Core Minutes 3/9/2010

• Trip report: (Richard) The HEAD meeting was held at the Hilton Resort on the west coast of the Big Island. About 400–500 people attended. The resort was Disneyland-like in its artificiality, but the natural surrounding were about as you would expect; see, e.g., a humpback whale and a sunset.

• ScienceTools: (Jim) spent the weekend fixing a bug that cropped up when using SummedLikelihood. He'll be making a new tag towards the end of this week. This will include SCons-related updates (current ST tag has everything needed for Linux but some patches for Windows and Mac are lacking).

• FSSC: (Eric W.) v9r16 is ported; making tarballs now. However there is a difference of a second between 32-bit and 64-bit builds, presumably having to do with handling of leap seconds. This is a show-stopper. Toby has been contacted.

• Documentation: (Chuck) sends the following:

  I'm making good progress on the pipeline documentation and hope to have a review copy ready by next week, and we should probably start a discussion of where and how we want to host it. As for the updates to the LAT GRBAnalysis tutorial to include an SCons build, I'm still holding off until we have released builds for all versions. No work has been done this past week on Pointlike.

• GR: (Leon) No news.
  (Anders) Nothing to say. [communicated via chat window. ed.]

• Disk problems: (Tom G.) sulky55 is still acting up, even after SCCS backed out a suspect OS patch. Best tactic at the moment is to apprehend users who are hammering the disk and retrain. (Jim) is surprised that pfiles are sometimes involved in such hammering since he believes they are only accessed a couple times per job. (Tom) Nonetheless, modifying pfiles access can make a substantial difference. Meanwhile, we are looking for recommendations from SCCS on alternate hardware.

• Core talk:

  • glastrm quota (Heather) The tags of Likelihood and pyLikelihood that Jim made didn't automatically trigger a LATEST build. Certain outputs from submitting a build go to glastrm's home directory. It appears the build didn't go through because that quota was used up. We erased a few old core files which was sufficient to let the build go through, but more extensive cleaning and reorganization is needed, or perhaps another volume. Also the mailboxes for both glast and glastrm tend to fill up. (Joanne) votes for another volume. It would be nice to understand where the core files came from, but they were not that large nor were they numerous. Over half of glastrm's puny quota (500 Mbytes) is dedicated to binaries RM must have, not leaving much breathing room. (Richard) More space is not a problem.

  • Beamtest releases restored (Heather) About 10 tags have been rebuilt on u09. They will be archived as well.

  • Externals

    • (Heather) The latest ROOT (5.26) has been built for rhel5 32-bit as well as rhel4 32 bit. The build process is a bit more complicated than for previous releases because Gaudi needs more things turned on. She will ask Kim to make remaining Linux builds of ROOT.
    • (Tom) There have been several iterations already of building the new python; we're closing in on a usable build.
    • Gregg and Jana (only just back at work) may be able to help with 64-bit obf.
    • The new Gaudi no longer supports the method we had been using to gather up tools of a certain type, as we had been doing, for example, in GlastRandomSvc and in HepRepSvc. This code has to be converted to use the new Observer pattern.

  • SCons, CHS and related matters (Heather) Jim Panetta made an SCons build of CHS and wanted to try it out on rhel5, but he was missing the ldf external. She decided to try to build it with SCons rather than CMT and, some hours later, emerged victorious. This should make it easier to build it on other new platforms.

    (Joanne) With our current motley build system, there is no easy path to producing an SCons-style release tag for CHS. When Jim tried it last week, he ran into a snag because the top-level CHS, unlike GR or ST, is not a pure container. The mapping between CMT organization and SCons organization is therefore different and not properly handled by the scripts which generate one sort of tag from the other sort. A second problem was that the old RM seemed not to be generating CHS builds, even the the web page indicated they were turned on. In the end, she had to tag most the packages belonging to CHS by hand with the new release tag, CHS-11-00-00. (However, mysteriously, CHS HEAD and release builds did start yesterday morning, nearly a week after they should have.)

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:46:35 -0700