A shell-based build tool
shmake is a build tool which is driven by POSIX shell scripts, specialized around gcc and compilers which behave like it. It deliverately lacks ambition to be an universal build tool like lake, but aims to make the simple stuff easy and the complicated stuff easier.
For instance, say we have a little project with two source files, 'hello.c' and 'common.c', with 'hello.c' including 'common.h'. This would be the shmakefile.
#!/bin/sh . /tmp/shmake.sh C hello *.c
Shmakefiles are executable shell scripts - the second line sources any special magic which shmake needs to provide. But they are not executed directly. When shmake is run, it will look for a shmakefile, in time-honoured fashion.
simple$ shmake compiling common.o compiling hello.o linking hello simple$ shmake simple$ # nothing happened (shmake is quiet) simple$ touch common.c simple$ shmake compiling common.o linking hello # rebuild after change to common.c simple$ touch common.h simple$ shmake compiling hello.o linking hello # rebuild hello.c after change to included common.h
So the single 'C hello *.c' gives us a build which already tracks all the dependencies. It does this by using GCC's -MMD flag and reading the resulting .d files, saving you the most tedious part of writing makefiles.
By default it gives you an optimized stripped executable, and does not show you the actual compilation. The '-g' flag will give you a debug build, and '-v' will make shmake more verbose and chattery:
simple$ shmake clean simple$ shmake -g -v gcc -c -Wall -MMD -g common.c -o common.o gcc -c -Wall -MMD -g hello.c -o hello.o gcc common.o hello.o -o hello simple$ shmake clean simple$ shmake -v gcc -c -Wall -MMD -O2 common.c -o common.o gcc -c -Wall -MMD -O2 hello.c -o hello.o gcc common.o hello.o -Wl,-s -o hello
Just to make things a little easier; to create an initial shmakefile, use the '-c' option
simple$ shmake -c 'C hello *.c' shmakefile created simple$ cat shmakefile #!/bin/sh . /tmp/shmake.sh C hello *.c
This is all very cute, but can it do real projects? Here is a shmakefile for shmake, in the 'tests/self' directory. First the simplified version, assuming that the source has been copied up:
#!/bin/sh . /tmp/shmake.sh SRC=../.. C99 shmake -I$SRC -L$SRC/llib -lllib shmake.c lib.c utils.c
It is not a complicated project (only about 1200 lines in total) but leans heavily on llib. Here we specify the include and library directories, and the libs to link against. C99 means we specifically want '-std=std99'. This is pretty much how we would write a build as a shell script, except that this takes care of the dependency tracking.
This is entirely equivalent, except using the S ('set') command for specifying default flags. It's clearer for more complicated projects, plus it applies to any program target unless they explicitly override the defaults.
#!/bin/sh . /tmp/shmake.sh SRC=../.. S includes $SRC S lib-dirs $SRC/llib S libs llib C99 shmake shmake.c lib.c utils.c
The set variables are, with their compile command flag equivalents:
- includes (-I) include directorie
- defines (-D) preprocessor defines
- lib-dirs (-L) directories to search for libraries
- libs (-l) libraries
- needs (-n) any 'needs' (see next section)
- need-path extra dir to find needs (ditto)
- out-dir (-d) directory for object files and dependency (.d) files
- cflags any extra compilation flags
- lflags any extra link flags
- opt (-O) optimization level (default '2' - hence '-O2')
- debug (-g) debug build
- exports (-e) executable exports its symbols
- slack (-S) don't use "-Wall"
- quiet (-q) only generate output on error
Most of these are additive, so that setting a variable multiple times will add new values.
The -d flag is useful for keeping your source directory nice and clean, since the .o and .d files will have their own directory. The output directory will be created if it does not already exist; if it is the special word 'auto', then its name is a combination of the compiler and the build type, e.g. 'gcc-release' or 'clang-debug'. This makes switching between a release and debug build more efficient.
If you had a number of projects depending on llib or some other external dependency, then needs are a useful concept. A borrowing from Lake, needs are a general way of looking up the compile and link flags. A need can be defined by a file with the extension '.need', either in current dir or in ~/.shmake. Additionally, a private need path can be provided with 'S need-path ' and this will be checked as well. It is defined as a script which echoes at least one of 'cflags ...' or 'libs ... ':
self$ cat ~/.shmake/llib.need #!/bin/sh # simple file providing a Need echo cflags -I/home/user/dev/c/llib echo libs -L/home/user/dev/c/llib/llib -lllib self$ cat need.shmak #!/bin/sh . /tmp/shmake.sh C99 shmake -n llib shmake.c lib.c utils.c self$ shmake -v -f need.shmak gcc -c -Wall -MMD -std=c99 -I/home/user/dev/c/llib -O2 shmake.c -o shmake.o gcc -c -Wall -MMD -std=c99 -I/home/user/dev/c/llib -O2 lib.c -o lib.o gcc -c -Wall -MMD -std=c99 -I/home/user/dev/c/llib -O2 utils.c -o utils.o gcc shmake.o lib.o utils.o -L/home/user/dev/c/llib/llib -lllib -Wl,-s -o shmake
If there's no such file, we ask
-n/---needs flag generally takes a space-separated list of needs - it cannot be used multiply like '-I'.
(Note: in the initial release of shmake, the format of .need files resembled .pc files as understood by 'pkg-config`. However, it was different from the actual .pc file format, and did not allow the flexibility of 'shell everywhere'. Defining need files as scripts allows for checking for the existence and location of packages; they may also be written in Python or any other language if you feel frustrated.)
A Step Back: Targets
Underneath, shmake is very much like make, and allows the same style of dependency-based programming.
For the simple example, we could perform the build explicitly:
#!/bin/sh . /tmp/shmake.sh COMPILE='gcc -c @(INPUT) -o @(TARGET)' LINK='gcc @(DEPS) -o @(TARGET)' T hello.o hello.c common.h "$COMPILE" T common.o common.c "$COMPILE" T hello hello.o common.o "$LINK" all hello
The T ('target') command is of the form 'T target (prequisites) shell-command'. 'Prerequisites' is the make term for 'inputs that our target depends on'. So 'hello.o' depends on hello.o and on common.h. (This is a clunky word so I will just call them 'inputs'.) The shell-command can contain @() expansions of the three variables TARGET (first name) INPUT (first input) and DEPS (all inputs). The targets are hello.o, common.o and hello itself. We need the 'dummy' target 'all' ('all a b c' is short for 'T all a b c none') so that shmake knows what default target to build. Otherwise, just as in make, it would just compile hello.c and stop.
This style is tedious for actually building projects because you must manually track all the inputs your targets depend on. For gcc and compatible compilers, the marvelous -MMD flag (create .d dependency file) helps tremendously and that's the big convenience of the C compile command. But it does mean you can use shmake with any set of commands, using all the power of shell scripting.
If there are a lot of targets using the same command, then rules are useful. For instance, in tests/self/shmakefile, there is:
SRC='../..' R copy-files -d $PWD ditto 'cp @(INPUT) @(TARGET)' $SRC/*.c $SRC/*.h
A rule takes the following arguments:
- it has a name
- an output extension. 'ditto' means that the out extension is the in extension.
- a command, just as with a target
- a set of input files
- can use -d to set the output directory, which you must do with 'ditto'!
The name is not a target, although it can be used in the same contexts; it is a group of related targets, generated by the rule. Rules can be seen as target factories.
Rules, Groups and the 'C' Command
It is common enough to want to make a set of simple executables using the same rule. The
can do this, but then you lose the special compiling support. We have a number of source files, and each
needs to be compiled to an executable of the same name. The
-R flag does this:
C progs -R '' *.c all progs
The name 'progs' is again a group name, and represents a list of targets. When a group name is
specified as an input or dependency, then it expands out that list of targets - it is not itself a target.
Like with the
R command, you need to specify the output directory, here just the empty string - you may
use the special name 'exe' instead. So
-R, like the
R command, is a target factory.
C command creates both the compile targets, plus the final link target. The
C only produce the compile targets. It is still necessary to give a name but then the
name becomes the group name. This name can be passed to another
C command and gets
added to the list of object files that will be linked.
C objs -G *.c C prog objs
Why would you want to separate out compiling and linking like this? When some files need to be compiled differently, say C files and C++ files.
I have already mentioned -v for verbose; if you double this up as -vv then you will get a detailed dump of dependency checks. -t for testing is useful if you just want to see what shmake will do, without actually executing the commands.
'-f' has the same meaning as in make - use a named file as the shmakefile. '-C' also means 'switch to directory first'.
shmake provides three predefined variables to shmakefiles: CC (the C compiler),
CXX (the C++ compiler) and PLAT, which is the value of
The compilers are initialized to the values found, e.g CC is either 'gcc' or 'cc' depending on what
exists on the path.
The non-flag arguments are either VAR=VALUE assignments or targets. So 'shmake CC=clang' will override the default value of CC. Currently only one explicit target is supported; the default is 'all'. There is a predefined target 'clean' which removes all targets representing files.
Please note an important 'gotcha' with shmake - the shmakefile is executed, and its output is then processed by shmake. Any variable assignments are not passed to shmake itself, because they are created in a subshell. Any actions are executed from shmake directly, and so do ensure that any variables within them have already been expanded.
Example: Building Lua
For a complete programming language, Lua is relatively easy to build. But the makefile is a bit convolved and obscures what's going on. This shmakefile makes things a lot clearer.
#!/bin/sh . /tmp/shmake.sh LUA_LIB=liblua52.a S defines LUA_COMPAT_ALL S exports true S libs readline S libs m echo "Building Lua for $PLAT platform" case $PLAT in freebsd) S defines LUA_USE_LINUX ;; Linux) S defines LUA_USE_LINUX S libs dl ;; ansi) S defines LUA_ANSI ;; posix) S defines LUA_USE_POSIX ;; solaris) S defines LUA_USE_POSIX LUA_USE_DLOPEN S libs dl ;; darwin) S defines LUA_USE_MACOSX ;; *) echo "unsupported platform $PLAT. One of freebsd,Linux,Darwin,posix,ansi,solaris" exit 1 ;; esac C $LUA_LIB *.c -x 'lua.c luac.c' C lua lua.c $LUA_LIB C luac luac.c $LUA_LIB all lua luac
Note the very useful '-x' (for 'exclude') flag that avoids the need to write out all those C files! Apart from being clearer, this shmakefile immediately can give youa debug build with the global '-g' flag.
If nothing else, shmake is a demonstration of a useful tactic: using shell script as your program's domain specific language (DSL). In this way, a deeper knowledge of shell translates directly into a more powerful way to build projects.