Core Minutes 1/10/2012
Core Minutes 1/10/2012ScienceTools: (Jim) There is one new item: a new integration scheme available for unbinned analysis. See Science Tools Development Notes for details. This tag is in the most recent LATEST build. Interested parties have been alerted so they can compare binned and unbinned with the new code.
Disk acquisitiion: (Tom G.) We were getting perilously close to running out of disk, but on the very last day before shutdown our purchase req. for 5 Dell servers with 400 Tbytes of usable space went in. The servers are already at SLAC and cables have arrived. SCCS is working on bringing up the servers. And the associated storage is due to arrive on the 18th. We're out of the woods!
Pass7 reprocessing (Leon) Nothing much to report in the new year. People are gearing back up to work on it.
GR/Pass8 (Tracy) Everything is fine. Not much was done over break. Philippe is finishing up Cal recon work which should have significant impact on high-energy events. Tracy is concentrating on performance. Any of the new truncation schemes will result in more very large events.
GR progress (Heather) There have been a flood of Gaudi updates into GR HEAD which are still being sorted through, especially for Windows. She is hoping to have a fix in today for GuiSvc. Much thanks to Leon and Joanne for helping sort out some additional unit test issues for win and rhel5. Still working through Gaudi upgrade.. where we have builds for RHEL4, RHEL5 (SCons only) and trying to get windows to work again. The first run of test_Gleam in RHEL4 and 5 ran to comopletino but the numbers didn't look that great. ..so that needs to be understood further..perhaps more from Joanne/Leon?
(Joanne) Concerning the performance numbers which show up near the end of the output — effective area and figure of merit — I initially compared the rhel5 SCons output with GlastRelease-17-35-24-gr17 since that was a recent SCons build (albeit rhel4) for which test_Gleam ran to completion. The rhel5 numbers looked much worse. But they are quite similar to the CMT build of the tag v19r4p1 so the problem, if it is a problem, has little or nothing to do with SCons or rhel5.
(Leon) Output from test_CalRecon looks ok even though it reports failure. Overlay is still not working and we need it!
(Joanne) After some issues with externals were straightened out, there were a few packages which needed patches for the new (i.e., gcc41 rather than gcc34) compiler; not too many. gcc41 seems to be fussier about namespace designations. Overlay is the one serious problem remaining. test_Overlay, which links directly to the Overlay shareable rather than dynamically loading it, crashes immediately, even before getting to a break set on "main".
(Heather) The GuiSvc problem is due to it being a static lib. I'm fixing that and trying to sort out things on Windows, going back to gui lib, which is also static and sets things up appropriately for Lin and differently on Windows
(Leon) The current Windows HEAD build, started 2 days ago, is still running. Can something be done about that? (Heather) Some jobs don't terminate — especially on windows
(Heather) M.E. has requested a RHEL5 trial for L1proc, hopefully for next week, which would also require the Gaudi upgrade
Core talk:
CMT RM (Heather) Kim is gone; Michael Kuss has agreed to fill in as maintainer.
SCons meeting not (Heather) SCons meetings will now merge with this meeting, since at this point we should all be reasonably interested in SCons goings-on.
Queue mix-up (Heather) Neal Adams reports there are some rhel4 jobs being submitted to the wrong RHEL4 queue (I'm investigatig whether that is CMT or SCons; I think it's CMT) which is weird, since we've had RHEL4 builds go through correctly for both SCons and CMT
rhel6 f2c (Heather) Johann reported trouble using the f2c external on rhel6.. apparently there was a problem with it not being build with -fPIC. I've fixed that and will have to create a new version fo the f2c external for the ST RHEL6 builds.
SCons RM (Heather) SCons LATEST builds have not being functioning since Jan 4..apparently due to an issue with the cron jobs and specifically the releaseManagerDaemon. Kelly Tom S is aware of the issue and can run the daemon by hand ..and that's how we got LATEST builds yesterday. (Jim) No word from Unix Admin yet? (Heather) they claim nothing has changed..and they see nothing wrong. (Joanne) hah! (Heather) I'm hoping we can find another way to make the case.. this started after they reset the glastrm password on Jan 3rd.
And we are still trying to sort out with them the longterm handling of the glastrm account.. sharing the password via escrow is apparently a no no now. (Joanne) What do they propose, since a passwordless account doesn't do everything we require and we need a back-up person with access? (Tom, Heather) They would prefer no passwords at all, but we can't change trscron with a passwordless account. Alternatives are:
(Heather) Booker is leaning towards 1., but we don't have a clear commitment yet. Meanwhile, I'm hoping that if Tom S can reset the cron job settings so that this issue with the daemon being unable to connect to CVS will be fixed.
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: