HandBrake 0.9.3 Build User Guide ******************************** Table of Contents ***************** 1 Introduction 2 Quickstart 3 Build Process 3.1 Checkout Sources 3.2 Configure 3.3 Build 3.4 Clean 3.5 Extra Clean 4 Make Targets 4.1 Global 4.2 General Modules 4.3 Contrib Modules 4.4 Contrib Aggregate 5 Platform Requirements and Notes 5.1 Mac OS X 5.2 Cygwin 5.3 Linux 1 Introduction ************** This guide documents the recommended process to build HandBrake from the official source-code repository. Building from any other source is not unsupported. 2 Quickstart ************ This chapter is for the impatient or those just looking for a quick summary of the commands used to launch a typical build with the fewest commands possible. For more control over the build process please skip this section and jump to *Note Build Process:: for full details. svn checkout svn://svn.handbrake.fr/HandBrake/trunk hb-trunk cd hb-trunk ./configure --launch The above is an special streamlined invocation of `configure' which performs the following steps automatically: * assert scratch directory `build/' does not exist * create scratch directory `build/' * change to directory `build/' * launch `make' * capture build output to `build/log.txt' * echo build output 3 Build Process *************** 3.1 Checkout Sources ==================== Checkout HandBrake from the official source-code repository. svn checkout svn://svn.handbrake.fr/HandBrake/trunk hb-trunk cd hb-trunk Sources are checked out from the `trunk' branch. This document was generated from that very branch, and for example purposes, we will use exactly the same branch. If you have write-access to the repository, then you may add the appropriate login/password information as needed. It is recommended to use Subversion 1.5.0 or higher. Lower versions should also work. 3.2 Configure ============= Configure the build system. rm -fr build/ mkdir build/ cd build/ ../configure Create a scratch directory which will contain all files created during the build process. The directory name is arbitrary but we recommend something simple and descriptive. One directory is required for each distinctly configured build. We name our directory `build' for example purposes. The `configure' utility accepts many options. It is recommended that you specify `--help' for the complete list of options. The following options are detailed: `--help' List available options. `--debug=MODE' Select debug mode. Must be one of `none', `min', `std', `max'. This generally maps to gcc options `-g0', `-g1', `-g2', `-g3'. `--optimize=MODE' Select optimize mode. Must be one of `none', `speed', `size'. This generally maps to gcc options `-g0', `-O0', `-O3', `-Os'. `--arch=MODE' Select build architecture. The available architectures vary by platform. Most platforms support exactly one architecture except Mac OS X which has support for various universal binary architectures. The available choices are hard-coded per platform and no sanity checks for the required tools are performed. `--gcc=EXE' Specify the `gcc' executable to use where EXE is the executable name which is either absolute or environment `PATH' is searched accordingly. Clean-room procedures dictate that when certain factors change, old builds should be scrapped and new builds configured. This is the main reason for requiring a scratch directory; to promote consistent, reliable and clean software builds. The following is a short list of some of the reasons why someone may choose to scrap an existing build: * configure with different options * subversion working dir is updated and you want configure to re-evaluate working dir metadata. * build corruption is suspected There are generally two methods for scrapping a build. The `build' directory can be recusrively removed which has the effect of loosing your existing configuration but does guarantee no residuals are left behind. The other method is to use ask the build system to perform an `xclean'. This is known to work well but will leave empty directories behind. However, the configuration is left intact. See *Note Extra Clean:: for further details. 3.3 Build ========= Build main product. All necessary dependencies are also built if required. make Parallel builds may optionally be enabled. Be aware that while a parallel build may save time on systems with additional cores, the output is often mixed, overlapped and sometimes even corrupted with binary characters. Thus if you experience a build issue, you should clean and redo the build in default serial mode to produce a readable log. The following command allows for up to 4 concurrent jobs via make: make -j4 3.4 Clean ========= Clean all build output excluding contrib modules. Configuration is retained. make clean 3.5 Extra Clean =============== Clean all build output including contrib modules. Configuration is retained. make xclean 4 Make Targets ************** The build system supports passing many kinds of targets some of which become very useful in normal development cycles. The targets by convention are lower-case words passed to `make'. Global targets are one-word targets. Scoped targets are usually two-words seperated by a period. 