Core Minutes 11/27/2007

• ssh: There are still some glitches, particularly for Windows users.
[post-meeting: the Workbook page has been evolving to address them. CVS from command line and from MRvcmt is now working for the most intrepid. See Tony Johnson's page in Confluence for instructions. And Navid has found a solution for RM on Windows.]
• GlastRelease/Big run: (Richard) We're nearly on our way. Working through holiday Friday, Eric C. found one last problem in ACD code. A fix was incorporated in v13r5. We're waiting for final validation from Bill. Tom has task definitions for the background run are ready to go. We're 3 weeks late according to the original schedule but right on time according to the newly-revised one.

  We're hoping for the extra core allocation to speed up the runs; also Lyon is nearly ready to go.

  (Heather) Is the new once-per-run random seed setting in job options? (Richard) Not yet. That needs to be tried. (Toby) Once it's been tested, code should be changed to make it the default.

• Meta ROOT files [Thanks to Tom for writing this up]

  Experimentally, it seems that requested meta ROOT files are not created unless there is something to put in them - in contrast to other ROOT files which seem to be created even if they end up with zero length. The problem is the staging code expects there to be a file if one is requested (I cannot think of a reliable way to inform the staging code that a particular output file is "optional" or that "we just didn't get one this time" without throwing an exception).

  For the moment, we will simply disable the meta ROOT files altogether (they purportedly contain junk in this GlastRelease). Heather is, I think, looking into the possibility of making the meta ROOT output consistent with other ROOT file output.

• Composite Event Lists (Heather) CEL will go into HEAD for Tracy to use. (David) is not quite ready to bless the code, but he would welcome feedback from other users.
• Skimmer [Thanks to David for writing up the following]

  A new skimmer release has been installed. It can now skim cal/svac/gcr tuples, and in theory you can also use those tuples for the definition of the cut (and not only for the merit tuple). Some low levels tests have been applied, but I am searching for and will support users who could try it with true data. This implies the use of the command-line interface. The web front-end should be upgraded in the coming weeks.

  [and for this further elucidation from Igor:]

  I will upgrade the front end of the skimmer in the next release. This will happen probably just after New Year, given the shutdown and new, not-tested-enough architecture.

• Swig on Windows, MRvcmt (Toby) There is a problem on Windows building packages which invoke swig to build a python interface: a dependency seems to be missing so that an attempt is made to build the wrapper library first. It probably takes only a small change to MRvcmt to fix it, but the Ruby code is difficult to decipher. Who is maintaining this? (Richard) We were supposed to get additional DataMind help through December to make up for their near disappearance during the summer, but it hasn't been forthcoming. Long-term, Joanne has been nominated, but in the immediate future MOOT will take priority. [Late-breaking news from Toby: he has found a method to make package functionality available to python which meets all requirements on Windows — hasn't been tried elsewhere yet. It generates just one dll including both package code and python interface code, so MRvcmt has no problem with dependencies.]
• OnBoardFilter: (Tracy) Has made a build for Flight Software release B1-0-5; needs to be exercised. (Richard) There were bugs in the GRB filter so there will be a -6 release shortly. (Tracy) but we're not using the GRB filter. At this point the effort to get a release to build on Windows is not a terribly big job; he knows what kind of things to look for and how to fix them. However in this last round there were substantial changes in the way statistics-gathering is done. The new version has tentacles into more fsw packages than before.
• SCons (Navid) had been waiting for Jim's mods to python code to use commonUtilities; that is now ready. Now he's working on references to XML files. Several (ScienceTools) packages have been modified in CVS, but Navid will wait to tag until he gets approval from the owners.

  Batch submission for the new RM is now in test; it looks ok so far.

  Should we move to SLAC's centrally-supported installation of MySQL? There are advantages: for example, they would handle back-ups. He would like a decisition soon. (Richard) The SCS-maintained MySQL dbs will be blocked at the firewall, so probably not suitable for calibration database and MOOT, but ok for RM. (Navid) Installer needs to work through the firewall [command line installer only, I think; gui doesn't talk to db directly.] (Bryson) elogbook needs MySQL 4. (Navid) Central installation offers both 4 and 5, but his preference would be to convert elogbook to use 5 anyway. It is the recommended release and at some point MySQL Central will stop supplying security patches for version 4.

• Anders' concerns (Heather)
  • The EBF blob problem of last week is not yet resolved.
  • hadd problems with the CAL ntuple: Warren has a specific example of merging 104 CAL ntuples that results in a file slightly larger than expected and when opening the file causes his machine to lock up. There's a core file. Warren has also tried merging without hadd, and this also results in an unreadable file.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:45:39 -0700