A makefile is a text file which controls the operation of the make
program. The make
program is typically used to manage the creation of programs from their source files, but it can be more generally used to handle any process where files (or targets) need to be regenerated after other files (or prerequisites) have been modified. The makefile describes the relationship between targets and prerequisites, and also specifies the commands needed to bring the target up-to-date when one or more of the prerequisites has changed. The only way that make
determines "out of date-ness" is by comparing the modification time of target files and their prerequisites.
Makefiles are somewhat unique in a few ways which can be confusing initially.
First, a makefile consists of two completely different programming languages in the same file. The bulk of the file is written in a language that make
can understand: this provides variable assignment and expansion, some preprocessor capabilities (including other files, conditional parsing of sections of the file, etc.) as well as the definition of targets and their prerequisites. In addition, each target can have a recipe associated with it which specifies which commands should be invoked to cause that target to be brought up-to-date. The recipe is written as a shell script (POSIX sh by default). The make
program doesn't parse this script: it runs a shell and passes the script to the shell to be run. The fact that recipes are not parsed by make
, but instead handled by a separate shell process, is central in understanding makefiles.
Second, a makefile is not a procedural language like a script: as make
parses the makefile it constructs a directed graph internally where targets are the nodes of the graph and the prerequisite relationships are the edges. Only after all makefiles have been completely parsed and the graph is complete will make
choose one node (target) and attempt to bring it up to date. In order to ensure a target is up to date, it must first ensure that each of that target's prerequisites are up to date, and so on recursively.
Name | Also Known As | Initial Version | Version | Release Date |
---|---|---|---|---|
POSIX make | 1992 | IEEE Std 1003.1-2008, 2016 Edition | 2016-09-30 | |
NetBSD make | bmake | 1988 | 20160926 | 2016-09-26 |
GNU make | gmake | 1988 | 4.2.1 | 2016-06-10 |
SunPro make | dmake | 2006 | 2015-07-13 | |
MSVS nmake | 2003 | 2015p3 | 2016-06-27 |
Consider writing a "hello world!" program in c. Lets say our source code is in a file called source.c, now in order to run our program we need to compile it, typically on Linux (using gcc) we would need to type $> gcc source.c -o output
where output is the name of the executable to be generated. For a basic program this works well but as programs become more complex our compilation command can also become more complex. This is where a Makefile comes in, makefiles allow us to write out a fairly complex set of rules for how to compile a program and then simply compile it by typing make on the command line. For instance here is a possible example Makefile for the hello wold example above.
Basic Makefile
Lets make a basic Makefile and save it to our system in the same directory as our source code named Makefile. Note that this file needs to be named Makefile, however the capitol M is optional. That said it is relatively standard to use a capitol M.
output: source.c
gcc source.c -o output
Note that there is exactly one tab before the gcc command on the second line (this is important in makefiles). Once this Makefile is written every time the user types make (in the same directory as the Makefile) make will check to see if source.c has been modified (checks the time stamp) if it has been modified more recently than output it will run the compilation rule on the following line.
Variables in Makefiles
Depending on the project you may want to introduce some variables to your make file. Here is an example Makefile with variables present.
CFLAGS = -g -Wall
output: source.c
gcc $< $(CFLAGS) -o $@
Now lets explore what happened here. In the first line we declared a variable named CFLAGS that holds several common flags you may wish to pass to the compiler, note that you can store as many flags as you like in this variable. Then we have the same line as before telling make to check source.c to see if it has been changed more recently than output, if so it runs the compilation rule. Our compilation rule is mostly the same as before but it has been shortened by using variables, the $<
variable is built into make (referred to as an automatic variable see https://www.gnu.org/software/make/manual/html_node/Automatic-Variables.html) and it always stands for the source so in this case source.c. $(CFLAGS)
is our variable that we defined before, but note that we had to put the variable in parenthesis with a $ in front like this$(someVariable)
. This is the syntax for telling Make to expand the variable out to what you typed before. Finally we have the $@ symbol, once again this is a variable built into make, and it simply stands for the target of the compilation step, so in this case it stands for output.
Clean
Make clean is another useful concept to learn about make files. Lets modify the Makefile from above
CFLAGS = -g -Wall
TARGETS = output
output: source.c
gcc $< $(CFLAGS) -o $@
clean:
rm $(TARGETS)
As you can see we simply added one more rule to our Makefile, and one additional variable that contains all of our targets. This is a somewhat common rule to have in makefiles as it allows you to quickly remove all of the binaries you produced simply by typing $> make clean
. By typing make clean you tell the make program to run the clean rule and then make will run the rm command to delete all of your targets.
I hope this brief overview of using make helps you speed up your workflow, Makefiles can become very complex, but with these ideas you should be able to get started using make and have a better understanding of what is going on in other programmers Makefiles. For more information about using make an excellent resource is https://www.gnu.org/software/make/manual/.
A rule describes when and how certain files (rule's targets) are created. It can also serve to update a target file if any of the files required for its creation (target's prerequisites) are newer than the target.
Rules follow the syntax below: (Note that commands following a rule are indented by a tab)
targets: prerequisites
<commands>
where targets and prerequisites are file names or special reserved names and commands (if present) are executed by a shell to build/rebuild targets that are out-of-date.
To execute a rule one can simply run the make
command in the terminal from the same directory where the Makefile resides. Running make
without specifying the target, will execute the first rule defined in the Makefile. By convention, the first rule in the Makefile is often called all or default, commonly listing all valid build targets as prerequisites.
make
only executes the rule if the target is out-of-date, meaning either it doesn't exist or its modification time is older than any of its prerequisites. If the list of prerequisites is empty, the rule will only be executed when it is first invoked to build the targets. However, when the rule does not create a file and the target is a dummy variable, the rule will always be executed.
Pattern rules are used to specify multiple targets and construct prerequisite names from target names. They are more general and more powerful compared to ordinary rules as each target can have its own prerequisites. In pattern rules, a relationship between a target and a prerequisite is build based on prefixes including path names and suffixes, or both.
Imagine we want to build the targets foo.o
and bar.o
, by compiling C scripts, foo.c
and bar.c
, respectively. This could be done by using the ordinary rules below:
foo.o: foo.c
cc -c $< -o $@
bar.o: bar.c
cc -c $< -o $@
where automatic variable $<
is the name of the first prerequisite and $@
the name of the target (A complete list of automatic variables can be found here).
However, as the targets share the same suffix, the above two rules can now be substituted by the following pattern rule:
%.o: %.c
cc -c $< -o $@
Implicit rules tell make
how to use customary methods to build certain types of target files, which are used very often. make
uses the target file name to determine which implicit rule to invoke.
The pattern rule example we saw in the previous section, does not actually need to be declared in a Makefile as make
has an implicit rule for C compilation. Thus, in the following rule, the prerequisites foo.o
and bar.o
will be build using the implicit rule for C compilation, before building foo
.
foo : foo.o bar.o
cc -o foo foo.o bar.o $(CFLAGS) $(LDFLAGS)
A catalogue of implicit rules and the variables used by them can be found here.
if a directory contain 2 files:
$ ls
makefile
example.txt
and makefile
contain the following text
%.gz: %
gzip $<
then you can obtain example.txt.gz
by typing in the shell
$ make -f makefile example.txt.gz
the makefile consist of only one rule that instruct make how to create a file whose name end with .gz if there is a file with the same name but the .gz suffix.
C:\makefile:
helloWorld :
[TAB]echo hello world
run results:
C:\>make
echo hello world
hello world
Note: [TAB] should be replaced by an actual tab, stackoverflow replaces tabs with spaces, and spaces are not used the same as tabs in a makefile.