John Idarraga | 26 Aug 15:01 2015
Picon
Picon

load libs with cling,

Hello all,

     With CINT I used to have a scripts of this sort

{
gROOT->Load("libMylib.so");
AClassInMylib a;
}

     With cling (ROOT6) I can not do it anymore because before loading 
the lib cling is already checking if the symbols are available, and 
warns me saying that I probably forgot to load the lib (if I understand 
correctly).  How do I proceed with ROOT6 in this case ?  (I am 
generating the dictionaries now with rootcling, that should be ok).

thanks,

J.

Alessandro Bertolin | 25 Aug 10:14 2015
Picon
Picon

roofit status() report

Dear root experts,

  I have a roofit fit that ends with:

  MINUIT WARNING IN HESSE
  ============== Second derivative zero for parameter3
   MNHESS FAILS AND WILL RETURN DIAGONAL MATRIX.
  FCN=320459 FROM HESSE     STATUS=FAILED          5 CALLS          55 
TOTAL
                      EDM=4.61359e-06    STRATEGY= 2  ERROR MATRIX UNCERTAINTY 100.0 per cent

the "->status()" for it that I get is 0 ... so is there a different way to 
identify automatically (without having to look to the text output) a case 
like this one ?

thanks a lot in advance for your time and patience,

        Alessandro

Alessandro Bertolin
Istituto Nazionale di Fisica Nucleare
Sezione di Padova
Via Marzolo 8, 35131 Padova - Italy
Tel: +39 049 9677094 (office)
E-mail: alessandro.bertolin <at> pd.infn.it
Home page: http://www.pd.infn.it/~bertolin/TOP.html

Rocky Bala Garg | 14 Aug 21:27 2015
Picon
Picon

MIGRAD terminates without convergence before reaching call limit

Dear Rooters,

I am using TMinuit to do a signal+background log likelihood fit using a four parameter background function.

The code i am using is quite big so i try to explain as much as possible.

My parameter of interest is the cross section of signal. So basically, i have 5 parameters to fit to the data : one is for signal and 4 are for background.

Initially, i put the cross section to zero and make it constant and do a background only fit to obtain initial values of the background parameters.
I then, use these obtained parameters values from bkg only fit as initial guesses to do the signal+background fit.
My problem is that when i do the background only fit then the fit converges and i get some initial values of the background parameters but when i do the S+B fit using these background parameters as initial guesses then the fit fails.
I have defined a call limit of 20000 but minuit terminates without convergence before reaching this call limit. In principle, MIGRAD should keep on progressing the minimization until the call limit reached but it is failing much before that.

I have attached the output log file for both bkgonly fit and signal+bkg fit.
I have also attached the whole code as a tar file with this mail. The code contains a Makefile as well as a README.
I am using the Root version 5.32 and that too on a remote machine of Fermilab by doing ssh.

One more strange thing that is happening is that the same code is working absolutely fine on my local computer but giving the problem on fermilab machine. I am using root version 5.34 on my local machine. I have also checked by changing the root version on fermilab machine but all in vain.

In the signal + bkg fit log file, It gives a message of the type
----------------------------------------------------------------------
 RELATIVE CHANGE IN COV. MATRIX=  0.7 per cent
 MIGRAD MINIMIZATION HAS CONVERGED.
 MIGRAD WILL VERIFY CONVERGENCE AND ERROR MATRIX.
   START COVARIANCE MATRIX CALCULATION.
 EIGENVALUES OF SECOND-DERIVATIVE MATRIX:
        -1.0770e-05  5.2672e-04  5.4036e-02  8.8171e-01  4.0637e+00
 MINUIT WARNING IN HESSE  
 ============== MATRIX FORCED POS-DEF BY ADDING 0.004075 TO DIAGONAL.
 MIGRAD TERMINATED WITHOUT CONVERGENCE.
 FCN=133.788 FROM MIGRAD    STATUS=FAILED        417 CALLS         418 TOTAL
--------------------------------------------------------------------------
I don't understand the behaviour of MIGRAD here. It is showing both 'MIGRAD MINIMIZATION HAS CONVERGED' and 'MIGRAD TERMINATED WITHOUT CONVERGENCE'.

Please try to help me out on this.

Please let me know if you need some other information as well.

