Hi,
the underlying problem is that you are mixing usage of TTreeReader and TTree/TChain.
TTreeReader assumes that everything you are doing with the data you do through its TTreeReaderValue/Arrays, but you are doing a clone on the same TChain that TTreeReader is reading, in the same event loop.
What is happening is that TTreeReader is trying to help your code run faster by only loading r1 and r2 in memory if absolutely necessary. The only way it can judge whether it is necessary, is to check whether r1 or r2 are used anywhere, that’s why when you do *r1 it “fixes” your problem.
I’m not 100% sure but you might be able to substitute *r1, *r2 etc. with a call to chain->GetEntry(reader.GetCurrentEntry()), which should load all active branches in the TChain.
If you have access to ROOT v6.14, though, this kind of application is exactly why we introduced RDataFrame:
ROOT::RDataFrame df("result", "./data/*.root");
df.Filter("x > 0") // some cut expression -- you can also pass a C++ function
.Snapshot("result", "RESULT.root"); // write to file all events that pass the filter
Those three lines do a cut and clone, you don’t need anything else.
Hope this clarifies things a bit!
Cheers,
Enrico