Core Minutes 9/6/2011

• ScienceTools: (Jim) A new package, BayesianBlocks, has been added to ScienceTools. There was a problem with tagging, now resolved [unraveled by Heather; the tag was along a branch other than MAIN].

• snowleopard ST build (Heather) Gino noticed that certain executables were missing from the snowleopard distribution he got from the Installer. In fact those executables aren't being built at all and there are complaints from SCons in the log. This has been the case for some time. The precipitating factor seems to be the move from python 2.6.5 to 2.7. There is a bug in SCons which may be relevant, but the patch for that isn't out yet, and besides it's likely that python 2.7 itself is the problem, so Heather will make a new HEAD, retreating to 2.6.5 for snowleopard, but not for other platforms currently at 2.7.

• Pass7 reprocessing: (Richard) If we proceed with it, it will be split between SLAC and Lyon. Andy Hanushevsky, father of xrootd, has taken an interest. He suggests moving xrootd outside the firewall will solve possible i/o problems. He believes Security can be convinced that it will cause no harm.

  (Leon) has no news yet on comparison; that is, whether we should proceed with the full reconstruction or not. There is a significant difference when one looks at all events between v15 recon and v17 recon, even though comments in release notes would lead one to believe otherwise. He will attempt to track down exactly what in the code might be causing the different behavior; over the weekend he was otherwise occupied (see below). It's not yet known whether there is still a significant difference if standard cuts are made, nor whether the new calibrations are sufficient reason for full reprocessing.

• System tests: (Leon) was working on this along with Liz (well past midnight in her time zone; thank you, Liz!) to add a couple histograms: one to count TkrClusters and the other to plot widths of the clusters. A combination of pilot error and systest idiosyncrasies made it tough, but we have managed to add them for at least the AllGamma test.

• GR/Pass8 (Tracy) Thanks to Heather for the new release tag, v19r3p7. It allows us to compare native G4 behavior to our homegrown multiple scattering. Work in Santa Cruz shows that native G4 includes some effects which our stuff is missing. The consensus (Franz, Bill, Leon, Tracy) is that we should make the new physics choice (QGSP_BERT_EPAX) the default. This will be done in the next GR tag, which will also include updates to the various recons.

• Core talk:

  • RMs on Windows (Heather) Windows builds on CMT are more or less ok. (Tom S.) SCons builds are failing because the information on how to connect to the database which is stored in the registry has been disappearing. This is managed through Qt. (Joanne) It's also possible to store the information in a file on Windows as GoGui does. But since we've also had problems with disappearing files this might not be any better. (Heather) At least a file could be made read-only.

  • SCons and GR (Heather) GR rhel4 builds are working in SCons. She would like to run SCons-built GR against CMT-built in systests. They should be identical.

  • Virtual machines (Joanne) first tried rhel4 as the guest os for the GlastRelease-VM project since GlastRelease doesn't currently run on any other Linux os, but VirtualBox documentation warns against old Linux kernels as guest os and, indeed, rhel4 took most of the host cpu but never allowed a login. rhel5 and rhel6, however, are fine and some of the pieces we need worked immediately and transparently: network, MySQL, and afs access (via OpenAFS). However, before this project can go too much further, we need to get GR running on rhel5.

  • tagCollector (Joanne) The new tagCollector which submits production tag collecting to batch and sends email to relmanlist was released last week and has been used several times since with no problems reported.

  • SCons and rhel5 (Heather, Joanne) Since SLAC plans to stop supporting rhel4 in December, closely followed by Redhat end-of-life next February, we need to get GR up and running on rhel5. The main obstacles are externals, specifically gaudi (Heather's bailiwick) and obf (Joanne's).

  • New Pass7 GR tag (Heather) is coming soon.

• Announcements

  • SCons meeting this week. Tomorrow (Wednesday), 10:30 AM PDT

previous

minutes index

next

J. Bogart, Last Modified: