C Essense (5) - Makefile

1. Basic Rule

Suppose we have a project, structure is like this:

├── main.c
├── main.h
├── maze.c
├── maze.h
├── stach.h
└── stack.c

stack is our self-implemented stack, and maze is a maze game.

To compile them together, I can run gcc main.c stack.c maze.c -o main.

If I modify maze.c, then I need to re-compile all the files.

It's better to do it this way:

$ gcc -c main.c
$ gcc -c stack.c
$ gcc -c maze.c
$ gcc main.o stack.o maze.o -o main

Then if I modified maze.c, the only thing I need to do is:

$ gcc -c maze.c
$ gcc main.o stack.o maze.o -o main

To avoid typo, we can create a Makefile:

main: main.o stack.o maze.o
    gcc main.o stack.o maze.o -o main

main.o: main.c main.h stack.h maze.h
    gcc -c main.c

stack.o: stack.c stack.h main.h
    gcc -c stack.c

maze.o: maze.c maze.h main.h
    gcc -c maze.c

The format of makefile rule is:

target ... : prerequisites ...

The first rule in makefile is the default rule.

For every command start with Tab in makefile, make will create a Shell process to execute it.

make will only recompile the target that is modified, by checking the last modified time of source file .c .h, and target file .o.

To Summarize, target needs to be updated if one of the criteria is meet:

  • target has not been run.
  • one of the prerequisites needs to be updated
  • some prerequisites's last modified time is later than current target

Normally Makefile will have a clean target:

    @echo "cleanning project"
    -rm main *.o
    @echo "clean completed"

So we can run make clean. @ means does not display command itself, and only display the result. By default if one command fails, the target is terminated. - means even if command fails, the target should continue.

There is a list of special built-in target, like .PHONY. Check this link about GNU Special Targets

GNUMake also has some common target, that normally we follow the convention:

  • all: default target
  • install
  • clean: clean the generated binary files
  • distclean

The name of makefile is not required to be Makefile, but it is recommended. The order of looking for makefile is GNUmakefile -> makefile -> Makefile.

2. Implicit rule & Pattern rule

One rule can be written in multiple sub-rules, but only one rule should have the commands.

main.o: main.h stack.h maze.h

main.o: main.c
    gcc -c main.c

Example in first section can be written as:

main: main.o stack.o maze.o
    gcc main.o stack.o maze.o -o main

main.o: main.h stack.h maze.h

stack.o: stack.h main.h

maze.o: maze.h main.h

main.o: main.c
    gcc -c main.c

stack.o: stack.c
    gcc -c stack.c

maze.o: maze.c
    gcc -c maze.c

    -rm main *.o

.PHONY: clean

There are some rules that can be omitted:

main: main.o stack.o maze.o
    gcc main.o stack.o maze.o -o main

main.o: main.h stack.h maze.h

stack.o: stack.h main.h

maze.o: maze.h main.h

    -rm main *.o

.PHONY: clean

When we run make:

$ make
cc -c -o main.o main.c
cc -c -o stack.o stack.c
cc -c -o maze.o maze.c
gcc main.o stack.o maze.o -o main

If all the prerequisites of a target do not have command list, then make will try built-in implicit rule. Use make -p to list all implicit rules.

In the example, we have used:

# default

# default
CC = cc

# default

%.o: %.c
# commands to execute (built-in):
        $(COMPILE.c) $(OUTPUT_OPTION) $<

CC is a makefile variable. CC=cc is the declaration, and $(CC) gets its value. cc points to the C compiler. It could be gcc or clang depends on your system. CFLAGS, CPPFLAGS, TARGET_ARCH here are all empty. COMPILE.c command becomes cc -c.

$@ fetches the target in the rule, and $< fetches the first prerequisites.

%.o: %.c is a pattern rule. main.o matches this %.o pattern, so %=main. Thus it will run cc -c -o main.o main.c.

stack.o and maze.o also match %.o target.

For multiple target rule:

target1 target2: prerequisite1 prerequisite2 command $&lt; -o $@

It becomes:

target1: prerequisite1 prerequisite2
    command prerequisite1 -o target1
target2: prerequisite1 prerequisite2
    command prerequisite1 -o target2

3. Variables

In makefile, variable declaration can be put behind calling.

main.o: main.c
    $(CC) $(CFLAGS) $(CPPFLAGS) -c $<
CC = gcc
CFLAGS = -O -g
CPPFLAGS = -Iinclude

Expanded command is gcc -O -g -Iinclude -c main.c.

  • CFLAGS: compilation options, like -O, -g.
  • CPPFLAGS: preprocessing options, like -D, -I.

When declaration is behind calling, it is easy to write calling cycle, like this:

A = $(B)
B = $(A)

An example:

foo = $(bar)
bar = Huh?
    @echo $(foo)

The value of foo is determined in all:, not at line 1.

If you want variable to be evaluated immediately, use :=.

Another example:

y := $(x) bar
x := foo

y will be evaluated as bar since x is empty string at the time of evaluating.

Operator ?= means, only assign value if it has not been assigned. It is like ||= in Ruby.

+= is also allowed in makefile.

Other special variables:

  • $?: all the prerequisite that needs to be updated in a rule, as a list
  • $^: all the prerequisite, as a list

    main: main.o stack.o maze.o gcc main.o stack.o maze.o -o main

can be writter as:

main: main.o stack.o maze.o
    gcc $^ -o $@

libsome.a: foo.o bar.o lose.o win.o
        ar r libsome.a $?
       ranlib libsome.a

This is for build library file. ar is a archive tool.

Some frequently used makefile variables:

Variable Name Default value Description
AR ar archive tool
AS as assembly compiler
CXX g++ C++ compiler name
CPP $(CC) -E C preprocessor name
LD ld linker name
TARGET_ARCH target platform options
LINK.o $(CC) $(LDFLAGS) $(TARGET_ARCH) link .o files.
LINK.c $(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH) link .c files
$(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH)` link .cc files, which are cpp files.
COMPILE.c $(CC) $(CFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c compile c files
COMPILE.cc $(CXX) $(CXXFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c compile cc files
RM rm -f remove command

4. Header file dependencies

Our makefile looks like this now:

all: main
main: main.o stack.o maze.o gcc $^ -o $@
main.o: main.h stack.h maze.h
stack.o: stack.h main.h
maze.o: maze.h main.h
	-rm main *.o
.PHONY: clean

One problem is that, we need to check the source code to determine header files dependencies. If we update source file, we may forget to update makefile.

gcc -M auto generate target file and source file dependencies:

$ gcc -M main.c
main.o: main.c /usr/include/stdio.h /usr/include/features.h \
/usr/include/sys/cdefs.h /usr/include/bits/wordsize.h \ /usr/include/gnu/stubs.h /usr/include/gnu/stubs-32.h \ /usr/lib/gcc/i486-linux-gnu/4.3.2/include/stddef.h \ /usr/include/bits/types.h /usr/include/bits/typesizes.h \ /usr/include/libio.h /usr/include/_G_config.h
/usr/include/wchar.h \ /usr/lib/gcc/i486-linux-gnu/4.3.2/include/stdarg.h \ /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
main.h \
  stack.h maze.h

If we don't want system header file dependencies, use -MM:

$ gcc -MM *.c
main.o: main.c main.h stack.h maze.h maze.o: maze.c maze.h main.h stack.o: stack.c stack.h main.h

Next problem is, how to put the dependencies into makefile:

sources = main.c stack.c maze.c
include $(sources:.c=.d)
%.d: %.c
    set -e; rm -f $@; \
    $(CC) -MM $(CPPFLAGS) $< > $@.$$$$; \
    sed 's,\($*\)\.o[ :]*,\1.o $@ : ,g' < $@.$$$$ > $@; \
    rm -f $@.$$$$

$(sources:.c=.d) is a substitution syntax, which means substitute all the .c to .d in sources. include $(sources:.c=.d) becomes:

include main.d stack.d maze.d

include means read other makefiles.

You don't have .d files for now, so make tries the pattern rule %.d: %.c. The four lines below, uses only one make process to run it. set -e means, in current process, if any of the command returns non-zero, the process terminates. rm -f $@ removes previous generated .d file. $$ is the process id. We use sed tool to generate .d makefiles.

5. make command options

  • -n print the compiled command, but not execute it.
  • -C is used change directory and run a makefile there.