Hi Caius, Your second file in the chain contains a Tree with 0 entries. This gives a problem when computing the entry offset table in case wildcarding has been used in the chain creation. This is effectively a nasty case for which we will have to find a solution. Meanwhile, you can circumvent the problem by doing: TChain* ch_ce = new TChain("t_cse"); ch_ce->Add("./f*.root",0); ch_ce->GetEntry(999999); //this will force reading of all Tree headers ch_ce->Draw("run:snarl", "projs==1"); Thanks for having reported this problem. Rene Brun On Tue, 5 Nov 2002, Caius Howcroft wrote: > > Hi, > > root 3.3.09 > > I have a set of files all with the same trees in (a t_cse and a t_filter). > > Here is what happens: > > root [0] TChain* ch_ce = new TChain("t_cse") > root [1] ch_ce->Add("./f*.root") > (Int_t)11 > root [2] ch_ce->Draw("run:snarl", "projs==1") > <TCanvas::MakeDefCanvas>: created default TCanvas with name c1 > (Int_t)20 > root [3] ch_ce->Draw("run:snarl", "projs==1") > (Int_t)21 > root [4] > > > you see that I run the same Draw command twice, but the first time it > selects 20 events and the second time it selects 21 events!, This really > worries me if I execute the same bit of code twice I get different > results. This effect seems to scale with files, and in the total data set > the difference is more than a factor of 2!!!. This isnt any fancy tree is > just a bog standard tree with bog standard Int_t and Float_t entries. > > Exactly the same thing happens for TChain::Scan. > > > (the files I used to do this are at > http://www.hep.phy.cam.ac.uk/~howcroft/data_trees.tar.gz) > > I'm using root 3.03/09 on redhat 7.3, gcc v3.0, i685. > > Am I doing somthing really stupid? > > Caius > > > ____________________________ > CAIUS HOWCROFT > +44 (0)1223 764 128 UK(office) > AOL IM: sumsmug > Yahoo IM: sumsmug >
This archive was generated by hypermail 2b29 : Sat Jan 04 2003 - 23:51:16 MET