Core Minutes 10/4/2011

• ScienceTools: (Jim) made some minor fixes, closing JIRA STGEN-126 Some packages have been promoted to HEAD; there will probably be a new release tag later this week. See Science Tools Development Notes for more information.

• FSSC: (Eric W.) No news.

• Pass7 reprocessing: (Tom G.) Concerning how: still waiting to hear about disk storage in order to make reprocessing happen at Lyon. (Leon) Concerning whether: it's very clear from Art's work that there is a significant difference, and biggest effect is due to the updated CAL asymmetry calibration. So in an ideal world it would be worthwhile to reprocess. But in the real world, taking into account factors like the impact on the Pass8 schedule, it's less clear-cut.

• GR/Pass8 (Tracy) The latest GR tag is what Bill needs to attack classification issues. There is no immediate need for another tag. Meanwhile Tracy is engaged in various projects..stay tuned for more detail next week.

• Documentation: (Richard) After the power outage last weekend we discovered hwo-to-fix information should not be just in Confluence, and that the list of maintainers was out of date. Heather has fixed the latter problem.

  (Heather) sent a query to the "How Do I?" forum but received no very useful suggestions; moving all the Workbook pages to a wiki is not a good use of resources. It's looking like the best path is to continue as we've been doing, editing the html directly.

• Core talk:

  • OS hopping (Toby) would like to know the timescale for the end of CMT support for GR as this will affect the UW terminal server. (Heather) GR will be moving to SCons-only builds on all platforms ultimately. Right now there are some more urgent issues along these lines:

    • rhel4 will no longer be supported at SLAC after this (calendar) year so GR has to build on rhel5
    • there is a proposal to move glastlnx* machines directly from rhel4 to rhel6. This could affect ASP, which in turn depends on ScienceTools. So, among other things, we would need a rhel6 build of ScienceTools. To this end, some space has been set aside for rhel6 externals and Kim is working on building them. FSSC had no difficulty porting their slightly restricted version of ScienceTools so maybe this won't be so bad.

    (Tracy) Since most of the Pass8 development is being done on Windows, getting Window (vc90) SCons builds to go work now rather than several months from now would be much more useful. (Joanne) was planning to revisit this now that obf B3-0-0 is well in hand on Linux and obf may be the only GR external remaining to port to vc90. "Now" is not in the cards, though.

  • Systests (Heather) Leon did a comparison of v17r35p23 with v17r35p24gr07. The latter has changes needed to build on SCons, but for this purpose he used the CMT build. The newer version takes much (10 times) longer. (Leon) it also writes a much bigger ROOT file. (Heather) understands where the change is coming from — new version of G4Generator — but not why it should have this effect. That advertised difference in G4Generator tags is a mechanism to get certain environment variables defined for SCons builds; it should have no effect on the CMT build.

    There is a separate problem in the p23 tag having to do with reading clusters back in.

    (Heather) would like to re-instate automated Systests, but this time to be driven by the SCons RM. She has asked Kim to look at the (long broken) implementation in the CMT RM as a start in understanding how to do this.

  • WIRED (Leon) Dima continues to work on merging plug-ins before tackling various requests from users. [See report from 9/20.]

  • SCons Windows RM (Tom S.) The Windows RM fell on its face but that seems to be largely due to lsf (lack of) recovery from the power outage rather than with RM itself.

  • obf (Joanne) Thanks to Tony W. and JJ's efforts, we can now get by with a much reduced set of obf libraries at link time: 8 instead of 26. A little more effort on both sides is going into cleaning up the use of include files, but what we have to date is adequate and we should be able to use the identical scheme on other Linux platforms (so far only attempted to build and run with rhel4-32bit since other GR externals are missing for rhel5).

• Announcements

  • SCons meeting this week at 10:30 on Wednesday.

• AOB

  • (Joanne) has been looking into profiling GR. This originally came up because Denis Bastieri asked about it with regard to porting compute-intensive parts of Recon to gpu's. It looks like it won't be too difficult to get detailed results with Callgrind, one of the set of Valgrind tools. She has output (from a little test run of 100 all_gamma's). Another tool, kcachegrind, provides a nice gui interface to the statistics gathered.

  • VM for ScienceTools development (Toby) See new entries in Tony's development notes. (Eric W.) is leery of any implementation which renames a library (or, what amounts to the same thing, makes a differently-named sym link) though there may be reasons to believe it's ok in this particular case. He will investigate other methods of satisfying dependencies for the external ape.

  • (Heather) M.E. pointed out that various additional packages need to be migrated to SCons.

previous

minutes index

next

J. Bogart, Last Modified: