Feb 11, 2020 learning C++ Makefile ROOT
main()
{
int a = 2;
int b = 3;
int c = a*b;
}
Let’s start with this short and not-quit-correct program. Save it to a file called test.cc. test.C or test.cpp works just as well.
Your computer does not understand this at all. You need to convert it to a format that the machine can understand using a program called compiler. A standard C++ compiler in Linux is called g++. So you can run
$ g++ test.cc # compile test.cc using g++
$ ls # check what is created by g++
a.out test.cc
to create a.out from test.cc using g++. a.out is an executable, you can run it this way:
$ ./a.out # run a.out in the current directory (./)
Of course, nothing will show up in your terminal. To print out the result of the calculation, you need to modify your program a bit:
#include <iostream>
main()
{
int a = 2;
int b = 3;
int c = a*b;
std::cout<<c<<std::endl; // print the value of c on screen
}
<< indicates the flow of data. std::cout is something declared in iostream.h. It means the standard output, or your terminal screen. std::endl is declared in iostream.h as well. It means the end of a line, or an Enter to start a new line. Everything behind // are comments and will be ignored by the compiler. #include <iostream> tells the compiler where to search for declarations of std::cout and std::endl.
Compile the modified test.cc again and run it:
$ g++ test.cc
$ ./a.out
6
Now the calculation result is printed in your terminal screen.
a.out is not a very good name. You can change the name of the comfile file using the o (outut) option of g++:
$ g++ test.cc -o test.exe
$ ls
test.exe test.cc
As I mentioned at the beginning, test.cc is not quit correct. To see what’s wrong, you can require g++ to print out useful warning messages to help us debug:
$ g++ -Wall test.cc # turn on "all" Warnings
test.cc:3:7: warning: ISO C++ forbids declaration of 'main' with no type [-Wreturn-type]
main ()
^
The warning message is quite clear. The C++ standard forbids to declear the main function without any return type. To get rid of this warning, you need to add int in front of main:
int main()
Try to compile again. The warning message should disappear.
From the simple program above we learned that you can call functions written by others in your program, for example, std::cout in iostream. Now we are going to call a random number generating function provided by ROOT:
#include <iostream>
#include <TRandom.h>
int main()
{
TRandom generator; // create an object of the class TRandom
std::cout<<generator.Rndm()<<std::endl; // call TRandom's public member function Rndm()
}
Run g++ test.cc and you will get
test.cc:2:22: fatal error: TRandom.h: No such file or directory
#include <TRandom.h>
^
compilation terminated.
The second include causes the problem while the first does not. This is because the second is not as standard as the first one. You need to tell g++ the location of TRandom.h:
$ g++ -I /path/to/include/ test.cc
This will fixed the previous error, but create more error messages, which is so long that you’d better save them to a log file so that you can check the start of them easily:
$ g++ -I /path/to/include/ test.cc > log 2>&1
Search on Google bash error redirect if you don’t understand the meaning of 2>&1. Open log and you should be able to allocate the following two lines close to the beginning of the output:
# error "ROOT requires support for C++11 or higher."
# error "Pass `-std=c++11` as compiler argument."
which says that ROOT related program needs to be compiled with c++11 standard. Follow this instruction:
$ g++ -std=c++11 -I /path/to/include/ test.cc
You will get some new error messages:
/tmp/ccmm4zlO.o: In function `main':
test.cc:(.text+0x1c): undefined reference to `TRandom::TRandom(unsigned int)'
test.cc:(.text+0x2b): undefined reference to `TRandom::Rndm()'
...
This is because TRandom.h only declares the class TRandom and the function Rndm(), but the real definition of them are saved in a separated file called libMathCore.so, which is a shared object (.so), or a shared library file. You need to tell g++ to link your executable with this library:
$ g++ -std=c++11 -I /path/to/include/ test.cc -L /path/to/ROOT/lib -lCore -lMathCore
Now that you have your test.cc compiled to test.exe, you can try to run it as ./test.exe. ./ means current directory. This is a way to tell your shell where to find an executable named test.exe. You can add . in your PATH so that you don’t have to type ./ all the time:
$ export PATH=.:$PATH # add . to the list of folders that contain executables
Check Linux 101 if you don’t know about the environment variable PATH.
Run test.exe and you will get yet again an error message:
test.exe: error while loading shared libraries: libCore.so: cannot open shared object file: No such file or directory
This is because test.exe uses the ROOT library libCore.so and your shell does not know where to find it. Yes, we already told g++ where to find it. But shell and g++ are two different things. We need to instruct them individually. To tell shell where to find some libraries, you need
export LD_LIBRARY_PATH=/path/to/ROOT/lib:$LD_LIBRARY_PATH
For MAC users, replace LD_LIBRARY_PATH with DYLD_LIBRARY_PATH.
It is too much to type such a long command, g++ -Wall -std=c++11 -I..., just to compile a simple program test.cc. You’d better save this command somewhere so that you can use it later. A standard way to do this is to create a Makefile in the same directory as your test.cc, which contains the following two lines of code:
test.exe: test.cc
g++ -std=c++11 -I /path/to/include/ test.cc -o test.exe -L /path/to/ROOT/lib -lCore -lMathCore
This is called a rule. Check the make manual to understand the structure of a Makefile rule. Basically, test.exe is the target. test.cc is the prerequisite of this target. If the target is older than its prerequisite, the command in the second line, or the recipe, will be used to update the target. Otherwise, no action will be taken. Be aware that a recipe must start with a real Tab instead of a few spaces.
Now you can run
$ make
to compile test.cc to test.exe.
We typed test.cc and test.exe in the recipe of the rule above. But they are simply the prerequisite and the target of that rule. In principle, you have already told make all the information it needs. Indeed, make remembers them. You can use two automatic variables in your recipe to refer to them without defining the two variables (that’s why they are called automatic ones):
test.exe: test.cc
g++ -std=c++11 -I /path/to/include/ $? -o $@ -L /path/to/ROOT/lib -lCore -lMathCore
where $? refers to test.cc and $@ refers to test.exe. There is a full list of automatic variables in the make manual.
There is a standard name for each part of the recipe. For example, g++ is called the compiler, -std=c++ -I... are flags. Make maintains a list of standard names (Implicit variables) to refer to individual parts in a recipe. If we use these standard names, the Makefile can be rewritten as
CXXFLAGS = -std=c++11 -I/path/to/include/
LDLIBS = -L/path/to/ROOT/lib -lCore -lMathCore
test.exe: test.cc
$(CXX) $(CXXFLAGS) $? -o $@ $(LDFLAGS) $(LDLIBS)
You don’t have to define $(CXX), since it has a default value of g++.
ROOT provides a command root-config for you to figure out the contents of these standard parts. Run root-config --help to learn more about it. Our Makefile can be modified to work with any machine that have ROOT properly installed:
CXXFLAGS = $(shell root-config --cflags)
LDLIBS = $(shell root-config --libs)
$(shell a-shell-command) is how you call a-shell-command in a Makefile and get the output of it.
Let’s add another C++ source file, gaus.cc into this directory. We need to add another rule in our Makefile to compile it:
test.exe: test.cc
$(CXX) $(CXXFLAGS) $? -o $@ $(LDFLAGS) $(LDLIBS)
gaus.exe: gaus.cc
$(CXX) $(CXXFLAGS) $? -o $@ $(LDFLAGS) $(LDLIBS)
However, if you run make in your terminal, only test.cc will be compiled. This is because make without any argument will only run the first rule. To run a specific rule, you need to pass the target name of that rule to make:
$ make gaus.exe
What if you want to compile both of them? You need to add a special rule that depends on both exe files:
all: test.exe gaus.exe
Keep this as the first rule so that you can run it when you call make without any argument. This rule does not have any recipe. The sole purpose of it is to call rules related to test.exe and gaus.exe.
Unlike test.exe and gaus.exe, all is not really a file. It is just the name of a target. We call it a phony target. To make this point clear so that make won’t do something for a file named all in the directory, we need to add the following line to your Makefile:
.PHONY: all
There are some standard phony targets you may consider to add in your Makefile:
.PHONY: all clean install debug
clean:
$(RM) *.exe
install:
install *.exe ~/bin
debug:
@echo $(CXXFLAGS)
@echo $(LDLIBS)
where $(RM) is another implicit variable. It has a default value of rm -f. The @ in front of echo is to suppress the print out of the recipe itself in the terminal window.
The two rules for test.exe and gaus.exe are very similar. It would be nice if we can combine them in one rule. You can achieve this through the patter-specific variable values:
all: test.exe gaus.exe
%.exe: %.cc
$(CXX) $(CXXFLAGS) $? -o $@ $(LDFLAGS) $(LDLIBS)
This rule can be applied to any pair of .exe and .cc files.
If we want to add another C++ source file, rndm.cc, to the directory, we just need to add rndm.exe to the list after all:
all: test.exe gaus.exe rndm.exe
It would be nice if we can automate this step. You can achieve this using a make function called wildcard:
SRC = $(wildcard *.cc)
EXE = $(SRC:.cc=.exe)
all: $(EXE)
The first line creates a list of all files that end with .cc and save it in $(SRC). The second line changes the suffix of every entry in $(SRC) from .cc to .exe and save the new list in $(EXE). The third line uses this list.
When you add a new .cc file in this directory, you can compile it with make, without modifying the Makefile.
A command in Linux normally does not end with .exe. We can remove it from our final executables:
%: %.cc
$(CXX) $(CXXFLAGS) $? -o $@ $(LDFLAGS) $(LDLIBS)
This way, test.cc will be compiled to test instead of test.exe. This rule is so commonly used that make includes it as an implicit rule. You don’ even need to write it in your Makefile. (Run make -np |less to get a list of all implicit rules.) Remove this rule and your Makefile still works! It seems crazy that we spend so much effort to improve our rules, only to get rid of them at last!
As a final touch, we’d like to add some protections, print out some instructions here and there. Your final Makefile would look like this:
# variables used by implicit rules to allocate ROOT headers and libs
CXXFLAGS = $(shell root-config --cflags)
LDLIBS = $(shell root-config --libs)
SRC = $(wildcard *.cc) # list all files that end with .cc
EXE = $(SRC:.cc=) # remove .cc from those file names
all: $(EXE)
@echo make install: copy $(EXE) to ~/bin
@echo make clean: delete $(EXE)
@echo make debug: check contents of Makefile variables
clean:
$(RM) $(EXE)
install:
mkdir -p ~/bin
install $(EXE) ~/bin
@echo Please add $(shell root-config --libdir)
@echo to your LD_LIBRARY_PATH before you run any executable
debug:
@echo CXXFLAGS = $(CXXFLAGS)
@echo LDLIBS = $(LDLIBS)
@echo SRC = $(SRC)
@echo EXE = $(EXE)
.PHONY: all clean install debug
References for “Final touch”:
Activities mentions in this site have been supported by the following grants: