Core Minutes 10/6/2009
Core Minutes 10/6/2009ScienceTools: No report this week.
FSSC: (Richard) FSSC would like a little more formality in determining and defining exactly what they are supporting, etc. They're setting up an advisory group which will include Eric, Seth, Richard, Jim and Julie.
(John) Eric has been working on building on Ubuntu 9.0.4. There have been lots of minor issues thanks to newer gcc version (4.3.3.) RHEL5 has gcc 4.1.2. I've been mostly dealing with small reliability/stability dataserver things.
Documentation: (Chuck) sends the following:
The new SCons pages were posted to the web last week, thanks to some great inputs from Heather, Navid, and Joanne; these pages are linked from each of the workbook's main "splash" pages. Currently, I'm working with Tom on updating the "Using the SLAC Batch Farm" page, with the intent of documenting "Good Practices"; this is an effort to reduce the overloading of key resources. I'm also continuing to work on an rmfit tutorial, and testing some of the latest SCons builds in the process.
User disks: (Richard) u31 and u33 will become read-only at about the end of this week.
GR RHEL4 Validation: (Richard) Michael has discovered that a couple unitialized variables in flux are responsible for some of the differences seen between behavior of RHEL3 and RHEL4 builds. (Toby) will take a look.
L1proc branch clean-up (Heather) Philippe had suggested removing two tools from CalRecon which were obsolete. The associated CAL merit variables had been removed from the main branch. We were wondering if we could do the same for the L1proc branch (GlastReleasev15r47p12) - where we still have those variables. Indeed we can; they are unused by the classification trees and apparently really are obsolete.
New mass model: (Leon) As of the end of last week, he had a pretty good, albeit not perfect, representation of materials. The most significant piece of work left is to better represent volumes in the vicinity of the bottom tray. For that tray, closeouts extend beyond the bottom of the tray and there is some extra stuff — flexures, cables.. — in the interior. Yesterday he figured out the problem he was having with xml representation of the volumes. Now he needs details of the materials. Him impression so far is that active materials are practically unchanged from the old model; the differences are in inactive materials like walls and cables. (Richard) Is the next step to make a simulation request to Johann? (Leon) Yes.
PassX news: (Tracy) Is still scoping out foundations for Pass8. See a Confluence page discussing the global pattern recognition known as "vector links".
Yesterday he got the latest Pass7 xml file for GlastClassify and was able to use it immediately. It needs validation. (Anders) Marcus will be back on or about Monday and can look at it then.
Announcements: Congratulations to Zach (sorely missed ex-CAL software wizard) and his wife Crystal on the birth of Julia SoonAe Fewtrell!
Core talk:
RHEL5 batch machines (Heather) The Computer Center is working to convert the remaining RHEL3 yilis on the batch farm to RHEL5. As of this time, we cannot handle having RHEL5 in our glastdataq queue since there are outstanding xrootd issues on RHEL5 that are actively being worked on, and we would need to move to the RHEL5 build of the ISOC software release. At the present time, the scripts do not easily allow one to limit the machine type when submitting the job. For now, we will request that the RHEL5 machines be removed from the glastdataq queue.
GR (Heather) Tag v15r47p12gr05 includes an update to AnalysisNtuple where Leon patched the initialization of FT1EventClass. GR v15r47p12gr06, coming soon, will remove the above mentioned CAL merit variables. We hope to use this new tag for another round of RHEL4 validation data runs.
Failure modes: (Heather) Code handling failure modes for ACD is about ready to go; it's just a question of collecting everything needed into a new tag.
obf (Heather) Has the correct flight software build been copied over? (Navid) Libraries are copied. He ran into a glitch with includes. He plans to copy includes from the previous version instead, expected to be just as good for our purposes.
RM and such (Navid) ST has been recompiled with the latest ROOT (-gl5). Compile output looks the same as with CMT; a few extra unit tests fail, as expected. The only problem currently is on RHEL4 64-bit and he knows how to go about fixing it. It has to do with f2c on that platform.
RHEL5 is nearly ready to go,
Developer releases are being created. A couple lines need to be added to RMViewer to make them available.
The very latest SCons release, 1.2.0.d20090919, doesn't work on Windows with our implementation. One of our infrastructure files, processExternals.scons, includes a somewhat dodgy reference to SCons internals in order to get the bindexplib command executed along with shared library links on Windows, but the SCons internals are a bit different in this release. Most likely a similar workaround can be found for the new version.
Tagging (Joanne) Latest versions of stag and GoGui now will tag along a branch and will update the version line in the requirements file if there is one.
Leon had encountered a failure while using the new stag to make a regular (not along a branch) tag, but it was most likely due to a transient CVS problem. He was later able to complete a similar operation successfully. When running stag on a machine with unreliable CVS access, or in any situation where stag fails for an unknown reason, use the -verbose option. It may help to narrow down the problem.
Version display (Joanne) As of stag version 0.2.2 stag will display version at the start of help output:
jrb@noric10 $ stag -help
stag version 0.2.2
Syntax for stag:
stag -cvspath=path-rel-to-cvsroot -notes="text-for-release-notes"
(-major | -minor | -patch | -custom=version-spec)
...
GoGui 0.9.8 displays version from an "About" menu item.
healpix made SCons-friendly by moving some includes out from under src.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified:
01-Jun-2010 15:48:14 -0700