Core Minutes 1/18/2011
Core Minutes 1/18/2011ScienceTools: (Jim)
Data access: (Richard, Eric, Julie) The HEASARC site was attacked on Saturday. As of Wednesday or Thursday static pages were back. In particular, weekly datasets are again available.
(Richard) At SLAC the LAT astroserver stopped ingesting data as of the 13th; no idea why yet. [As of 1 PM PST today, astroserver is back in business.]
Documentation: (Chuck) sends the following:
I've completed fairly extensive updates to the Binned Likelihood Tutorial, and it's now posted to the web. The subject matter experts have been notified but, so far, there's been no feedback. My primary concern is that I may have performed some redundant, or even self-canceling steps. New gtexpcube2 Help and Parameter files have also been created and posted to the web, and these also need to be reviewed. Please refer to the LAT Workbook Change Log for links to the new files.
All files have been checked in to CVS and, as soon as they have the SME's okay, I'll create a new tag so Berrie can update the mirror site. In the meantime, work is continuing on updates to the Science Tools.
The next newsletter is due to be released on Feb. 18th.
Pass8: (Tracy) The latest GR tag, v18r8p3, cleaned up problems discovered in earlier tags in the series and appears to be fit for generating large-scale simulations. There is one remaining issue: random sequences are not quite repeatable. After a (variable) number of events, two supposedly identical runs will diverge. In particular this interferes with the task of creating simulations with and without overlays which are otherwise identical. One suggested work-around was to start with the same mc root file for both, but this hasn't been smooth sailing, either. Other simulations tasks are not affected by this problem and are proceeding apace.
(Leon) is concerned about the randoms issue. This used to work. Something has apparently changed in GlastRelease to break it.
New Pass7 tag First some background from Eric Charles, passed on by Heather:
We are looking to deploy pass 7 to the collaboration shortly. The only outstanding issue of some note is the detector model. The improved model that Leon has developed gives a few % better agreement between the front and back fluxes. It would be nice to be able to get rid of issue so that we can focus on the other things that are giving us disagreement between the front/back fluxes.
(Heather) Leon passed on the necessary tags and I followed the pass7 breadcrumb trail to toss in a few other patches made over the last few months, to produce v17r35p14. (Liz) ran systests, but not with the new geometry (it's not selected by default). Another round of systests is in order.
Core talk:
u15 (Tom G., Heather) is being eaten up by certain temp files, perhaps from CMT RM. We need to track down where these are coming from, delete those which are no longer needed, and perhaps routinely direct them elsewhere.
RM woes (Heather) Since shortly before the holiday shutdown distributions resulting from the RM builds have not been copied to u09 as they should be. They have to be copied by hand from the V disk or built independently from source. Now there is a new problem preventing the builds from starting up at all. This new problem may affect SCons RM Windows builds also. Symptoms are similar to what occurs when ssh files are missing, but this time they're not missing. [Subsequent investigations by Tom S. and Kim indicate that the jobs sit in "pending" limbo in LSF indefinitely.]
New Windows box (Heather) got a note from Neal (of Computer Center) that LSF is now installed. The machine is ready for us to do some testing.
Snow leopard (Heather) The clock is ticking! Less than 60 days to go. (Eric W.) may be able to help with a testbed. (Tom S.) has been working on his laptop. The FSSC distribution builds and runs there without any problems. Ours builds, but test_optimizers doesn't run.
SCons and TMine (Joanne) has switched strategies: she'll try breaking up the massive TMine library into a collection libraries, built by separate packages, depending on each other. TMine was meant to have this sort of structure. The TMine classes fall into 8 or 9 subsets. Unintended dependencies had crept in in a few cases, but Eric Charles fixed these late last week. Since then I've rearranged the code into separate packages and have been able to build 4 of the libraries (on Linux) with only minimal code changes.
SCons and Windows miscellany (Heather) has been building GR, v18r7p1 branch (includes code for new Gaudi and additional updates Joanne has made for SCons support) with some success. The G4Generator library builds with nothing worse than warnings, however the test program doesn't run. (Joanne) That's farther than I got. (Heather) rearranged the debug G4 library she had built (which included both shared and static libraries), eliminating dll's and renaming .a's to .lib.
libXm Different versions of this library are installed on different machines, depending on OS, possibly also more subtle factors.
SCons meeting Wednesday at 11:30 this week.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: