Core Minutes 12/7/2010
Core Minutes 12/7/2010ScienceTools: (Jim) There have been two bug fixes and one enhancement since last week. The bug fixes concern 1) latResponse PSF integration in certain special cases and 2) optimizers occasional large negative TS values seen with newMinuit. The enhancement (pyLikelihood) added a new function for binned analysis, allowing the caller to set energy bands directly. See the last three entries (starting with the one dated Nov. 30) in the Science Tools Development Notes for details.
FSSC: (Eric W.) has been porting v9r18p6 to various Linux platforms.
(Tom S.) The Goddard Fermi website is back! Security was willing to let us do the verification for the most part, rather than insisting on checking over everything themselves, which would have taken much longer.
Documentation: (Chuck) sends the following:
The Likelihood tutorials have all been checked into CVS and are awaiting an okay from the SMEs; Liz was reviewing them while flying back from Japan yesterday. I'm making another editorial pass through both tutorials and all of the support documents, most of which were ported from the Cicerone and "SLAC-centricized" and are linked in from a new Likelihood Overview page, also ported from the Cicerone. Looking towards next week, I've also created the basic framework for the last newsletter of the year. Looking towards next year, I'm reviewing the rest of the Science Tools to determine update priorities.
Pass8: (Tracy) The Pisa guys have lots of plans for further enhancements to Cal recon, beyond those already in place. The work to come will be quite extensive and will involve revamping tds output. However at the moment they are stymied by a crash.
Tree-based pattern recognition code has been committed to CVS since Tracy needs to run batch jobs to test adequately. He has tried running on real data. For the most part it does well, but events with bad Cal axis direction are a problem. He's working on a new approach.
(Leon) continues to work on truncation. It definitely impacts high-energy events. He also has embarked on another event display project to allow the viewer to toggle x hits and y hits independently.
(Tracy) There will be a Pass8 workshop in Pisa Feb. 14-18. At a minimum Tracy, Leon, Alex and Eric will go from SLAC.
Johan wants to make a run with alpha particles but it's crashing. He suspects Overlay.
Pass7: (Anders) L1 will move to a Pass7 release in the forseeable future
Core talk:
Obf (Heather) There was a meeting on Dec. 2 and things are looking good! The bulk of the work necessary to support 64-bit and rhel5 has been done. About a month of JJ/TonyW/Owen time is still required for adequate testing. They estimate this will happen around February/March. See also detailed meeting minutes.
vc90 and debug (Heather) has made debug builds of Gaudi and ROOT for vc90 and is working on Geant4. Joanne made debug Xerces libraries and Kim is looking into some of the others. (Joanne) Using debug versions of Guadi and ROOT (didn't even bother with debug Xerces, even though it was available), she was able to build several packages including GlastSvc and AcdDigi and successfully run their test programs. G4Generator had linking problems; we're hopeful they'll be resolved when a debug G4 is available.
SCons RM LATEST (Tom S.) It had been the case that SCons LATEST builds, unlike CMT LATEST builds, would pick up the very latest (last-created) package tags, regardless of form of tag name. This has now been fixed so that only tags of the form packageName-DD-DD-DD where the D's are (possibly distinct) digits. This is the form reserved for tags along the main branch.
SCons RM and GR (Joanne) Now that Tom has fixed a bug having to do with package selection for LATEST, these builds are looking pretty good. There are relatively few errors. Some of them also show up in CMT builds.
Snow leopard (Tom S.) With Jim's help the problem with Snow Leopard builds has been further narrowed down to test_optimizers and libf2c. Heather made a new native build of this library; we'll try using that.
New Windows box (Tom S.) The ball is in our court. We need to install SCons and RM programs. However, there still are concerns with disk access speed since it seems we cannot just access a local disk and VS 2008 installed. We need to add our stuff. (Tom) That would be SCons and all the RM programs.
TMineRelease (Tom S., Heather, Joanne) Neither the user nor the development distribution for vc90 is usable because the build itself did not complete successfully. The rootcint command fails because it can't find a file. It's difficult to see why since the very same command works on Linux.
GR and vc90 (Joanne) While awaiting the debug Gaudi build I've attempted to build several GR packages non-debug, just to see what goes wrong. So far, not much; for example, GlastSvc and AcdDigi build correctly and their test programs run. There are compiler errors in other packages. This is not surprising; new compilers tend to be fussier than old ones.
TMine (Joanne) has been engaged in discussions with Alex concerning TMine-within-GlastRelease and TMine-within-TMineRelease. Currently in both contexts it builds but the test program fails. This is because TMINEDATAPATH is not defined. There is a standard way to do this (call commonUtilities::setupEnvironmet( ) but Alex would prefer not to increase the dependence of TMine on facilities. To be discussed further, perhaps at
SCons meeting Wednesday at 10:30, as usual
Disks: the saga continues (Richard) Recall we had ordered 300 Tbytes of disk from Oracle at bargain prices since the hardware was about to be discontinued. We have now heard back from them: they won't fill the order because they don't have any more of the discontinued disks, so we have to look elsewhere. DDN is a logical candidate; LCLS is using their stuff. We have about 3 months to come up with something (75 Tbyte left, using about 25 a month) so long as KIPAC and BaBar don't demand we return their borrowed hardware sooner.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: