Compile my C/C++ program with certain version of Python interpreter

If you have more than two Pythons in your system, such as in the case that Python2 and Python3 are installed in your system, you may have to specify one explicitly to compile that with your application program, which is written not in Python. Here is how to specify PythonX.Y with the python version number X.Y in a Makefile to build your application.

First, you need to find out compiler and linker options specified at the time of installation of PythonX.Y:

% pythonX.Y-config --cflags

will show you the compiler options <CFLAGS> and

% pythonX.Y-config --ldflags

will show you the linker options <LDFLAGS>.

Second, you add lines in the Makefile to include <CFLAGS> in the compilation and load <LDFLAGS> in the linkage.

That's it!

How to CMake MyProject together with ROOT GUI

I have MyProject composed of source files

myProject/c++/src/myMain.cpp
myProject/c++/include/myMain.h
myProject/c++/include/myMainLinkDef.h

to run on ROOT GUI. In myMain.cpp, a class MyMainGUI is called in a loop of  TApplication. In myMain.h,  this class is declared with the first line

    RQ_OBJECT("myMainGui").

myMainLinkDef.h includes pragma comment lines which look like

#pragma link C++ class MyMainGUI;

#pragma link C++ class mylib::MyFunction;

where mylib::MyFunction is a class used in ROOT GUI and given in

/somewhere/myLibrary/myLib.h
/somewhere/myLibrary/libmyLib.a

Problem: Compose a cmake package to build MyProject in an arbitrary directory myWKDIR.

I describe a solution in the following.

1. Prepare myProject/c++/CMakeLists.txt (presupposition: ROOT version 5.34):
==========================================================
cmake_minimum_required(VERSION 2.8)

set(CMAKE_VERBOSE_MAKEFILE 1)

#is convenient for debugging the make process.

project(MyProject)

set(CMAKE_CXX_COMPILER g++)
# since I used g++ to compile libmyLib.a

set(EXECUTABLE_OUTPUT_PATH ${PROJECT_BINARY_DIR}/..)
#since I'm going to build MyProject in myWKDIR/build/
#so that ${PROJECT_BINARY_DIR}/.. means myWKDIR.

#Prepare for the ROOT GUI

#-------------------------------------

#imitating geant4/examples/analysis/AnaEx02

#/////////////////////////////////////////////////////////////////////

find_package(ROOT REQUIRED)

find_program(ROOT_CINT_EXECUTABLE rootcint PATHS ${ROOT_DIR}/bin NO_DEFAULT_PATH})

#External Libraries to be linked

#---------------------------------------

set(myLibrary_DIR /somewhere/myLibrary)

add_library(MyLib STATIC IMPORTED)

set_property(TARGET MyLib PROPERTY

IMPORTED_LOCATION ${myLibrary_DIR}/libmyLib.a)

#Other ways to include libmyLib.a in the ld process is to specify this library in

#CMAKE_EXE_LINKER_FLAGS as

#-L${myLibrary_DIR} -lmyLib

#or find_library(MyLib myLib PATHS ${myLibrary_DIR})

#and include MyLib in target_link_libraries.

#building a dictionary for ROOT GUI

#------------------------------------------------

set(INCLUDE_DIRECTORIES

${PROJECT_SOURCE_DIR}

${PROJECT_SOURCE_DIR}/include

${myLibrary_DIR}

${ROOT_INCLUDE_DIR}
)

include_directories(${INCLUDE_DIRECTORIES})

#generate the dictionary

#the following part is imported from http://lcio.desy.de/v01-60/cmake/MacroRootDict.cmake

set(MAINPRGM "myMain")

set(headers include/myMain.h)

set( _dict_includes )

FOREACH( _inc ${INCLUDE_DIRECTORIES} )

SET( _dict_includes "${_dict_includes}\t-I${_inc}")

ENDFOREACH()

ADD_CUSTOM_COMMAND(

OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${MAINPRGM}Dict.cpp ${CMAKE_CURRENT_BINARY_DIR}/${MAINPRGM}Dict.h

COMMAND ${ROOT_CINT_EXECUTABLE} -f ${CMAKE_CURRENT_BINARY_DIR}/${MAINPRGM}Dict.cpp -c ${_dict_includes} ${headers} include/${MAINPRGM}LinkDef.h

WORKING_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}"

DEPENDS ${headers} include/${MAINPRGM}LinkDef.h

COMMENT "generating: ${MAINPRGM}Dict.cpp ${MAINPRGM}Dict.h"

)

list(APPEND headers ${CMAKE_CURRENT_BINARY_DIR}/${MAINPRGM}Dict.h)

#build the target

#-------------------------

#prepare for the ld flag

execute_process(COMMAND root-config --glibs OUTPUT_VARIABLE     

               ROOT_LD_FLAGS OUTPUT_STRIP_TRAILING_WHITESPACE)                 

set(CMAKE_EXE_LINKER_FLAGS "${ROOT_LD_FLAGS}")

#destination

set(sources src/myMain.cpp

${CMAKE_CURRENT_BINARY_DIR}/${MAINPRGM}Dict.cpp

)

add_executable(${PROJECT_NAME}.exe ${sources} ${headers})

target_link_libraries(${PROJECT_NAME}.exe MyLib)

2. build MyProject in myWKDIR

============================
> mkdir myWKDIR
> mkdir myWKDIR/build
cd myWKDIR/build
myWKDIR/build> cmake /directory_path_to/myProject/c++
myWKDIR/build> make

Any comments are welcome.

gcc and gfortran on OS X 10.9 Marvericks

A new release of OS X 10.9 Marvericks was installed on my MacBook 15-inch Retina Pro 2012.
I kept all softwares previously installed on OS X 10.8 as they were. All of them are working
well without a change except for the following.

gcc 4.8.1
========
I needed to install the Command Line Tools by issuing 
%xcode-select --install
to make gcc work. This is necessary regardless whether one has
already installed Xquartz or not.

gfortran
=========
I need
-fno-underscoring
option at the load time to use gnu extensions. 
This was not required before in my case.

A primitive way to pass a pointer to CINT macro

I have a TTree pointed to by *tree in my C++ program and want to pass it to MyMacro.C to use it in a CINT session invoked from the C++ program. Here is a primitive-but-it-does-work way as below.

Lines in my C++ program read

TTree *tree = ....
std::stringstream ss;
ss<< tree;
std::string macroName = "MyMacro.C("+'"'+ss.str()+'"'+")";
gROOT->Macro(macroName.c_str());

and in MyMacro.C, I have

void MyMacro(std::string &tp){

TTree *tree;
std::stringstream ss(tp);
ss>> tree;

int nentries = (Int_t)tree->GetEntries();
TObject *tobj;
...
}

Using MyClass in a CINT macro

When I am evaluating something in my C++ program and calling a CINT macro by

gROOT->Macro(mymacro.C);

to draw figures, I need to pass parameters of the evaluation to this macro.
When the number of parameters increases and their structures get complicated,  passing them by

mymacro.C(hard coded parameter list)

is not useful. In this case, I prepare a class MyClass to define a set of parameters to use
in my evaluation and pass it to the macro. Results of the evaluation kept in MyClass
can certainly be passed to the macro for drawings. The way to do this is well described in
The CINT Dictionary Generator.

Here I note my sample case where MyClass is stand alone and not inheriting from TObject class.
MyClass is defined in MyClass.h and implemented in MyClass.cpp
MyClass has a constructor without arguments.
Parameters are defined as static members of MyClass.
MyClass is called in mymacro.C as

MyClass passParams;

In the Makefile which I showed on Aug. 13, the following lines are added to create
a dictionary MyClassDict.o to compile with the main part:
CINTCLASS = MyClass
UOBJS := $(UOBJS) $(CINTCLASS).o $(CINTCLASS)Dict.o
$(CINTCLASS)Dict.cpp: $(CINTCLASS).h
<<tab>> rootcint -f $(CINTCLASS)Dict.cpp -c $(CINTCLASS).h

Now I can draw figures of results from a C++ computation interactively in CINT.

Geant 4.9.6 p02 installation

I use macports to install supplemental softwares:

sudo port -v install cmake (necessary)
sudo port -v install iAIDA (optional, I was just curious)

sudo port -v install xercesc3 (needed for GDML)

sudo port -v install qt4-mac +demos +examples (optional. This takes very long.)

I downloaded the newest available source from cern Source Download Page.
I am going to install it in /usr/local/geant4:

$ sudo mkdir /usr/local/geant4
$ cd /usr/local/geant4
$ mkdir build4.9.6.p02
$ sudo mv  ~/Downloads/geant* ./
$ sudo tar -xvf geant4.9.6.p02.tar
$ cd build4.9.6.p02

$ls -lO /
$sudo chflags nohidden /usr (to see files on finder)

$export G4INSTALL=/usr/local/geant4/geant4.9.6.p02

Now we use cmake

$sudo cmake -DCMAKE_INSTALL_PREFIX=/usr/local -DXERCESC_ROOT_DIR=/opt/local -DGEANT4_USE_OPENGL_X11=ON -DGEANT4_USE_RAYTRACER_X11=ON -DGEANT4_INSTALL_DATA=ON -DGEANT4_USE_GDML=ON -DGEANT4_USE_QT=ON $G4INSTALL

For build options, refer to section 2.3 in ‎geant4.web.cern.ch/geant4/UserDocumentation/UsersGuides/InstallationGuide/fo/BookInstalGuide.pdf

Now proceed usual make:
$sudo make -j4
$sudo make install

$ls -lO /
$sudo chflags hidden /usr (to hide /usr from finder)

Build an example application:
$cd /my/private/directory
$mkdir B1-build
$cd B1-build
$source /usr/local/bin/geant4.sh
$cmake -DGeant4_DIR=/usr/local/lib/Geant4-9.6.2 $G4INSTALL/examples/basic/B1
$make
$./exampleB1

Build another example which requires a slight different targeting:
$cd ..
$mkdir rdecay01-build
$cd rdecay01-build
$cmake -DGeant4_DIR=/usr/local/lib/Geant4-9.6.2 $G4INSTALL/examples/extended/radioactivedecay/rdecay01

That's it.

#This installation was performed after an unsuccessful installation of a macports distribution of geant4.9.4 by

$sudo port -v install geant4 +aida +gdml +raytracerx

Precisely speaking, the installation process was successful. However, when I try to build an example, clang called in Darwin-g++.gmk couldn't find files such as libgcov.a in my OS X 10.8 system to generate a library. I therefore uninstalled the macports distributed geant4.9.4.p02 from my system.

Makefile for my c++ lib with ROOT

For later reference, I put my Makefile in the following.


#===========================================================
#
##        The Standard Makefile
#
#(Usage)
# make U=<program_name>;  for compiling a single program
#              <program_name> = bar for bar.cpp
# make ; for compiling all (listed below all:)
#
#(Default Library) ROOT
#
#(History)
# 130813 updated
#===========================================================
main:
make library
make compile PROGRAM=$(U)

all:
make compile_all PROGRAM=testReadAData
make compile_all PROGRAM=testA2RT
make compile_all PROGRAM=testReadTTree
make compile_all PROGRAM=testSetDoPrm

WRKDIR = tmp
ULIBOBJS    = ReadAData-v03.2.o \
              A2RT-v03.5.o \
              ReadTTree-v03.3.o \
              SetDoPrm-v01.8.o

LIBRARY_NAME     = mycpplib
LIBRARY          = lib${LIBRARY_NAME}.a

ULIBDIR = /My/cpplib/is/here

#====================================================
### INCLUDES ###
include /opt/local/etc/root/Makefile.arch
INCLUDE = -I/opt/local/include
#------ include user headers
ULIBINC = -I$(ULIBDIR)/include -Iinclude -I.
INCLUDE := $(INCLUDE) $(ULIBINC)

### LIBS ###
LIBS := -L./ -l${LIBRARY_NAME} $(ROOTGLIBS) -L/opt/local/lib -lgsl

UOBJS    = <list your objects to be linked>
UOBJS    =

LANGSFX=cpp
COMPILER = g++
LINKER = g++

COMPOPT = $(ROOTCFLAGS)
LINKOPT =
GOPTION = -DDEBUG
#====================================================
SRC         = $(PROGRAM).$(LANGSFX)
OBJ         = $(PROGRAM).o

TARGET    = $(PROGRAM)_cxx

LDFLAGS    = $(LIBS)

COMPOPT := $(COMPOPT) $(GOPTION)
LINKOPT := $(LINKOPT) $(GOPTION)

LIBSRCS = $(addprefix $(WRKDIR)/,$(ULIBOBJS))

ULIBSRC = $(ULIBDIR)/src
UOBJDIR = $(WRKDIR)
UOBJS := $(addprefix $(UOBJDIR)/,$(UOBJS))
#----------------------------------------------------
library:
@if [ ! -d $(WRKDIR) ]; then mkdir $(WRKDIR); \
echo "## --> tmp/ created...";fi
rm -f ${LIBRARY}
rm -f ${LIBSRCS}
make ${LIBRARY}

${LIBRARY}: ${LIBSRCS}
ar cru $@ ${LIBSRCS}
ranlib $@

compile:
@if [ -f $(SRC) ]; \
then make $(TARGET); \
rm $(OBJ); \
else make all; fi

compile_all:
rm -f $(TARGET)
make $(TARGET)
rm $(OBJ)
#----------------------------------------------------

.SUFFIXES: .exe .f .c .o .a .cpp .cxx .cc

#=================================================  
$(TARGET):  $(OBJ) $(UOBJS)
$(LINKER) $(LINKOPT) $(OBJ) $(UOBJS) -o $@ $(LDFLAGS)

$(ROOTLINKSRC)Dist.cpp: $(ROOTGUIHDRS).h $(ROOTLINKSRC)LinkDef.h
rootcint -f $(WRKDIR)/$(ROOTLINKSRC)Dist.cpp -c $(INCLUDE) $(ROOTGUIHDRS).h $(ROOTLINKSRC)LinkDef.h

#The following line must be the first one
$(UOBJDIR)/%.o :  $(UOBJDIR)/%.cpp
$(COMPILER) -c $(COMPOPT) $< $(INCLUDE) -o $@

$(UOBJDIR)/%.o :  $(ULIBSRC)/%.cpp
$(COMPILER) -c $(COMPOPT) $< $(INCLUDE) -o $@

$(UOBJDIR)/%.o :  src/%.cpp
$(COMPILER) -c $(COMPOPT) $< $(INCLUDE) -o $@

$(UOBJDIR)/%.o :  %.cpp
$(COMPILER) -c $(COMPOPT) $< $(INCLUDE) -o $@

.f.o:
$(FCOMPILE) -c $(FCMPOPT) $(FINCLUDE) $<

.c.o:
$(CCOMPILE) -c $(CCMPOPT) $(CINCLUDE) $<

My c++ library with ROOT: test run

 A Makefile to compile my c++ library with ROOT is adjusted for the new environment. A flag -m32 is removed, an include path for Makefile.arch is changed for the new ROOT set (which I needed to look for in /opt/local/etc,) a link path for gsl is explicitly added. (Specifying  LD_LIBRARY_PATH in ~/.bashrc did not work. My make processes seem not referring to this env. variable.)
 All sample programs included in my library were complied with ROOT and run without any problem. I was expecting some mysterious problems occur to bother me. But no mysterious thing happen. It's I call a miracle!

Preparing for numerical works

After moved from the old 32-bit mode mac, the first thing to do is to recompile sources of my private c++ library with gcc without -m32 flag. This step is going well and will be finished in 3 hours.

Plan for a new MacBookPro: Writing specification

When we have a plan to purchase something, our administrators require us to write a specification to open for a bidding. I can not, however, just write "I want a MacBook Pro quad-core Intel Core i7, 2.8 GHz model." They don't accept names of commercial products and I can not use words like "Mac" nor "Intel". They accept only descriptions of characteristics which I need for my purpose. So, only "quad-core" and "2.8 GHz" are accepted in the sentence above. This is something like

namespace admin{

class Specification{

     public:
        Specification & cpu_arch(std::string const& arg);
        Specification & cpu_speed(double const& arg);
        Specification & memory_size(double const& arg);

        Specification & memory_speed(double const& arg);

        Specification & storage_size(double const& arg);

       ........

};

} // admin namespace

to use it as

using namespace admin;

Specification I_want_a_MacBookPro(

          Specification.cpu_arch(X86-64)

                            .cpu_speed(2.8)

                            .memory_size(16)

                            .memory_speed(1.6)

                            .storage_size(512));

ROOT5.22 doesn't have TFitResult

I was to include root/include/TFitResultPtr.h in my c/c++ fitting routine. But I was told "No such file..." I found this class was not included in the root version 5.22, which I got from fink! (Yes, I'm getting old.) I have 3 choces:
(1) to upgrade 5.22 in my current MacBookPro,
(2) to install a brand new 5.34, or
(3) to find another way in 5.22 without TFitResultPtr.
My estimate for required elapse time for each case is as follows.
(1) a week, (2) 3 days, (3) 1 day
I take (3).
....Result
I was right. It took a few 10 minuits to find the information I needed was
available through TF1 object used in the fit for TGraph.

lesson:
If you don't find an object in your old root, find another way to obtain the information you need. It can not be happen that there is no way in 5.22.

Specifying a file path at runtime

1. In c/c++ source code, specify the file with an environment variable:

std::string fileName = getenv("package_SRCDIR");
fileName = fileName+"<file.name>";
doSomething(fileName.c_str());

2. Ask bash users to prepare a run.sh which reads

export package_SRCDIR=/users/share/package_dir
$package_SRCDIR/source_cxx

where source_cxx is an executable of the source code.

Environment: Mac OS X 10.6, bash