Core Minutes 17 July 2012

Core Minutes 7/17/2012

Present: Joanne Bogart, Jim Chiang, Richard Dubois, Elizabeth Ferrara, Warren Focke, Heather Kelly, Michael Kuss, Julie McEnery, Leon Rochester, Tom Stephens, Tracy Usher little boat

ScienceTools: (Jim) There is just one new item: a new pyLikelihood tag enabling the flux density calculation for diffuse sources. See Science Tools Development Notes for details.

Keith will be tagging updates for a new EarthPhenom model later this week.
Disk order: (Richard) We have a quote for our next order which has come in at about the same as for the last order: 450 Tbytes for $200K. At current rates we'd run out by October so we need to get the new order going.
Server upgrades: (Richard) The upgrade of sulkys to wain-class machines is imminent.

The upgrade of the 25 glastlnx machines to 10 new machines, each about 3 times more powerful, continues to cause some consternation. We'll move to the new machines, keeping the old machines around temporarily in case it turns out the reduced number of machines is a problem. If so, eventually we would need to get more new machines.
Server glitches: (Richard, Heather) glastlnx03 and glastlnx12 are now fully back after recent hardware problems. glastlnx14 was the next to go but is now back online after its hard drive was replaced and the OS reinstalled. glastlnx14 is notable for being the one and only rhel4 machine available for our GR builds.
Automated calibrations: (Leon) is working on automating dead strip calibrations.
Windows developer woes: (Leon) One problem he encountered, quite apart from the task at hand, was that the installer builds he's been using which are supposed to include all necessary binaries don't. He's had to make the rest of the build products locally.

(Tracy) typically fetches source and builds everything locally, but for recent tags there has been a problem with CRflux when he builds it locally. (Heather) It's a problem for the CMT RM as well. She'll look into it.
Pass8: (Tracy) Stephen Zimmerman was gone last week so has not yet made test data sets using the new GR tag, but that should now happen shortly.

Mysterious crashes which had been plaguing Bill on the UW terminal server seem to be gone with this latest tag.
Core talk:
- relh4-to-rhel5 migration (Heather) The second and last stage of database migration is done! CMT RM is running all right off the newly-migrated db. It's referring to the node as mysql-node01 rather than glastDB because it turned out the port had to change, so this move was not going to be transparent under any circumstances. It seemed just as well to use the preferred node name while we were at it.
  
  It's time to get off rhel4! There is still some work to be done validating the L1proc rhel5 SCons build, but developers should be using SCons builds, preferably rhel5 SCons builds. (Leon) may need some guidance.
  
  (Julie) There are flight tools, written by JJ, which we use on the ground to understand GRB handling and which were not ported to rhel5 when other parts of fsw were ported. (Richard) Is it not possible to run rhel4 versions on a rhel5 machine? (Julie) Gregg has tried; it apparently can't be done for this code. Sylvia or I need to be able to log in to glastlnx14. (Heather) That list is supposed to get smaller; we probably can't add anyone. (Joanne) In any case it would only be a very temporary solution: rhel4 is going away. If we need to use this code, JJ will have to be convinced to port it to rhel5. (Julie) will write up a description of what's needed. (Joanne) can act as on-site intermdiary of that's of any use.
- Windows password change (Heather) has taken place. Those who need it should be able to obtain it from escrow. She also needs to and will change the lsf password
- vc90 GR adventures (Tracy) The goal is to run Gleam with vc90. It's tantalizingly close. There are issues at various levels, some fully dealt with, others in progress.
  - The compiler — certain constructions in common use in our C++ code are not acceptable to vc90. Tracy has been fixing them as he encounters them.
  - CLHEP bug — In the CLHEP version we're using on vc90 there is an unitialized-variable bug which was causing a crash. He found a work-around. (Heather) There is a newer version where this bug is fixed (it doesn't exist in certain older versions, either, like the one we're using on Linux, but we can't use that version with vc90 for other reasons). She now has a build of this version for vc90 and is making a G4 build against the new CLHEP.
  - Dependencies — VS often thinks it needs to rebuild things it doesn't really need to rebuild. Fortunately, his new machine is fast so this is a relatively minor annoyance.
  - Gleam crash — Gleam now gets to the point of putting up the first event in the WIRED event display, but crashes immediately afterward in the clean-up of the TDS.
  - test_Gleam crash — Since various other test programs run, the TDS clean-up problem is not universal. Tracy tried running test_Gleam instead in order to side-step any problems related to the event display but it just crashed in a new way.
  - user unfriendliness — use of this stuff is not for the faint of heart. There is a fair amount of magic that has to take place before you can bring up VS successfully. But this was to be expected in the initial release.
  (Tracy) will document experiences to date. None of these things looks like a show-stoppers.
- Virtual machines (Tom S.) has released version 2 of the GR "bare bones" VM. It has some additional (system) packages which Joanne discovered were required and is in the form of an exported appliance rather than a tarball. It will be a tad simpler for users to start it up than version 1 was: just use 'Import Appliance' in the VirtualBox File menu. It can be downloaded via ftp It's about 200 Mbytes larger than version 1. Next up: GR standard. (Joanne) would like to strip some unneeded stuff from externals. See this table for a first cut.
- rhel6 (Joanne) found a little 64-bit bug in the ldf external. This is now fixed in v07-01-02 and test_ldfReader runs properly on rhel6 (64-bit; we don't plan to support rhel6 32-bit). The only remaining test program failures are for Gleam and userAlg. I tried running Gleam in the debugger — and it succeeded! However, there is strong evidence of a problem, most likely another 64-bit bug, in the TMineExt external. The program could fail or succeed depending on what happened to be in the upper half of some 64-bit pointer.
- externals paths (Joanne) proposes that we eliminate the compiler component in external paths for new versions of non-Windows externals. That is, the component of the path named after the version (v07-01-02 for the new ldf build) will have subdirectories with names like 'lib' and 'include' rather than having just a single subdirectory with a name like 'gcc44'. At SLAC the compiler information is already embedded in the path higher up. The extra component can cause problems for remote users trying to use a closely-related but not identical compiler. (Jim) would like to see this happen, but would prefer not to wait for new "naturally occurring" versions. (Heather) also would prefer more uniformity. [After the meeting she made new -glX versions with sym links pointing to the appropriate spot within the old version. This should be compatible with Installer operation and avoids a lot of copying.]
Announcements
- Congratulations to Tony J. and Seth for being named Senior Staff Scientists!

minutes index

J. Bogart, Last Modified: