The first look at the job run by Urs with the breco overlap fix in is not
good.
I got ~4% of IERR < 0 events (usually they are of the order of the 0.2%).
This is a clear sign that something is screwed up in the overlap flag.
Daniele
>
> Hoi,
>
> 1. PidKilling was still running with the old PidKilling setup. This is
> fixed by sourcing the correct tcl in IbuIsl.tcl:
>
> #old sourceFoundFile BetaMicro/BetaMicroSmear.tcl
> sourceFoundFile BetaMicro/BetaMicroPidKilling.tcl
>
> 2. PidKilling crashes in MuonMicroKillDispatch::BtaMicroMapCreator.cc.
> No solution to that yet. I think we use the output of the dispatch
> module when filling the ntuple.
>
> 3. Our low-level tcl is (imvho) a bit suspect: We get the PidBuildEnv
> from the Roo implementation and not the Bta version. This is most
> probably not what we want (though it might not hurt). I guess this
> happens maybe at the islNew.tcl level (or in the initial
> reading-setup tcl sourced by that tcl file). This was discovered
> by cleaning up things, in particular doing
>
> mod disable RooBuildPidEnv
>
> We are missing printouts like
>
> BtaBuildPidEnv::AbsEnv.cc(334):Overriding BtaEnv object in global environment.
> DchBuildEnv:mapconfig = dch_dtag
> DchBuildEnv:mapfile = DchMaps/dch_dtag.map
> DchBuildEnv: Configuration database disabled for channel maps
> DchBuildEnv: Configuration database disabled for MapBrowsers...
> SvtBuildEnv:BdbCondPathSingleton::loadFromConfigFile() -- info.
> A configuration file describing a revision/owner path was not found.
> Assuming the default behavior of BdbDatabase::fetch() interface --
> the most recent version of the condition data will always be
> fetched from the database.
> SvtBuildEnv: Pulse height ratio cut is 1.1
>
>
> I have heard that other people have successfully implemented the
> recipes for PidKilling. Given the problems in the BuildPidEnv, I tend
> to suspect that our low-level job setup is maybe at the source of the
> problems.
>
>
> A few jobs are running on data, they'll end up in
>
> ~ursl/output/sx-080202/.
>
> These are with the RooBuildPidEnv, but should be sufficient to test
> the breco overlap business.
>
>
> Cheers,
> --U.
>
|