Core Minutes 12/6/2011
Core Minutes 12/6/2011ScienceTools: (Jim) He made an upgrade to gtobsim concerning efficiency corrections in certain special cases; most analyses are unaffected. Yesterday he made a new ST tag, 09-26-00, incorporating all changes since Oct. 9. See Science Tools Development Notes for more information on recent ST activity.
FSSC: (Eric W.) has been working on the tip overhaul; a first version should be ready by the end of the week.
Changxue Deng made several minor upgrades before leaving which need to be tagged. (Heather) volunteers to handle it.
(Eric W.) A new hire, Changxue Deng's replacement, will be starting next year.
rhel6 (Tom G.) The upgrade took several hours, not just two as expected. It took 36 hours for L1 to recover and catch up, but catch up it did and all is well. Jim moved a couple cron jobs to rhel4. glastlnx01 and glastlnx02 are still at rhel4, pending a MySQL upgrade. Arash expects to do this upgrade early next year.
Pass7 reprocessing (Leon) It's still under discussion. His impression is that the decision will be to try it for a year's worth of data, then decide whether to reprocess all data.
Systests: (Leon) Next week Liz will have daycare and should be able to quickly finish up the automation of Systests. In the meantime he and Heather (along with occasional contributions from others) have been making comparisons by hand. There is one last issue to be addressed — warnings from the SCons build indicating that certain physics tables have not been found. [Some of the warnings were eliminated with a patch to G4Generator. It appears that the remainder also show up for CMT builds. We're ready to compare CMT and and SCons for tag 17-35-24-gr15.]
Random numbers (Leon) In order to be able to compare SCons and CMT builds on Linux, we changed the behavior of GlastRandomSvc to just (explicitly) use a single engine. This was happening anyway on Linux, but in such a way that the two builds were starting with different seeds. On Windows we had true separate streams, desirable behavior for developers. (Heather) We might be able to get this behavior on Linux as well if we linked to static rather than shared CLHEP. (Joanne) tried some experiments along these lines but was unable to create a build which would run because our G4 external is linked against shared CLHEP. G4Generator is linked against both CLHEP and G4; when I try to run Gleam the G4Generator component can't be loaded. We could perhaps get around this by creating a new build of G4 in which those libraries (many) using CLHEP Random services were linked against a static version of the CLHEP-Random library, but I would prefer a different kind of solution where we don't use a static engine. It remains to be seen whether CLHEP supports this. (Heather) We should have a meeting to decide on a course of action, preferably including Toby, who wrote the original code, and Michael Kuss, who has made recent upgrades, along with any other interested parties.
GR/Pass8 (Tracy) Everything is fine. Particularly notable is Philippe Bruel's work using standalone G4 simulations to better understand the effects of the gaps. This should pay off especially at ultra-high energies.
Concerning truncations — no real change in status since last week. People are examining implications of eliminating the 14-hit limit.
astrotools (Heather) A recent user question demonstrated we have some documentation work to do at a minimum. The Workbook recommends the installation under $GLASTROOT, which is quite old. That probably should be changed to the KIPAC distribution. (Tom G.) Users should note that the default KIPAC distribution is updated reasonably frequency, but they do keep old distributions. One can specify a paricular version by means of the .xrayrc file. (Jim) ASP is still using the version under $GLASTROOT (has to do with AFS tokens, not because of any dependency on this particular version) so we can't get rid of it just yet.
Core talk:
CMT RM (Heather) Michael Kuss has agreed to fill in as maintainer after Kim's departure next Friday.
GR, gaudi, and rhel5 (Heather) We're close to the finish line for rhel4 Systest comparison of SCons and CMT builds. The next step is to compare rhel5 and rhel4. rhel5 externals are ready to go. Heather is in the process of merging new-gaudi upgrades into the Pass7 branch (and also HEAD).
rhel6 and ST (Heather) Seems ok. (Jim) There may be an issue with readline. He will look into it.
SCons RM (Tom S., Heather) ST builds look fine. Heather notes that a GR opt build failed partway through checkout; the debug build of the same tag succeeded. Tom confirms that all checkout jobs have been modified to use the medium rather than short queue. He'll take a closer look at this failure.
VMs (Joanne) has made notes on most of what has transpired to date. She was able to get a sample program, using the C++ API, to run through some of its paces. It failed to attach a (virtual) disk, but this is not something we're likely to want to do via this API. Next I intend to test other features of the API which we probably will need.
Disk acquisition (Tom G.) is handling this in Richard's absence. We're looking for 400 Tbytes of usable space (equivalent to about 600 raw Tbytes) in a form we can afford. It will probably be in the form of Dell servers. We have 82 Tbytes remaining which, at the last measured burn rate, will last us until about Feb. 11th. The new space will be used for
EVO news (Heather) There have been rumors swirling around EVO's future, as it seems CERN has dropped support and has moved to a commercial product (Vidyo). Our contacts at EVO assure us that EVO will continue through 2012 and they are working on expanding funding through a commericial offering coming soon. We'll keep an eye on this breaking news story and provide additional information as it becomes available.
Announcements
NO SCons meeting this week
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: