Deploying the new Acd Recon will take some time for two reasons.

  1. Lots of software has been added, final testing and verifying backward compatiblity or the various packages will take some times.
  2. Some of the changes are fairly major.  We should understand them before we proceed.

First off, here are the packages that are affected.

  • Data structures and conversion:  Event reconRootData RootConvert
  • ACD software:  AcdUtil AcdDigi AcdRecon
  • High Level:  AnalysisNtuple HepRepSvc Gleam (jobOptions only) svac

I propose we do this in three steps:

  1. Put in the new data structure and verify backward compatility (should be easy, already largely done).
  2. Put in new versions of ACD software, High Level software, but set job options to use current versions, this allows further testing
  3. Activate it all
  • No labels

1 Comment

  1. I'm probably misunderstanding the ultimate goal... but I would not think you would want to have 'both" ACD recons in the code going into the future. So, I would suggest branching these packages, removing the old code and going with the new. You can test by running parallel data sets, one with the non-branched code, one with the branched, to cross check "final" results. Once this has been confirmed then it is straightforward to bring the branch back into the mainline.

    The concern is long term maintainability, especially if/when that falls to someone not directly involved in developing the code. Having legacy code laying about can be confusing, especially if its not used.