PierrePronchery These are the documentation notes for configure. The configure project is a Makefiles generator. It uses project description files to propose simple Makefiles. It is intended to keep generated code as simple as possible.

configure generates Makefiles needed in an entire project. Instead of trying every little trick to let it work in a number of potential uses, it sticks to the simplest possible code generation. Consequently, it will mostly be useful for projects running on UNIX platforms first.

Every software developer could gain using configure. However, due to its intentional simplicity, it may not be appropriate to use along with code compilation on some non-portable platforms. Of course an advanced user, trying to modify a given project Makefile on his platform, inside a project using configure may have to use configure too.

configure will certainly not be appropriate alone for cross-plaform projects, and maybe not for big projects. However one should be able to keep its configuration files inside a project without creating any trouble for potential other Makefile files generators.

configure has been created to be efficient at writing simple and compliant Makefiles for small to big software development projects. It should be useful: for people learning software development using Makefiles; for developers of some small to big projects; for developers concerned by the readability and efficiency of their Makefiles.

The configure utility may be invoked from the command line this way: $ configure [-nv][directory] The "-n" option just parses the project definition files, without actually re-generating the Makefiles. The "-v" option gives information about the progress of the operation. It then processes the current directory, or one given at the command line, according to the project configuration files.

This file should be found in every project directory. It must be named "project.conf". It is organized in sections, and string variables. Section names are written on their own line, between brackets (eg "[section]"). Variables are given on their own line too, like this: variable=value Configuration files may be commented, comment lines being prepended with a hash sign "#". The significant variables are the following: in the default section (has an empty name, eg "[]" in the file) * subdirs: subdirectories to look for too * cflags_force: CFLAGS to force globally * cflags: optional global CFLAGS * ldflags_force: LDFLAGS to force globally * ldflags: optional global LDFLAGS * targets: targets to handle in the Makefile * dist: additional files to include in a source archive - in sections named like the target they define: * type (mandatory): type of the target (eg "binary", "library", "object", ...) * cflags: additional CFLAGS for this target * ldflags: additional LDFLAGS for this target * sources: source files to compile

The following target types are currently supported: "binary": produces binary files, linked from every object file produced with their source files. "library": produces a static and a shared version of the target, linked from every object file produced with their source files, and respectively appending ".a" and ".so" extensions to the target name. "object": produces a binary object file from the given source.

You may first create all necessary "project.conf" files with the subdirectories definitions. Then for every binary or library built, specify the adequate target along with its section. When migrating from automake/autoconf, the existing subdirectories are defined in the "Makefile.am" files, in the "SUBDIRS" variable. The binary targets are defined in the same file, as the "bin_PROGRAMS" variable, each declined to "program_SOURCES" for their respective source files.

configure is not user-proof when necessary development files are not available (relies on appropriate third-party programs error messages, eg "project-config" or "pkgconfig")

run-time system detection or selection; generation of the Makefiles altered acordingly; tweakable installation path.