For reasons still to be determined [since determined by Heather - not straightforward!], there were problems in various parts of the code in the recon step which expected MC tuple values to be filled sensibly. (We've run with real data input before; the procedural difference would appear to be ingesting the FT2 file as well.) Some of this can be handled with a new tag of AnalysisNtuple which Toby and Leon have made, and some careful pruning of job options to eliminate use of various algorithms and tools which are not appropriate in this context, but there are still difficulties with ClassifyAlg.
(Anders) Goals (to be achieved by Wednesday) are
It's ok if there are some nonsense values in the tuple if they are unused.
[from Heather post-meeting: Leon has offered to create a new tool in AnalysisNtuple, aptly named McKludgeTool, that will fill the three MC variables with -1 for ClassifyAlg, to allow the algorithm to run, yet making it clear to mortals that the MC values are bogus for this "real" data run.]
cmt broadcast cmt config
, then build the new
packages. [Anders tried this and it worked!]
[Related matters I didn't think to mention when I had the chance: 1) More recent tags to mootCore fixed Windows compilation problems and made some small improvements for Online users, independent of any of the code Martin will use. 2) Much of the code supporting access to MOOT information via the calibration TDS has been written even though we won't need it for OktoberTest. I expect to commit and tag it later this week, disabled by default.]
previous | minutes index | next |