Thanks
Rocky


Attachment (BkgOnlyFit.log): text/x-log, 65 KiB
Attachment (SigPlusBkgFit.log): text/x-log, 90 KiB
Attachment (LimitCode.tar.gz): application/gzip, 136 KiB
Jonas Rademacker | 10 Aug 21:24 2015
Picon
Picon

compiling on OS X with gcc/g++, not clang

Hi

I am trying compile root on mac OSX 10.10.4. I want to use openMP in my programme, hence clang does not work. By
default, on OSX, gcc and g++ are just links to clang, but I did have installed “proper" gcc and g++ and can
now do parallel/threaded openMP programming with my new g++ compiler, but I cannot compile root.

I edited the configure script like this:

altcc=gcc
altcxx=g++
altf77=gfortran
altld=g++
altccset=
gcctoolchain=
usercflags=
usercxxflags=
host=

This is what I get:

0 rademack <at> artixphybrisacuk:root-6.04.02 79> make
g++ -O2 -DNDEBUG -m64 -std=c++11 -pipe -Wshadow -W -Wall -Woverloaded-virtual -fsigned-char
-fno-common -Iinclude     -pthread -UR__HAVE_CONFIG -DROOT_STAGE1_BUILD
-I/Users/rademack/Desktop/root-6.04.02/core/utils/src
-I/Users/rademack/Desktop/root-6.04.02/core/metautils/src
-I/Users/rademack/Desktop/root-6.04.02/interpreter/cling/include
-I/Users/rademack/Desktop/root-6.04.02/interpreter/llvm/inst/include -DNDEBUG -D_GNU_SOURCE
-D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__STDC_LIMIT_MACROS -stdlib=libc++
-std=c++11 -fno-exceptions -fno-rtti -fno-common -fno-strict-aliasing -DNDEBUG -Wno-shadow
-Wno-unused-parameter -MMD -MP -o core/utils/src/rootcling_tmp.o -c core/utils/src/rootcling_tmp.cxx
g++: error: unrecognized command line option ‘-stdlib=libc++’

Any good idea how I can fix this?

Many thanks,

Jonas

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Jonas Rademacker, H H Wills Physics Lab, Tyndall Av, Bristol,
 BS8 1TL, UK.     Tel: +44-117-3317904   Fax: +44-117-9255624

John Haggerty | 7 Aug 06:03 2015

Access to std::map in TTree branch?

Is it possible to access a TTree branch containing a std::map?  A branch 
like this:

> xx.ecells : map<unsigned int,float> ecells[xx_]

seems to be saved to a root file correctly, and one can 
Draw("xx.ecells.first") which seems to show sensible keys, but I can't 
see sensible float values, although perhaps there is a problem in the 
way they were written.  I'm unclear whether this should work or will 
work in the future.  This is with root v5-34-26, I haven't experimented 
with other versions.  Thanks.

--

-- 
John Haggerty
email: haggerty <at> bnl.gov
cell: 631 741 3358

Marc de Kamps | 5 Aug 23:35 2015
Picon

Obtaining TColor from a scalar

Hi,


The ROOT documentation is quite clear on how to set a palette for colouring histograms based on their variable values. I find it harder to find out how I can convert a scalar value directly  into a TColor instance. I want to do this do because I want to convert a scalar value into the fill color of a TPolyLine.  I understand how I can set a palette globally, but they look useful for converting scalars  to color values in general.


I want to be able to select a grey value (or indeed a color value, as some of the  palettes look useful), based on e.g. a scalar value between 0 and 1. How can I do that?


Thanks,

Marc

Lynn Garren | 5 Aug 01:25 2015

building root 6.04.02 with c++14

We successfully built root 6.04.00 with c++14 enabled, but the build of 
root 6.04.02 is failing with the following error:

Generating PCH for core/base core/thread io/io math/mathcore net/net 
math/matrix hist/hist tree/tree graf2d/graf graf2d/gpad graf3d/g3d 
gui/gui math/minuit hist/histpainter tree/treeplayer tree/treeviewer 
math/physics graf2d/postscript core/rint math/smatrix math/smatrix 
hist/spectrum hist/spectrumpainter gui/fitpanel graf2d/x11 graf2d/x11ttf 
graf3d/gl graf2d/asimage graf2d/asimage bindings/pyroot math/genvector 
math/genvector math/minuit2 roofit/roofitcore roofit/roofit 
roofit/roostats roofit/histfactory hist/hbook tmva/tmva tmva/tmvagui 
math/genetic

error: C++14 was enabled in PCH file but is currently disabled

I've identified and tried using a couple of patches from the head, but 
haven't found anything yet that works.  We are using configure and gcc 
4.9.2:

configure linuxx8664gcc
   --with-cc=gcc --with-cxx=g++ --with-ld=g++
   --enable-http --enable-gdml
   --enable-cxx14 --enable-roofit
   --with-fftw3-incdir=... --with-fftw3-libdir=...
   --disable-mysql
   --with-xrootd-incdir=...
   --with-xrootd-libdir=...
   --enable-minuit2 --disable-xft --disable-sqlite
   --with-xml-incdir=... --with-xml-libdir=...

Lynn

Jonas Rademacker | 4 Aug 20:48 2015
Picon
Picon

cannot compile root 6.04.02 on OS X v 10.3.3 with clang on

Hi

I tried to compile root version 6.04.02 on OS X v 10.3.3 and failed at the linking stage.

I executed the following commands

$ ./configure macosx --prefix=/usr/local --enable-soversion --enable-minuit2 >& configOutput.log
$ make >& makeOutput.log &

The last few lines of output are:

 "vtable for clang::UsedAttr", referenced from:
     cling::DeclCollector::AddedCXXImplicitMember(clang::CXXRecordDecl const*, clang::Decl const*)
in DeclCollector.o
     cling::IncrementalParser::markWholeTransactionAsUsed(cling::Transaction*) const in IncrementalParser.o
     cling::DeclCollector::AddedCXXImplicitMember(clang::CXXRecordDecl const*, clang::Decl const*)
in DeclCollector.o
     cling::IncrementalParser::markWholeTransactionAsUsed(cling::Transaction*) const in IncrementalParser.o
 NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
ld: symbol(s) not found for architecture i386
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make: *** [core/utils/src/rootcling_tmp] Error 1

I also attached the full output to this email

Do you know what might be wrong? Help would be much appreciated.

Best,

Jonas

PS: I tried also with 6.02, getting the same problem.

Attachment (configOutput.log): application/octet-stream, 7353 bytes
Attachment (makeOutput.log): application/octet-stream, 366 KiB

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Jonas Rademacker, H H Wills Physics Lab, Tyndall Av, Bristol,
BS8 1TL, UK.     Tel: +44-117-3317904   Fax: +44-117-9255624

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 Jonas Rademacker, H H Wills Physics Lab, Tyndall Av, Bristol,
 BS8 1TL, UK.     Tel: +44-117-3317904   Fax: +44-117-9255624

Mathieu ROGER | 2 Aug 11:19 2015
Picon

UNSUBSRIBE


Diego Casadei | 31 Jul 10:50 2015
Picon
Picon

TFormula issue in 6.03/04?

Hello.

I want to plot a compound power-law function:

TF1 fMbkg("fMbkg","[0]*(x/[2])^[1]+[3]*(x/[5])^[4]",1.0,200.0);

but I get a wrong plot with 6.03/04 (it looks reasonable with 5.34/19) 
on my mac.  The commands

fMbkg.SetParameters(0.1, -8, 5, 0.0, -5, 30); // only 1st active
fMbkg.Draw();

produce a monotonically decreasing function with 5.34/19 which is always 
positive, as it should be.  On the other hand, I get all negative values 
in 6.03/04.

Testing the second part only with

fMbkg.SetParameters(0.0, -8, 5, 0.1, -5, 30); // only 2nd active

gives me a positive and monotonically increasing function in 6.03/04 but 
looks OK with 5.34/19.

Is it a bug in TFormula?

Best regards,
Diego

--

-- 

Diego Casadei
-------------------------------------------------------
           Visiting Scientist, University College London
                     Department of Physics and Astronomy
-    -    -   -   -  -  - - - -  -  -   -   -    -    -
Senior Research Fellow, FHNW
School of Engineering, Institute of 4D Technologies
-------------------------------------------------------
diego.casadei <at> cern.ch              skype: diego.casadei
mobile: +41-76-213-5376          http://cern.ch/casadei

Omar Andrés Zapata Mesa | 30 Jul 21:32 2015
Picon

Possible Bug in master code

Hi Dear ROOTers,

I am defining in a namespace  a global static variable 

#include<Rcpp/Named.h> //header with the class  Rcpp::internal::NamedPlaceHolder 
namespace ROOT {
   namespace R {
      static Rcpp::internal::NamedPlaceHolder Label;
   }
}


IncrementalExecutor::executeFunction: symbol '_ZN4ROOT1RL5LabelE' unresolved while linking [cling interface function]!
You are probably missing the definition of ROOT::R::Label

I try using pragma in my LinkDef.h
#pragma link C++ global ROOT::R::Label;

but still dont work,

It stopped working after update the master.

Thanks

Omar.


Gmane