4.1 Global ========== `make' Alias for `make build'. `make build' Build main product. All necessary dependencies are also built if required. `make clean' Clean all build output excluding contrib modules. Configuration is retained. `make xclean' Clean all build output including contrib modules. Configuration is retained. 4.2 General Modules =================== General modules such as `libhb' and `test' have the following scoped targets: `make MODULE.build' Build MODULE. `make MODULE.clean' Clean build output for MODULE. 4.3 Contrib Modules =================== Contrib modules such as `a52dec', `bzip2', `faac', `faad2', `ffmpeg', `lame', `libdca', `libdvdread', `libmkv', `libmp4v2', `libogg', `libsamplerate', `libtheora', `libvorbis', `mpeg2dec', `x264', `xvidcore' and `zlib' have the following scoped targets: `make MODULE.fetch' Download source tarball from the Internet and save to `TOP/downloads' directory. No checksumming is performed. `make MODULE.extract' Extract source tarball into `build' tree. `make MODULE.patch' Apply appropriate patches (if any) to module sources. `make MODULE.configure' Configure module sources. This usually invokes autotool configure. `make MODULE.build' Build module. This usually invokes autotool build. `make MODULE.install' Install module products such as headers and libraries into `build' tree. This usually invokes autotool install. `make MODULE.uninstall' Uninstall module products; generally the reverse of install. This usually invokes autotool uninstall. `make MODULE.clean' Clean module; generally the reverse of build. This usually invokes autotool clean. `make MODULE.xclean' Extra clean module; first invokes uninstall then recursively removes the module build directory. 4.4 Contrib Aggregate ===================== For convenience, the following targets aggregate the all contrib modules' respective targets together: * make contrib.fetch * make contrib.extract * make contrib.patch * make contrib.configure * make contrib.build * make contrib.install * make contrib.uninstall * make contrib.clean * make contrib.xclean 5 Platform Requirements and Notes ********************************* The build system supports various platforms of interest to the project. However this does not mean it supports all plaforms. If the platform is not listed in this chapter, then it is not supported. The following tools are used on various platforms and it is recommended you use these versions or newer: * python - Python 2.4.6 * curl - curl 7.19.3 (or wget) * m4 - GNU M4 1.4.6 * make - GNU Make 3.81 * patch - Patch 2.5.8 * tar - GNU tar 1.15.1 * wget - GNU Wget 1.11.4 (or curl) 5.1 Mac OS X ============ Building on Mac OS X is well supported. It is the reference platform for HandBrake. The following are the recommended specifications for this platform; but is not necessarily the only configuration that is possible: * Mac Intel hardware * Mac OS X 10.5.6 * Xcode-3.1.2 * gcc 4.0.1 (Apple Inc. build 5490) * yasm 0.7.2.2153 (for i386 or x86_64 architectures) Note: It is recommended to use the platform distribution's bundled compiler for maximum C++ compatibility. If you build with a custom compiler it will likely introduce non-standard runtime requirements. There are of course many valid reasons to build with unbundled compilers, but be aware it is generally unsupported and left as an exercise to the reader. 5.2 Cygwin ========== Building on Cygwin is supported. The following are the recommended specifications for this platform; but is not necessarily the only configuration that is possible: * Intel 32-bit or 64-bit hardware * Cygwin, gcc 4.3.2 * yasm 0.7.1.2093 (for i386 or x86_64 architectures) Note: As of this writing, Cygwin has available to it several versions of gcc; only one of which may be found and used in the path as `gcc' and `g++'. Configure will thus find what is probably the older version of gcc in a typical Cygwin environment. If you desire to build with the newer gcc, it is found in the path as `gcc-4' and `g++-4' respectively and you must indicate to configure the desired versions. The following syntax should do the trick: ../configure --gcc=gcc-4 5.3 Linux ========= Building on Linux is supported. The following are the recommended specifications for this platform; but is not necessarily the only configuration that is possible: * Intel 32-bit or 64-bit hardware * Fedora 8, gcc 4.1.2, yasm 0.6.2.1985 * Fedora 9, gcc 4.3.0, yasm 0.6.2.1985 * Fedora 10, gcc 4.3.2, yasm 0.7.1.2093 * gcc 4.0.0 or higher is reported to work Note: It is recommended to use the platform distribution's bundled compiler for maximum C++ compatibility. If you build with a custom compiler it will likely introduce non-standard runtime requirements. There are of course many valid reasons to build with unbundled compilers, but be aware it is generally unsupported and left as an exercise to the reader.