OCamlMakefile - A Simple Generic Makefile for OCaml-Projects
- GNU-Make version 3.80 or higher
-
It is well-tested across multiple platforms and has been used in many projects.
-
It generates dependencies correctly by ensuring that all automatically generated OCaml-files exist before dependency calculation. This is the only way to guarantee that
ocamldep
can do its job. -
Convenience. Even fairly complex compilation processes (see example
calc.ml
) need only little information to work correctly, sometimes just about the minimum (filenames of sources).
-
It may not be a good choice in projects where many compilation units require different flags.
-
Though it can scale to medium-sized projects, large projects with, for example, dependencies across multiple libraries in different directories are not well-supported.
This is a general shortcoming of the already somewhat dated
make
. You may want to investigate the following tools to approach larger projects:
It is recommended that first-time users take a look at the examples in the
distribution for a quick introduction. OCamlMakefile
-projects are often so
simple that they are self-explanatory.
To create your own project, first edit a project-specific Makefile
in the
appropriate directory. There are two ways of making use of OCamlMakefile
:
-
Have a look at the default settings in
OCamlMakefile
and set them to the values that are valid on your system. For example, check whether the path to the standard libraries is ok, what executables shall be used, etc. Copy it into the directory of the project to be compiled. Add the following statement as last line to yourMakefile
:-include OCamlMakefile
-
Put
OCamlMakefile
somewhere else in your system. In this case you will have to set the variableOCAMLMAKEFILE
in your project-specificMakefile
. This is the way in which the examples are written. Now you only need one version ofOCamlMakefile
to manage all of your projects! See the examples for details.
You will usually need to specify two further variables for your project:
SOURCES
(default:foo.ml
)RESULT
(default:foo
)
Put all the sources necessary for a target into variable SOURCES
. Then set
RESULT
to the name of the target. If you want to generate libraries,
you should not specify the suffix (.cma
, .cmxa
, .a
). It will be
added automatically if you specify that you want to build a library.
** Don't forget to add the `.mli`-files, too! **
** Don't forget that the order of the source files matters! **
The order is important, because it matters during linking due to potential
side effects caused at program startup. This is why OCamlMakefile
does not
attempt to partially order dependencies by itself, which might confuse users
even more. It just compiles and links OCaml-sources in the order specified
by the user, even if it could determine automatically that the order cannot
be correct.
The minimum of your Makefile
looks like this (assuming that OCamlMakefile
is in the search path of make
):
-include OCamlMakefile
This will assume that you want to compile a file foo.ml
to a binary foo
.
Otherwise, your Makefile will probably contain something like this:
SOURCES = foo.ml
RESULT = foo
-include OCamlMakefile
Be careful with the names you put into these variables. If they are wrong,
a make clean
might erase the wrong files!
A simple make
will generate a byte-code executable. If you want to change
this, you may add an all
-rule that generates something else. For example:
SOURCES = foo.ml
RESULT = foo
all: native-code-library
-include OCamlMakefile
This will build a native-code library foo.cmxa
(+ foo.a
) from file
foo.ml
.
You may even build several targets at once. To produce byte- and native-code
executables with one make
, add the following rule:
all: byte-code native-code
You will probably want to use a different suffix for each of these targets so that the result will not be overwritten. See the optional variables below for details.
You may also tell make
at the command-line what kind of target to produce
(e.g. make nc
). Here all the possibilities with shortcuts between
parenthesis:
byte-code (bc)
byte-code-nolink (bcnl) - no linking stage
byte-code-library (bcl)
native-code (nc)
native-code-nolink (ncnl) - no linking stage
native-code-library (ncl)
debug-code (dc)
debug-code-nolink (dcnl) - no linking stage
debug-code-library (dcl)
profiling-byte-code (pbc)
profiling-byte-code-library (pbcl)
profiling-native-code (pnc)
profiling-native-code-library (pncl)
byte-code-dll (bcd)
native-code-dll (ncd)
pack-byte-code (pabc)
pack-native-code (panc)
toplevel (top)
subprojs
Here is a short note concerning building and linking byte code libraries with C-files:
OCaml links C-object files only when they are used in an executable. After compilation they should be placed in some directory that is in your include path if you link your library against an executable.
It is sometimes more convenient to link all C-object files into a single C-library. Then you have to override the automatic link flags of your library using
-noautolink
and add another link flag that links in your C-library explicitly.
Concerning maintenance:
-
make clean
removes all (all!) automatically generated files. So again, make sure your variables are ok! -
make cleanup
is similar tomake clean
but keeps executables.
Another way to destroy some important files is by having OCamlMakefile
automatically generate files with the same name. Read the documentation about
the tools in the OCaml-distribution to see what kind of files are generated.
OCamlMakefile
additionally generates (%
is the basename of source file):
%_idl.c
-camlidl
generates a file%.c
from%.idl
, but this is not such a good idea, because when generating native-code, both the file%.c
and%.ml
would generate files%.o
which would overwrite each other. Thus,OCamlMakefile
renames%.c
to%_idl.c
to work around this problem.
The dependencies are stored in three different subdirectories (dot dirs):
._d
- contains dependencies for .ml-files._bcdi
- contains byte code dependencies for .mli-files._ncdi
- contains native code dependencies for .mli-files
The endings of the dependency files are: %.d
for those generated from
%.ml
-files and %.di
for ones derived from %.mli
-files.
This is easy: if you discover a bug, just do a make clean; make dc
to
recompile your project with debugging information. Then you can immediately
apply ocamldebug
to the executable.
To generate code that can be profiled with ocamlprof
(byte code) or gprof
(native code), compile your project with one of the profiling targets (see
targets above). E.g.:
make pbc
will build byte code that can be profiled withocamlprof
.make pnc
will build native code that can be profiled withgprof
.
Please note that it is not currently possible to profile byte code with
threads. OCamlMakefile
will force an error if you try to do this.
A short hint for DEC Alpha-users (under Digital Unix): you may also compile
your sources to native code without any further profiling options/targets.
Then call pixie my_exec
, my_exec
being your executable. This will produce
(among other files) an executable my_exec.pixie
. Call it and it will produce
profiling information which can be analyzed using prof -pixie my_exec
.
The resulting information is extremely detailed and allows analysis up to
the clock cycle level...
Because any kind of program that reads from standard input and prints to standard output can be used as a preprocessor, there cannot be any default way to handle all of them correctly without further knowledge.
Therefore, you have to cooperate a bit with OCamlMakefile
to let
preprocessing happen automatically. Basically, this only requires that you
put a comment into the first line of files that should be preprocessed, e.g.:
(*pp cat *)
(* ... rest of program ... *)
OCamlMakefile
looks at the first line of your files, and if it finds a
comment that starts with "(*pp
", then it will assume that the rest of
the comment tells it how to correctly call the appropriate preprocessor.
In this case the program cat
will be called, which will, of course, just
output the source text again without changing it.
If, for example, you were an advocate of the "revised syntax", which is
supported by the camlp4
preprocessor, you could simply write:
(*pp camlp4r *)
(* ... rest of program in revised syntax ... *)
If you want to write your own syntax extensions, just take a look at the
example in the directory camlp4
: it implements the "repeat ... until
"
extension as described in the camlp4
-tutorial.
OCamlMakefile
contains two targets using ocamlfind
for this purpose:
libinstall
libuninstall
These two targets require the existence of the variable LIBINSTALL_FILES
,
which should be set to all the files that you want to install in the
library directory (usually %.mli, %.cmi, %.cma, %.cmxa, %.a and possibly
further C-libraries). The target libinstall
has the dependency all
to force compilation of the library so make sure you define target all
in your Makefile appropriately.
The targets inform the user about the configured install path and ask for
confirmation to (un)install there. If you want to use them, it is often a
good idea to just alias them in your Makefile to install
and uninstall
respectively.
Two other targets allow installation of files into a particular directory
(without using ocamlfind
):
rawinstall
rawuninstall
There is just one target for this:
top
The generated file can be used immediately for interactive sessions - even with scanners, parsers, C-files, etc.!
The following targets are supported:
htdoc - generates HTML-documentation
ladoc - generates Latex-documentation
psdoc - generates PostScript-documentation
pdfdoc - generates PDF-documentation
doc - generates all supported forms of documentation
clean-doc - generates all supported forms of documentation
All of them generate a sub-directory doc
. More precisely, for HTML it
is doc/$(RESULT)/html
and for Latex, PostScript and PDF the directory
doc/$(RESULT)/latex
. See the OCamldoc-manual for details and the optional
variables below for settings you can control.
You can have several targets in the same directory and manage them from
within an single Makefile
.
Give each subproject a name, e.g. p1
, p2
, etc. Then you export settings
specific to each project by using variables of the form PROJ_p1
, PROJ_p2
,
etc. E.g.:
define PROJ_p1
SOURCES=foo.ml main.ml
RESULT="p1"
OCAMLFLAGS="-unsafe"
endef
export PROJ_p1
define PROJ_p2
...
endef
export PROJ_p2
You may also export common settings used by all projects directly, e.g.:
export THREADS = y
Now is a good time to define which projects should be affected by commands by default. E.g.:
ifndef SUBPROJS
export SUBPROJS = p1 p2
endif
This will automatically generate a given target for all those subprojects
if this variable has not been defined in the shell environment or in the
command line of the make-invocation by the user. E.g., make dc
will
generate debug code for all subprojects.
Now you need to define a default action for your subprojects if make
has been called without arguments:
all: bc
This will build byte code by default for all subprojects.
Finally, you'll have to define a catch-all target that uses the target provided by the user for all subprojects. Just add (assuming that OCAMLMAKEFILE has been defined appropriately):
%:
@make -f
See the threads
-directory in the distribution for a short example!
* LIB_PACK_NAME - packs all modules of a library into a module whose
name is given in variable LIB_PACK_NAME.
* RES_CLIB_SUF - when building a library that contains C-stubs, this
variable controls the suffix appended to the name of
the C-library (default: _stubs).
* THREADS - say THREADS = yes if you need thread support compiled in,
otherwise leave it away.
* VMTHREADS - say VMTHREADS = yes if you want to force VM-level
scheduling of threads (byte-code only).
* ANNOTATE - say ANNOTATE = yes to generate type annotation files
(.annot) to support displaying of type information
in editors.
* USE_CAMLP4 - say USE_CAMLP4 = yes in your Makefile if you
want to include the camlp4 directory during the build
process, otherwise leave it away.
* INCDIRS - directories that should be searched for .cmi- and
.cmo-files. You need not write -I ... - just the
plain names.
* LIBDIRS - directories that should be searched for libraries
Also just put the plain paths into this variable
* EXTLIBDIRS - Same as LIBDIRS, but paths in this variable are
also added to the binary via the -R-flag so that
dynamic libraries in non-standard places can be found.
* RESULTDEPS - Targets on which results (executables or libraries)
should additionally depend.
* PACKS - adds packages under control of findlib.
* PREDS - specifies findlib-predicates.
* LIBS - OCaml-libraries that should be linked (just plain names).
E.g. if you want to link the Str-library, just write
str (without quotes). The new OCaml-compiler handles
libraries in such a way that they "remember" whether
they have to be linked against a C-library and it gets
linked in automatically. If there is a slash in the
library name (such as ./str or lib/foo) then make is
told that the generated files depend on the library.
This helps to ensure that changes to your libraries
are taken into account, which is important if you are
regenerating your libraries frequently.
* CLIBS - C-libraries that should be linked (just plain names).
* PRE_TARGETS - set this to a list of target files that you want
to have built before dependency calculation actually
takes place. E.g. use this to automatically compile
modules needed by camlp4, which have to be available
before other modules can be parsed at all.
** WARNING **: the files mentioned in this variable
will be removed when make clean is executed!
* LIBINSTALL_FILES - the files of a library that should be installed
using findlib. Default:
$(RESULT).mli $(RESULT).cmi $(RESULT).cma
$(RESULT).cmxa $(RESULT).a lib$(RESULT).a
* OCAML_LIB_INSTALL - target directory for rawinstall/rawuninstall.
(default: $(OCAMLLIBPATH)/contrib)
* DOC_FILES - names of files from which documentation is generated.
(default: all .mli-files in your $(SOURCES)).
* DOC_DIR - name of directory where documentation should be stored.
* OCAMLFLAGS - flags passed to the compilers
* OCAMLBCFLAGS - flags passed to the byte code compiler only
* OCAMLNCFLAGS - flags passed to the native code compiler only
* OCAMLLDFLAGS - flags passed to the OCaml-linker
* OCAMLBLDFLAGS - flags passed to the OCaml-linker when linking byte code
* OCAMLNLDFLAGS - flags passed to the OCaml-linker when linking
native code
* OCAMLMKLIB_FLAGS - flags passed to the OCaml library tool
* OCAMLCPFLAGS - profiling flags passed to ocamlcp (default: a)
* PPFLAGS - additional flags passed to the preprocessor
(default: none)
* LFLAGS - flags passed to ocamllex
* YFLAGS - flags passed to ocamlyacc
* IDLFLAGS - flags passed to camlidl
* OCAMLDOCFLAGS - flags passed to ocamldoc
* OCAMLFIND_INSTFLAGS - flags passed to ocamlfind during installation
(default: none)
* DVIPSFLAGS - flags passed to dvips
(when generating documentation in PostScript).
* STATIC - set this variable if you want to force creation
of static libraries
* CC - the C-compiler to be used
* CXX - the C++-compiler to be used
* CFLAGS - additional flags passed to the C-compiler.
The flag -DNATIVE_CODE will be passed automatically if
you choose to build native code. This allows you to
compile your C-files conditionally. But please note:
You should do a make clean or remove the object files
manually or touch the %.c-files: otherwise, they may
not be correctly recompiled between different builds.
* CXXFLAGS - additional flags passed to the C++-compiler.
* CPPFLAGS - additional flags passed to the C-preprocessor.
* CFRAMEWORKS - Objective-C framework to pass to linker on MacOS X.
* LDFLAGS - additional flags passed to the C-linker
* RPATH_FLAG - flag passed through to the C-linker to set a path for
dynamic libraries. May need to be set by user on
exotic platforms. (default: -R).
* ELF_RPATH_FLAG - this flag is used to set the rpath on ELF-platforms.
(default: -R)
* ELF_RPATH - if this flag is yes, then the RPATH_FLAG will be
passed by -Wl to the linker as normal on ELF-platforms.
* OCAMLLIBPATH - path to the OCaml-standard-libraries
(first default: $(OCAMLC) -where)
(second default: /usr/local/lib/ocaml)
* OCAML_DEFAULT_DIRS - additional path in which the user can supply
default directories to his own collection
of libraries. The idea is to pass this as an
environment variable so that the Makefiles do not
have to contain this path all the time.
* OCAMLFIND - ocamlfind from findlib (default: ocamlfind)
* OCAML - OCaml interpreter (default: ocaml)
* OCAMLC - byte-code compiler (default: ocamlc)
* OCAMLOPT - native-code compiler (default: ocamlopt)
* OCAMLMKTOP - top-level compiler (default: ocamlmktop)
* OCAMLCP - profiling byte-code compiler (default: ocamlcp)
* OCAMLDEP - dependency generator (default: ocamldep)
* OCAMLLEX - scanner generator (default: ocamllex)
Applies to .mll files.
* OCAMLYACC - parser generator (default: ocamlyacc)
Applies to .mly files. A good alternative to the default is
"menhir" if installed.
* OCAMLMKLIB - tool to create libraries (default: ocamlmklib)
* CAMLIDL - IDL-code generator (default: camlidl)
* CAMLIDLDLL - IDL-utility (default: camlidldll)
* CAMLP4 - camlp4 preprocessor (default: camlp4)
* OCAMLDOC - OCamldoc-command (default: ocamldoc)
* LATEX - Latex-processor (default: latex)
* DVIPS - dvips-command (default: dvips)
* PS2PDF - PostScript-to-PDF converter (default: ps2pdf)
* CAMELEON_REPORT - report tool of Cameleon (default: report)
* CAMELEON_REPORT_FLAGS - flags for the report tool of Cameleon
* CAMELEON_ZOGGY - zoggy tool of Cameleon
(default: camlp4o pa_zog.cma pr_o.cmo)
* CAMELEON_ZOGGY_FLAGS - flags for the zoggy tool of Cameleon
* OCAML_GLADECC - Glade compiler for OCaml (default: lablgladecc2)
* OCAML_GLADECC_FLAGS - flags for the Glade compiler
* OXRIDL - OXRIDL-generator (default: oxridl)
* NOIDLHEADER - set to yes to prohibit OCamlMakefile from using
the default camlidl-flag -header.
* NO_CUSTOM - Prevent linking in custom mode.
* QUIET - unsetting this variable (e.g. make QUIET=)
will print all executed commands, including intermediate
ones. This allows more comfortable debugging when
things go wrong during a build.
* REALLY_QUIET - when set this flag turns off output from some commands.
* OCAMLMAKEFILE - location of (= path to) this OCamlMakefile.
Because it calls itself recursively, it has to know
where it is. (default: OCamlMakefile = local directory)
* BCSUFFIX - Suffix for all byte-code files. E.g.:
RESULT = foo
BCSUFFIX = _bc
This will produce byte-code executables/libraries with
basename foo_bc.
* NCSUFFIX - Similar to BCSUFFIX, but for native-code files.
* TOPSUFFIX - Suffix added to toplevel interpreters (default: .top)
* SUBPROJS - variable containing the names of subprojects to be
compiled.
* SUBTARGET - target to be built for all projects in variable
SUBPROJS.
* MINGW - variable to detect the MINGW-environment
* MSVC - variable to detect the MSVC-compiler
Please submit bugs reports, feature requests, contributions and similar to the GitHub issue tracker.
Up-to-date information is available at: https://mmottl.github.io/ocaml-makefile