Core Minutes 1/24/2012

• ScienceTools: (Jim) Enabled specification of diffuse model integrals via xml files, as opposed to the existing method that used environment variables (The env var method is still available). This will be incorporated in the next release, which will probably occur within the week. See Science Tools Development Notes for a record of all recent developments.

• FSSC: (Eric W.) He finished importing Science Tools v9r26p2 to the HEASARC build system, and there were a few minor issues which he will send to Jim Chiang.

• Pass7 reprocessing (Tom G.) It's starting! One more round of calibration adjustment between Leon and Sasha is needed. Then we'll be poised for the first milestone: reprocessing a month of data. That will give people something to look at and should give us an idea of the resources needed for the full run. (Leon) communicated with Sasha this morning. He will deliver brand new set of 18 or so CAL calibration files of various kinds which will need to be copied to the official place and entered into the database. He proposes another item be added to Eric's list: run Vela and stacked AGN skims for calibration validation.

• GR/Pass8 (Tracy) Planning for the engineering run is front and center. There was a meeting last Wednesday of interested parties. (Leon) The plan decided upon involves taking two 3-orbit data sets (2 limb-pointing, one standard rocking), one with standard config, one with new config separated by 54 orbits so that position of the instrument will be approximately the same for both sets (therefore similar background conditions). The goal is to determine whether the MC for the new config matches reality acceptably. (Tracy) It will be necessary to produce Pass7 IRFs.

  (Tracy) The next Pass8 Face-to-face is coming up at the end of February in Pisa. Johan saw crashes when processing solar data. Once this bug is fixed, we'll be ready to make datasets suitable for studies to be presented at the meeting.

• Core talk:

  • GR and new/old gaudi (Leon) His GR/Pass 8 work recently mostly has to do with moving to the new Gaudi. Pass8 development is now taking place on a separate branch without the new Gaudi as well as on the main branch, a treachorous situation for those not accustomed to it [everybody except Heather. ed].

    (Leon) Almost everything is now working for the main, new-Gaudi branch. There is still a problem with the crucial Overlay package and with the event display. (Heather) Understands the problem with the event display; the fix is on its way. For certain facilities, including random number service and event display, we were using a non-standard method to determine whether and by what packages the service was needed. This no longer works for new gaudi. She modified the random number service to use, the recommended method (an Observer) and needs to complete a similar upgrade for the event display. She'll be making a new tag on the old-gaudi branch for the Pisa meeting, as well as one on the main branch

  • GR and gcc44 (Heather) Johann has pretty much gotten GR to build with gcc44 (which rhel6 also uses), using rhel5 externals

  • SCons for rhel5, vc90 (Joanne) GR on rhel5 (which of necessity uses new Gaudi) has an event display problem, presumably the same as for CMT builds, so most likely Heather's fix will take care of it. Otherwise things are in pretty good shape, so she switched focus to vc90. There has been some progress — small but necessary Windows-specific fixes for CalRecon and lsfData among others — and discovery of some as-yet-unresolved problems. One is a failure in making a rootmap for G4Generator. It would be great if we could do without rootmaps altogether, but the jury is still out on that question. She made a trial build without rootmaps on vc90 to see if component libraries can still be loaded — and ran into another build problem, this one involving OnboardFilter and the obf external. However, the rest of the GR libraries used by Gleam did build; it should be possible to try out the loading test.

  • Systests (Heather) Systests need to be upgraded for SCons, but for now keeping ability to compare CMT and SCons builds. We'll need to validate rhel5 GR. (Leon) Systests need to be in CVS. Getting them to go for a given build is a tricky process and difficult to reconstruct. (Heather) agrees completely and has already committed it.

  • Infrastructure to-do (Heather) Richard asked for a list, and here it is. [Discussion ensued on several of the items.]

  • CMT RM (Leon) Windows RM builds have always been slow, but they are now in the outrageous realm, taking anywhere from 1 to 3 days. This is too long to be of much use to developers. (Heather) The extreme times are probably primarily due to the very slow (V) disk access. It may be possible to apply some partial remedies (e.g., move all external libraries to local disks). The more radical among them have a greater cost in time to implement and possible disruption. [Discussion continued offline in an email thread.]

  • SCons RM (Tom S.) Short summary: it works! Since the switch to using a local path to the repository (/nfs/...) on Linux plus some patches to RM code to make use of different dbs entries for CVSROOT for different platforms the problems we were seeing have not recurred. Also, since we're not using cron rather than trscron for the RM daemon, we can manage with a passwordless glastrm account.

previous

minutes index

next

J. Bogart, Last Modified: