ScienceTools: (Jim) No report this week.
FSSC: (Eric) Scheduled outage will be delayed until at least next week due to weather. Last week we found the reason for the Mac failures: it had to do with changes in the HEASARC script. This is now fixed. (John) In fact they were in fact not exclusively Mac failures; they only showed up as such because of changing test procedures.
Documentation: (Chuck) sends the following:
The newsletter has top priority this week, and is scheduled to hit the streets on Thursday. Next highest priority is documentation for the generic pipeline, and no. 3 on the "todo" list is Pointlike.
Also he added a link on the home page for Release Manager II (SCons) web pages right next to the one for Release Manager I (CMT).
Support of Python code: (Toby) Doxygen already does the right thing. See e.g. this example from pointlike
Clarification/correction of discussion last week on SCons handling of python code: SCons will correctly install python packages as well as python modules.
It appears that the upgrade to SwigModule to eliminate the need for the .in files still needs some tweaking. SCons RM builds of pointlike and skymaps are failing. [Turns out this was only true in the last HEAD build, which does not contain the most recent versions. Newer LATEST builds are ok. ed]
GR: (Leon) Tracy put in a fix for CAL Overlay energy problem. It looked ok on Windows, but systests results seem a little odd. (Joanne) made a suggestion to Tracy on an alternate way to implement. [Post-meeting: Tracy tried it. It is more robust than the original quick and dirty code, but appears to have the same behavior.]
(Leon) Continuing the saga of chi-square from last week: Tracy has gone after this one as well. What he has done is to compare our normal pattern recognition on simulated events with one that uses the MC information to assign the clusters and energy to the track (existing feature, Tracy again!). What he sees is that, for high-energy elecrons, the strange behavior of the track chi-squared goes away when the "correct" clusters are used.
For the normal patrec, the probability to find the incorrect cluster increases along the track,and this is reflected in the chi-squared. The chi-squared also increases with angle... this could be a consequence of the cluster distribution becoming compressed, and therefore more confusing, at large angle, or it may reflect real physics, with more deltas making it to the detectors as the angle increases. We don't have a convincing explanation at the moment.
There's an issue concerning what "correct cluster" means in a EM shower. We're getting more convinced that a global patrec strategy that "follows the energy flow" may ultimately be the best way to characterize the shower, and to extract the track parameters.
[No, I didn't remember all that. Thanks to Leon for supplying the above text. ed.]
GR, Root and TMine (Heather) made a new Pass 8 tag, v18r2p1, which has the patched version of ROOT which fixed the problem seen with TMine. TMine itself is now at the HEAD of GlastRelease. Ultimately we plan to phase out GlastClassify in favor of TMine plus a to-be-written package to interface TMine to Gaudi.
Branches not (Heather) Kim is looking into the lack of branch creation. It used to be the CMT rhel3 tag builds would trigger creation of a branch tag. Kim thought he had switched that function over to rhel4, but it's not happening.
7.2 branch (Anders) requests a patch needed to reprocess LEO data and pick up some misclassified events.
Python versions (Heather) Most likely SCons will work with Python 2.6, but we don't have a guarantee. Therefore Kim will first make a 2.5.1-based version with the new add-ons, then will do the same for 2.6. The latter will be tested with SCons.
More externals (Heather) ape 2.6 is ready to go. (Joanne) obf for Linux rhel4, 32-bit, reorganized for building with SCons, has been installed at SLAC as version B1-1-3-gl3.
(Heather) has been working on Root 5.26A, which has the virtue of supporting gcc 4. It requires a slightly newer version of Gaudi than she had been planning to move to (to support Windows vc90) so she's had to redo some work there, but it's coming along. She will be upgrading GR HEAD to use the new Root version soon.
SCons activity
previous | minutes index | next |