Core Minutes 9/24/2013

• FSSC: (Joe) is working with Jim on rspgen changes for the new CALDB stuff. The data release has been delayed till Thursday.

• Reprocessing: (Tom G.) Pass7: With the release now scheduled for Thursday, there is about a week of backlog to do.

  Pass8 reprocessing has passed the 61% mark, hence 3 whole years are complete. Standard procedure (whenever a year boundary is crossed) is to halt ROOT generation temporarily in favor of quickly catching up with FITS generation to make the data more generally accessible.

  In other news, the batch system suffered a melt-down last night from which it has not yet recovered. It affects jobs needing afs tokens. As a result reprocessing is at a standstill; most likely L1 is as well. [There was a note from Renata shortly after 8 AM Pacific that this problem has been fixed. ed.]

• Geometry: (Tracy) Leon made a first attempt to add more material (by changing composition of ACD support structure) between tracker and ACD so that MC would better match real data. The changes were in the right direction, but not sufficient. Before leaving on vacation Leon showed Philippe how to adjust the xml mass model himself. Philippe has now tuned it to optimal values [but it appears these changes have not yet been committed to CVS. ed.]

• Pass7 L1 release: (Heather) There is a new candidate release out there, largely motivated by the desire to use the same versions of shared packages (such as astro) which ST uses.

• New flavor: (Joanne) Before leaving on vacation Leon laid out a plan — first just make the new flavor (L1current) entries for those calibrattions needed by L1; then if all goes well do the rest after returning — and made the first batch of entries. However, testing of it by L1 folk actually required that calibrations for a slightly earlier period be available. Leon was able to give me instructions remotely on how to proceed. It took several rounds, but we now believe we have a complete set (back to Aug. 2008) of L1current calibrations.

• Core talk:

  • Windows (Heather) Bill reported a stack-overflow issue with G4 (seen by Tracy as well a couple years ago) which only manifests itself during long-running jobs. There is a fix for it — rebuild with an additional link-time flag — and it also is fixed in later GR versions. It does not seem worthwhile to advance versions just for this Windows-only issue so Heather will make a rebuild.

  • Mac Pro (Tom S.) has determined what compilers, etc. we need to install to be compatible with FSSC. The next question is: how to we make it happen? (Heather) This may be covered by the support we're entitled to from David Leung. She will send Tom an appropriate email address since he is in the best position to describe what we need.

  • GR tracking (Heather) We need to do a better job of tracking changes such as those connected with the L1current calibration entries and the changes to the mass model. This seems like a job for JIRA [Leon opened a JIRA for the mass model but there have been no updates for recent work. ed.]

  • Release diffs (Heaather) We would like to have at least release note diffs as the CMT used to provide. The effectiveness is of course dependent on how conscientious people are about entering release note comments. There have also been requests for full source diffs, probably on a per-package basis. Requirements will be firmed up in a future meeting. Can Jenkins be made to do this? (Tom S.) Yes, in fact the regular RM (non-Jenkins) system could do this as well. (Heather) Ultimately we probably want to move to Jenkins-only builds, even for Linux. That's another discussion.

previous

minutes index

next

J. Bogart, Last Modified: