Our analysis groups have a specific framework that converts miniAOD to
TTree format. Recently, I’ve launched it in CMSSW_8_0_26_patch1 which is
linked to ROOT 6.06/01. If I open my tree there, I see nearly all
variables in “EventTree” filled with events.
However, if I open this tree with newer version of ROOT, like 6.09 and
6.10 a lot of those variables are shown “Empty”. I tried this on both
MAC and Linux. If I reinstall older versions (tested with 6.06 and 6.08)
it works fine.
Any idea why? You can find a small (~21MB) version of this tree here:
~hdelanno/public/HyperNewsROOT/tree.root
Thanks for your test and reply. I realize I forget to precise one important point: I see this issue with the TBrowser.
Indeed if I open the tree like you did, I see everything filled like it should be.
However, if I navigate through the tree thanks to a TBrowser, then I have a lot of “Empty” histograms for ROOT verions 6.09 and 6.10 (didn’t test on 6.11).
So I tried, and the problem is not related to the TBrowser, but trying to draw some of the branches already fails, For example, this:
root [0] TFile f("tree.root", "read");
root [1] TDirectoryFile* d = (TDirectoryFile*) f.Get("tupel");
root [2] TTree* t = (TTree*) d->Get("EventTree");
root [3] t->Draw("METsigy2")
Info in <TCanvas::MakeDefCanvas>: created default TCanvas with name c1
root [4]
Gives an “Empty” histogram. Maybe @pcanal or @Axel could help with this issue
The underlying problem is solved in the master and v6.10 patch branch (i.e. the upcoming 6.12/00 and 6.10/10.
The trigger for this problem is the fact that the TTree (very very oddly) was stored after being put in MakeClass mode; this is odd as TTree is MakeClass mode usually do not work very well for output/writing. [It is even odder as many of the branches, for example those containing a vector, do not work in MakeClass mode …]
Nonetheless the MakeClass-ness of the branch is supposed to be transient (but was not until the fix I made).