README ------ This document provides a description of the MLT project organisation. It provides *CRITICAL* architecture information, so please read carefully if you plan to extend or use the MLT code base. Directories ----------- The directory heirarchy is defined as follows: + docs - Location of all documentation + src - All project source is provided here + framework - The mlt media framework + modules - All services are defined here + core - Independent MLT services + dv - libdv dependent services + ffmpeg - ffmpeg dependent services + avformat - libavformat dependent services + vorbis - vorbis dependenent services + sdl - SDL dependent services + resample - libresample dependent services + gtk2 - pango and pixbuf dependent services + xine - xine-derived services + bluefish - Bluefish dependent services (*) + mainconcept - mainconcept dependent services (*) + inigo - A media playing test application + valerie - Client API to access the server + miracle - The server implementation + humperdink - The evil demo + albino - The simple but nice demo + tests - Reserved for regression and unit tests Additional subdirectories may be nested below those shown and should be documented in their parent. (*) Not posted to CVS due to licensing issues. Configuration ------------- Configuration is triggered from the top level directory via a ./configure script. Each source bearing subdirectory shown above have their own configure script which are called automatically from the top level. Typically, new modules can be introduced without modification to the configure script and arguments are accepted and passed through to all subdirectories. Top level usage is: ./configure --help - report all configure options ./configure --prefix=[dir] - target install directory (default: /usr/local) ./configure --disable-[mod] - do not compile specified module(s) ./configure --[other] - pass through to children NB: This script must be run to register new services after a CVS checkout or subsequent update. Compilation ----------- Makefiles are generated during configuration and these are based on a per directory template which must be provided by the developer. Testing ------- To execute the mlt tools without installation, or to test a new version on a system with an already installed mlt version, you should run: . setenv NB: This applies to your current shell only. Installation ------------ * NOT IMPLEMENTED YET * The install is triggered by running make install or make install-strip from the top level directory. The framework produces a single shared object which is installed in $prefix/lib/ and public header files which are installed in $prefix/include/mlt/framework. The client produces a single shared object which is installed in $prefix/lib/ and public header which are installed in $prefix/include/mlt/client. The server produces a single exectuable which is installed in $prefix/bin/. This is linked against the framework shared object and posix libs but not against any of the modules. The modules produce a shared object per module and update text files containing a list of modules provided by this build. These are installed in $prefix/share/mlt/. It is at the discretion of the module to install additional support files. To allow the development of external components, mlt-client-config and mlt-framework-config scripts are generated and installed in $prefix/bin. After install, only those modules listed are usable by the server. No module is loaded unless explicitly requested via server configuration or usage. External modules are also placed in this $prefix/share/mlt, and the installation of those must modify the text file accordingly before they will be considered at runtime. Development ----------- All compilation in the project has {top-level-dir}/src on the include path. All headers are included as: #include All external modules have {prefix}/include/mlt on the include path. All headers should also be included as: #include This allows migration of source between external and internal modules. The configuration and Makefile template requirements will require attention though. Summary ------- 1. The server will interact with public interfaces from the framework only; 2. The modules must expose public framework interfaces only; 3. All modules are dynamically loaded at runtime.