (Heather) will send out an email to interested parties; they'll have a week to respond.
    2. Tagging conventions We would like to move to the new convention: xx-yy-zz rather than vXrYpZ. Users would universally use the new convention, preferably via the command line tool stag, available via ftp or on central SLAC linux at /afs/slac/g/glast/applications/stag/@sys/stag, which updates SConscript and prompts for release.notes input as well as making the tag (the same holds for tagging from GoGui). It's of course also possible to make tags directly with CVS. In any case, RM will then be responsible for making the corresponding CMT-style tag. Those with objections should speak up; otherwise we'll make this move in a week or two.
  • RHEL5 (Navid) Compile failure in pointlike (complaint about a missing #endif which is actually there) is still an issue. Contrary to earlier reports Eric has not encountered this problem at FSSC and so has no suggestions.

  • RMviewer (Navid) Latest version can now download source (just tar.gz for now; zip is coming) and will download externals as well as user releases. It's available via ftp.

  • SCons and Windows (Joanne) No real news this week, just cleaning up.

  • Externals (Heather) A new version of ape, 2.5.1, with a bug fix of interest concerning pfiles is built and available for all platforms. It will be integrated into CMT and SCons builds ASAP.

    We'll also be picking up Xerces 2.8.0 (Windows) or Xerces 2.8.0-gl1 (Mac, Linux). There is no particular urgency in this case; it's only needed for versions of Studio newer than 7.1.

    (Emmanuel) has built python 2.5.1-gl1 for 64-bit Linux.

    (Heather) is starting to look into upgrading Gaudi. It's not pretty. They officially support gcc 3.4 and Studio 7.1. They also use a newer CMT that we do. And there are interface changes.

  • Nagios (Emmanuel) First draft of the Nagios documentation should be done in a day or so, at which time he'll give it another going-over and solicit comments. The impatient can get a glimpse now.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:44:59 -0700