Difference between revisions of "SBR600 make and Makefiles"
Chris Tyler (talk | contribs) |
Chris Tyler (talk | contribs) |
||
Line 1: | Line 1: | ||
− | + | [[Category:SBR600]] | |
''make'' is a specialized scripting language used to build software. Unlike most scripting languages, commands are not executed in linear (start-to-finish) sequence; instead, command sequences are defined in terms of what input they accept and what output they produce, and ''make'' automatically sequences the commands to produce the required output. | ''make'' is a specialized scripting language used to build software. Unlike most scripting languages, commands are not executed in linear (start-to-finish) sequence; instead, command sequences are defined in terms of what input they accept and what output they produce, and ''make'' automatically sequences the commands to produce the required output. | ||
Revision as of 09:48, 13 January 2011
make is a specialized scripting language used to build software. Unlike most scripting languages, commands are not executed in linear (start-to-finish) sequence; instead, command sequences are defined in terms of what input they accept and what output they produce, and make automatically sequences the commands to produce the required output.
Running the make
command by itself will execute the makefile script named Makefile
or makefile
in the current directory.
Targets and Dependencies
Picture a very simple build, where the file test.c
is compiled by gcc into the executable binary named test
.
test.c -> compiled by the command 'gcctest.c -o test' -> test
The binary executable product file, test
, is considered the target -- the object to be built. The file test.c
is a dependency - a file that is required in order to produce the target. gcc is the command that builds the target from the dependency.
In a make script (typically called a Makefile), the syntax looks like this:
target: dependencies commands
(Note that the commands must be indented by tabs and not spaces in most versions of make).
The example above could be written:
test: test.c gcc test.c -o test
Here is the result when this Makefile is executed:
$ ls -l total 8 -rw-rw-r--. 1 chris chris 35 Jan 10 19:21 Makefile -rw-rw-r--. 1 chris chris 40 Jan 10 19:15 test.c $ cat Makefile test: test.c gcc -o test test.c $ make gcc -o test test.c
When executed a second time, make does nothing:
$ make make: `test' is up to date.
This is because the timestamp on the target (test
) is later than the timestamp on the dependency (test.c
). If the dependency has been changed since the target was built, though, then make will rebuild the target.
Complex Dependencies
A more complicated build will involve a number of targets and dependencies. C programs, for example, can be compiled into intermediate files, called object files (.o extension), which can then be combined to produce executables.
Picture this scenario:
- There are three object files:
- double.c, number.h, and sauce.h compile to make: double.o
- half.c, number.h, and sauce.h compile to make: half.o
- sauce.c compiles to make: sauce.o
- There are two binary targets:
- double.o and sauce.o can be linked to produce: double
- half.o and sauce.o can be linked to produce: half
The Makefile for these relationships may be written like this:
CC=cc CFLAGS=-O3 all: half double half: half.o sauce.o ${CC} ${CFLAGS} -o half half.o sauce.o double: double.o sauce.o ${CC} ${CFLAGS} -o double double.o sauce.o half.o: half.c number.h ${CC} ${CFLAGS} -c half.c double.o: double.c number.h ${CC} ${CFLAGS} -c double.c sauce.o: sauce.c ${CC} ${CFLAGS} -c sauce.c
There are several things worth noting about this Makefile:
- Variables are used for the name of the compiler and the compiler flags. This makes it very easy to change these values -- to use the gcc compiler, for example, the CC variable could simply be changed to gcc. If variables were not used, you would have to change every line that invoked the compiler.
- all is a dummy target. Since it appears as the first target in the file, it is executed first. It depends on the half and double files, which will be built in order to satisfy the dependency. However, the all target does not specify any commands, and the file
all
will never be built.
When make is executed the first time, five compilations are performed:
$ make cc -O3 -c half.c cc -O3 -c sauce.c cc -O3 -o half half.o sauce.o cc -O3 -c double.c cc -O3 -o double double.o sauce.o
Note that the commands are not being executed in the order in which they appear in the file -- instead, they are ordered according to dependencies.
When executed a second time, no compilations are performed:
$ make make: Nothing to be done for 'all'.
If the file half.c
was edited or the datestamp was updated, running make would execute two compilations:
$ touch half.c $ make cc -O3 -c half.c cc -O3 -o half half.o sauce.o
This reveals the power of make -- it does the absolute minimum in order to build the specified target.
On a large programming project, a binary may be comprised of hundreds or even thousands of source files, and compiling all of those files may take hours. If a software developer edits just one file, it's a waste of time to rebuild everything, so make can save a lot of time -- especially when the software is rebuilt many thousand times.