Myke - My Make Replacement

Myke is a make-like program I wrote several years ago to circumvent the tedious detail you have to put into Makefiles. Myke's syntax is a superset of make's so if you're already adept with make you're halfway there. Most of the myke extensions have to do with expressivity. For example, you can say this:

	CFILES	= $("*.c" G)		# glob up all the C sources
	OFILES	= $((CFILES P)).o	# enumerate the CFILES, strip the .c,
					# add a .o
	it: $(OFILES)
		cc -o $@ $(OFILES)
	%.o: %.c
		cc -c $?

Myke also allows you to put these rules in a configuration file, thus saving enumerating them in every Mykefile. Taken to an extreme you can omit the Mykefiles altogether.

Here is my .mykerc file as an example.

Here is the manual entry: web page, source form.

Installing Myke

Prebuilt Binaries

If you have one of the machines below you can save yourself the trouble of building myke. The following executables have libcs staticly linked, so you don't need to fetch the library, just the myke executable.

linux.x86.linux: bin/myke, lib/libcs.a.
sgi.mips.irix: bin/myke, lib/libcs.a.
sun.sparc.solaris: bin/myke, lib/libcs.a.
linux.alpha.linux: bin/myke, lib/libcs.a.

The Source

If you don't have one of the above machines, you'll want the source.

Licence: you can do what you like with this, but if you give it to someone else make sure they know I wrote it, and if you modify it make sure the mods are clearly marked as yours. If you make fixes or improvements, I'd appreciate patches and descriptions sent to me for incorporation in my central source. Naturally you'll be credited with such changes.

The source for myke comes in three parts:

myke.tar.gz - the myke source itself
libcs.tar.gz - the supporting library libcs.a on which many of my tools are built
csinc.tar.gz - the libcs include files: a toplevel libcs.h and a subdirectory cs/ holding the bits

It's maintained with itself, so you will also want tools.tar.gz which is a small collection of shell scripts used to bootstrap the compile. Currently these are: cccs, a simple wrapper for the cc command line, and gccanal, a simple wrapper for gcc which turns on lots of warnings. cccs calls gccanal, so if you want to use another cc command line then cccs is the script to hack. (Oh yeah, you'll have to hack cccs anyway to change the csinc= line at the top to point at the include directory you'll be making below.)

Extraction

Make a directory in which to work:
mkdir myketmp cd myketmp
Download the three components above.
Unpack the components:
gunzip <myke.tar.gz | tar xvf - gunzip <libcs.tar.gz | tar xvf - mkdir include ( cd include gunzip <../csincs.tar.gz | tar xvf - ) mkdir bin ( cd bin gunzip <tools.tar.gz | tar xvf - )

Configuration

Porting myke is mostly a matter of porting libcs, a usually straightforward task. With luck, there will already be a target for your architecture.

Go to the include/cs/config directory.
Bring up the file config in an editor.
Choose a name for your site. This is a four component name of the form:
systemid.vendor.cpu.os
An ls of the site subdirectory will show you typical names used so far.
The systemid is some useful name for the group of machines for which you'll be compiling myke. For example, while at OTC doing some of the early work on myke I used the name "otc" for the systemid there. A local domain name is also often good.
The vendor is the general name for the machines to run this binary.
The cpu is the general class of CPU used. For example, m68k denotes the whole Motorola M680x0 series of CPUs. Hopefully you'll find your CPU listed already.
The os is the general type of operating system you're running and is typically a vendor-specific name. For example, most Silicon Graphics machines run IRIX, and the name irix is used.
Add a clause at the bottom of the config file for your site. Hopefully you can just pick a promising looking one from the collection already there.

If not, you'll probably want to make new _vend_* and possibly _cpu_* clauses.
On most modern machines the best approach it to claim to have a POSIX-compliant ANSI-C machine and to add exceptions as needed if libcs fails to build. Thus clause for most modern machines can usually start out as a simple:
_vend_vendor_osname : STDC : POSIX M vendor vendor M system osname
You may later have to add some extra options if you run into compile problems. Contact me if things start getting weird.
Set the environment variable SYSTEMID to the systemid string chosen above.
Set the environment vaiable ARCH to the vendor.cpu.os string chosen above.
Rebuilt the templates and site-specific includes. Run the config.pl script:
cd include/cs/config perl config.pl
You will need to repeat this step whenever you amend the config file.

Building `libcs.a`

Go to the libcs directory.
Edit the cccs script to set csinc to the full pathname of the include directory. Make sure cccs is in your execution path.
Use cccs to do a test compile of a2gid.c:
cccs -c a2gid.c
If this fails, amend the config file you edited earlier, rerun config.pl, and try it again.
If you're using gcc (cccs does by default), expect to have to ignore warnings about:
- "value computed is not used": this comes out of the assertions in the code.
- "assignment discards `const' from pointer target type": this comes out of a bunch of functions; I've really got to clean this stuff up with regard to const.
Once a2gid.c compiles, attempt a compile of the whole library:
cccs -c *.c 2>cc.err && ar cv libcs.a *.o
If you run into problems, amend the configuration and try again. On some platforms you may need to run ranlib on the final library.
On GNU/FSF platforms (where they can't seem to write compatible tools) you want to say:
cccs -c *.c 2>cc.err && ar rsv libcs.a *.o

Building `myke`

Go to the myke directory.
Compile myke:
cccs -c -DGLOBALMYKERC=/usr/local/etc/mykerc *.c cc -o myke *.o ../libcs/libcs.a

Testing `myke`

Install my .mykerc file as your own. Install the myke executable you just made somewhere suitable (eg $HOME/bin). Go to the libcs directory and say:

myke