ScienceTools: No news
FSSC: (John) is still awaiting test files in order to verify that they can handle parallel ingest. (Warren) will remind himself of what's required and supply one this week.
(Eric) is finishing up import of v9r17p0; strictly routine.
Documentation: (Chuck) sends the following:
Most of last week was spent working with Dan on the Pipeline Installation Guide being developed in Confluence, and I believe he will have additional material later this week. Tony also mentioned that Max is working on some documentation for the web frontend.
On the WB front, Tom has offered major help in documenting the attributes and methods for the pyLikehood Binned and Unbinned Analyses. At the moment, I'm working to come back up to speed on both. In the process, I've also come across some of the earlier documentation that is still web accessible using the search command, but has not been flagged as archived.
Lastly, I've also started developing a new section of the workbook that will better accommodate the new software infrastructure tools.
GR isues (Richard) At yesterday's C & A meeting there was a plea from the calorimeter guys. It appears that longitudinal calibration varies significantly with time, hence they are even more eager to reskim the 8 months of data from which heavy ion filter events had been discarded. (Tracy) To recap: GCR recon currently gets its information from offline obf, which is doing the wrong thing for real data. The quickest way to get around this is to change it to use LAT obf instead. Johann is looking into this; it will take him a little time to come up with a plan which also correctly handles MC, when there is no LAT obf. No decision yet on which path to follow longer term (to fix the problem with offline obf). (Richard) How long would the reskim take? About a week? (Tom G.) Yes, perhaps; it's not a huge job. Need to make some tests.
(Richard) See the C & A schedule of projects for the complete picture. Note it includes reprocessing with the new GCR calibs, expected to take months.
Pass8: (Tracy) Not much to report. There will be a full meeting this Thursday. Last week's was canceled due the power outage in Pisa among other things.
(Leon) is working on the geometry, in particular a better description of materials. He's hoping to have some results in about a week, but is making no promises. Issues to be addressed include:
ST binned analysis bugs: (Tom G.) Rsp jobs now have too many photons for unbinned analysis — that's good news in a way — but the move to binned has uncovered various problems. gtsourcemap, when given a model with lots of sources, or with sources having special characters (e.g. blanks, +) in their names, sometimes crashes; other times it doesn't crash but produces incomplete or possibly corrupted output. Not clear whether this is a problem with gtsourcemap itself or with underpinnings like tip. Jim recommended changing names to eliminate the unusual characters. That eliminated the crashes, but the output was incomplete. He has also seen a possibly erroneous range error from gtmodel when it starts up. He reported it to Jim Chiang and James Peachey.
Windows RM builds now working (Tom S.) The problem — inability to run ssh from certain machines — went away after a routine reboot as happens periodically to install security patches. No one knows why this fixed it.
CMT rhel4_gcc34 oddness (Heather) Results from builds are not always properly entered into the database. The rhel4_gcc34opt builds don't have this problem. Kim is looking into it.
Python upgrade: (Heather) Python 2.6.5, including scipy, pyfits, and the kitchen sink, is now available for all platforms except Studio 2003, which will stay at 2.5.1. ST LATEST builds are already using it. (Eric) When will we have a tag with the new python? Will it be v9r18 or still in the v9r17 series? (Heather) Guesses we'll have a tag within about a week, but timing as well as naming is up to Jim.
ROOT: (Heather)New patched ROOT, v5.26.00a-gl2, has been built for all platforms, including snow leopard.
Gaudi upgrade: (Heather) Just LdfConverter and HepRepSvc remain to be upgraded. Now that the flurry of python activity is over maybe she can get to it. (Richard) What about a retrofit for Pass7? (Heather) had envisioned the upgrade for Pass8 only, but a retrofit would be possible. Only about 5 packages have substantial changes; many more had to be touched.
Exit sourcelike (Heather) sourcelike will be removed from ST. All its functions are handled by pointlike.
SCons topics
Installers (Heather) Tony has released a new version of the gui scons installer. It provides access to all three distribution types, not just user. (Tom) Has integrated Tony's improved functionality into the command line installer and fixed some minor bugs reported by David Landriu.
Creating Windows distributions (Tom) can create them by hand, but batch jobs executing the same command fail because the "=" between option and value is getting eaten somewhere along the way; the resulting command is gibberish as far as SCons is concerned. Even manually-created distributions are not available since they go on the V: disk. We need to get them to a publicly-accessible location. (Richard, Heather) u05 maybe? The old RM was able to do this.
HEAD and Release tag builds (Heather) They're working again. The root problem, discovered by running the relevant script by hand, was an old stranded lock file in the CVS repository.
Tag collector for SCons (Joanne) Enlightening email from Navid revealed it to be not as grim a situation as previously thought. The RMViewer tag collector probably is usable though there are display bugs. Furthermore, the core part of the procedure just consists of a few CVS commands. She has written a script which will probably serve, described in Confluence. It can make a new release tag from the most recent HEAD or can make a new HEAD incorporating specified mods to the previous one. (Heather and Toby) point out that this latter function is used not just by Heather and Jim (container managers for GR and ST, resp.) but also by other ST developers wishing to promote packages they own to HEAD.
GR and SCons (Joanne) has written a script to reorganize the obf external on Windows into a form suitable for SCons (modeled on a similar script for Linux) but hasn't been able to test it because of other niggling problems with certain externals on Windows. She hopes to have worked through them all later today.
CMT shut-off target (Heather) What with various problems and vacation schedules, it's not going to be possible to shut off CMT builds of ScienceTools at the end of this month. We're now aiming for Core Week.
previous | minutes index | next |