RooSimultaneous::generate Segm.Violation in ROOTv>= 5.32

Roofit and RooStats
1

Dear experts,

I am generating some data sample with RooSimultaneous but for ROOT versions 5.32 or later I get a segmentation violation (while for ROOT 5.30 it seems to work properly).
Before the crash I get errors like:

[#0] ERROR:Generation – RooAddGenContext::bin4:generate: cannot calculate expected number of events

which are due to the fact that the number of events I am generating is not sufficient to have at least one event for each bin (category).
I report down the stack trace.

Thanks in advance,

Francesco

===========================================================
There was a crash.
This is the entire stack trace of all threads:

#0 0x00002ad539ca8df5 in waitpid () from /lib64/libc.so.6
#1 0x00002ad539c4c491 in do_system () from /lib64/libc.so.6
#2 0x00002ad532ffd746 in TUnixSystem::StackTrace() ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libCore.so
#3 0x00002ad532ffd01c in TUnixSystem::DispatchSignals(ESignals) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libCore.so
#4
#5 0x00002ad53839a9c0 in RooDataSet::RooDataSet(char const*, char const*, RooArgSet const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#6 0x00002ad5384926ad in RooSimSplitGenContext::generate(int, bool, bool) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#7 0x00002ad53830a2b7 in RooAbsPdf::generate(RooArgSet const&, int, bool, bool, char const*, bool) const ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#8 0x00002ad53830e640 in RooAbsPdf::generate(RooArgSet const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#9 0x00002ad53830e7b3 in RooAbsPdf::generate(RooArgSet const&, int, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&)
()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#10 0x000000000040f076 in confinte::addRandom (this=0x7fff2b15a570,
n_events=10, pdf_name=) at confinte.cpp:86
#11 0x0000000000408a1e in main (argc=12, argv=0x7fff2b15ca48) at calca.C:210

The lines below might hint at the cause of the crash.
If they do not help you then please submit a bug report at
root.cern.ch/bugs. Please post the ENTIRE stack trace
from above as an attachment in addition to anything else
that might help us fixing this issue.

#5 0x00002ad53839a9c0 in RooDataSet::RooDataSet(char const*, char const*, RooArgSet const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#6 0x00002ad5384926ad in RooSimSplitGenContext::generate(int, bool, bool) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#7 0x00002ad53830a2b7 in RooAbsPdf::generate(RooArgSet const&, int, bool, bool, char const*, bool) const ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#8 0x00002ad53830e640 in RooAbsPdf::generate(RooArgSet const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&) ()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#9 0x00002ad53830e7b3 in RooAbsPdf::generate(RooArgSet const&, int, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&, RooCmdArg const&)
()
from /afs/cern.ch/sw/lcg/app/releases/ROOT/5.34.00-rc1/x86_64-slc5-gcc43-opt/root/lib/libRooFitCore.so
#10 0x000000000040f076 in confinte::addRandom (this=0x7fff2b15a570,
n_events=10, pdf_name=) at confinte.cpp:86
#11 0x0000000000408a1e in main (argc=12, argv=0x7fff2b15ca48) at calca.C:210

2

Hi,

Can you please post a example reproducing this problem or send us a workspace file containing your pdf ?
Thank you

Lorenzo

3

Hi Lorenzo,

I attach the workspace containing my pdfs.
The one giving this crash is “simBsPdf”.

Thank you,
Francesco
RootWorksp_FD_outp.root (199 KB)

4

Hi,

Thank you for the workspace. Which command exactly did you use for generating ? Did you include the categories in the observables ?

For me doing

simPdf->generate(allobs, 0, RooFit::Extended() )

where allobs is

root [37] allobs.Print("V")
  1) RooRealVar:: m_{#mu#mu} = 5300
  2) RooCategory::    binning = bin1(idx = 0)

works fine

Lorenzo

5

Hi,
I use the following code:

RooDataSet * datasig = 
    pdf_sig->generate(*m_ws->set(m_obs_set), n_events, Name("mysignal"+pdf_name));

where of course “m_obs_set” is the name of the set of observables which includes the two that you mention (so including the category).
With or without the Extended option it does not change, I get always the same error.

cheers,
Francesco

6

Strange, because this code works fine for me:

{
   TFile f("RootWorksp_FD_outp.root");
   RooAbsPdf * simPdf = w->pdf("simPdf");
   RooArgSet * obs = simPdf->getObservables(w->data("combData"));
   RooAbsData * data = simPdf->generate( *obs, 0, RooFit::Extended, RooFit::Name("mydata") );
   RooPlot * pl = ((RooRealVar*) obs->first())->frame();
   data->plotOn(pl, RooFit::Cut("binning == binning::bin1"));
   pl->Draw();
}

Please past your full code showing your problem
Cheers
Lorenzo

7

Hi,

your test works because you are using the wrong pdf.
The crash is with the pdf named “simBsPdf”, while you are using “simPdf”.
If I use your code I can reproduce the problem with the first but not with the second pdf, as expected.

cheers,
Francesco

8

Hi,

Sorry, I have not read carefully your post.
yes, I can reproduce now your problem with 5.34, but I have already the fix, which I am going to commit soon in the trunk and the patch releases. The problem is due to a rounding in the number of events and it has been reported few days ago in the RooStats mailing list, see
groups.cern.ch/group/roostats-d … 3A416E9071

Thank you for your report

Lorenzo

9

Hi Lorenzo,

thanks for this.
I have subscribed to the mailing list to be able to see the thread, and I will update to the trunk.

cheers,
Francesco

10

Hi Francesco,

The fix is now committed in the trunk (revision 44630 and also in 5.32 and 5.34 patches)

Lorenzo

11

Thanks a lot for the prompt reply and reaction! :slight_smile:

cheers,
Francesco