Recipes
Introduction A bitbake recipe is a set of instructions that describes what needs to be done to retrieve the source code for some application, apply any necessary patches, provide any additional files (such as init scripts), compile it, install it and generate binary packages. The end result is a binary package that you can install on your target device, and maybe some intermediate files, such as libraries and headers, which can be used when building other applications. In many ways the process is similar to creating .deb or .rpm packages for your standard desktop distributions with one major difference - in OpenEmbedded everything is being cross-compiled. This often makes the task far more difficult (depending on how well suited the application is to cross compiling), than it is for other packaging systems and sometimes impossible. This chapter assumes that you are familiar with working with bitbake, including the work flow, required directory structures and bitbake configuration. If you are not familiar with these then first take a look at the chapter on bitbake usage.
Syntax of recipes The basic items that make up a bitbake recipe file are: functions Functions provide a series of actions to be performed. Functions are usually used to override the default implementation of a task function, or to compliment (append or prepend to an existing function) a default function. Standard functions use sh shell syntax, although access to OpenEmbedded variables and internal methods are also available. The following is an example function from the sed recipe: do_install () { autotools_do_install install -d ${D}${base_bindir} mv ${D}${bindir}/sed ${D}${base_bindir}/sed.${PN} }It is also possible to implement new functions, that are not replacing or complimenting the default functions, which are called between existing tasks. It is also possible to implement functions in python instead of sh. Both of these options are not seen in the majority of recipes. variable assignments and manipulations Variable assignments allow a value to be assigned to a variable. The assignment may be static text or might include the contents of other variables. In addition to assignment, appending and prepending operations are also supported. The following example shows some of the ways variables can be used in recipes:S = "${WORKDIR}/postfix-${PV}" PR = "r4" CFLAGS += "-DNO_ASM" SRC_URI_append = "file://fixup.patch" keywords Only a few keywords are used in bitbake recipes. They are used for things such as including common functions (inherit), loading parts of a recipe from other files (include and require) and exporting variables to the environment (export). The following example shows the use of some of these keywords:export POSTCONF = "${STAGING_BINDIR}/postconf" inherit autoconf require otherfile.inc comments Any lines that begin with a # are treated as comment lines and are ignored.# This is a comment The following is a summary of the most important (and most commonly used) parts of the recipe syntax: Line continuation: \ To split a statement over multiple lines you should place a \ at the end of the line that is to be continued on the next line. VAR = "A really long \ line" Note that there must not be anything (no spaces or tabs) after the \. Using variables: ${...} To access the contents of a variable you need to access it via ${<varname>}:SRC_URI = "${SOURCEFORGE_MIRROR}/libpng/zlib-${PV}.tar.gz" Quote all assignments All variable assignments should be quoted with double quotes. (It may work without them at present, but it will not work in the future).VAR1 = "${OTHERVAR}" VAR2 = "The version is ${PV}" Conditional assignment Conditional assignment is used to assign a value to a variable, but only when the variable is currently unset. This is commonly used to provide a default value for use when no specific definition is provided by the machine or distro configuration of the user's local.conf configuration. The following example:VAR1 ?= "New value"will set VAR1 to "New value" if it is currently empty. However if it was already set it would be unchanged. In the following VAR1 is left with the value "Original value":VAR1 = "Original value" VAR1 ?= "New value" Appending: += You can append values to existing variables using the += operator. Note that this operator will add a space between the existing content of the variable and the new content.SRC_URI += "file://fix-makefile.patch" Prepending: =+ You can prepend values to existing variables using the =+ operator. Note that this operator will add a space between the new content and the existing content of the variable.VAR =+ "Starts" Appending: _append You can append values to existing variables using the _append method. Note that this operator does not add any additional space, and it is applied after all the +=, and =+ operators have been applied. The following example shows the space being explicitly added to the start to ensure the appended value is not merged with the existing value:SRC_URI_append = " file://fix-makefile.patch"The _append method can also be used with overrides, which results in the actions only being performed for the specified target or machine: [TODO: Link to section on overrides]SRC_URI_append_sh4 = " file://fix-makefile.patch"Note that the appended information is a variable itself, and therefore it's possible to use += or =+ to assign variables to the _append information:SRC_URI_append = " file://fix-makefile.patch" SRC_URI_append += "file://fix-install.patch" Prepending: _prepend You can prepend values to existing variables using the _prepend method. Note that this operator does not add any additional space, and it is applied after all the +=, and =+ operators have been applied. The following example shows the space being explicitly added to the end to ensure the prepended value is not merged with the existing value:CFLAGS_prepend = "-I${S}/myincludes "The _prepend method can also be used with overrides, which result in the actions only being performed for the specified target or machine: [TODO: Link to section on overrides]CFLAGS_prepend_sh4 = " file://fix-makefile.patch"Note that the appended information is a variable itself, and therefore it's possible to use += or =+ to assign variables to the _prepend information:CFLAGS_prepend = "-I${S}/myincludes " CFLAGS_prepend += "-I${S}/myincludes2 "Note also the lack of a space when using += to append to a prepend value - remember that the += operator adds space itself. Spaces vs tabs Spaces should be used for indentation, not hard tabs. Both currently work, however it is a policy decision of OE that spaces always be used. Style: oe-stylize.py To help with using the correct style in your recipes there is a python script in the contrib directory called oe-stylize.py which can be used to reformat your recipes to the correct style. The output will contain a list of warnings (to let you know what you did wrong) which should be edited out before using the new file.$ contrib/oe-stylize.py myrecipe.bb > fixed-recipe.bb vi fixed-recipe.bb mv fixed.recipe.bb myrecipe.bb Using python for complex operations: ${@...} For more advanced processing it is possible to use python code during variable assignments, for doing search and replace on a variable for example. Python code is indicated by a proceeding @ sign in the variable assignment.CXXFLAGS := "${@'${CXXFLAGS}'.replace('-frename-registers', '')}"More information about using python is available in the section. Shell syntax When describing a list of actions to take, shell syntax is used (as if you were writing a shell script). You should ensure that your script works with a generic sh and not require any bash (or other shell) specific functionality. The same applies to various system utilities (sed, grep, awk etc) that you may wish to use. If in doubt you should check with multiple implementations - including those from busybox. For a detailed description of the syntax for the bitbake recipe files you should refer to the bitbake user manual.
Recipe naming: Names, versions and releases Recipes in OpenEmbedded use a standard naming convention that includes the package name and version number in the filename. In addition to the name and version there is also a release number, which indicates changes to the way the package is built and/or packaged. The release number is contained within the recipe itself. The expected format of recipe name is:<package-name>_<version>.bb where <package-name> is the name of the package (application, library, module, or whatever it is that is being packaged) and version is the version number. So a typical recipe name would be:strace_4.5.14.bbwhich would be for version 4.5.14 of the strace application. The release version is defined via the package release variable, PR, contained in the recipe. The expected format is:r<n>where <n> is an integer number starting from 0 initially and then incremented each time the recipe, or something that effects the recipe, is modified. So a typical definition of the release would be:PR = "r1"to specify release number 1 (the second release, the first would have been 0). If there is no definition of PR in the recipe then the default value of "r0" is used. It is good practice to always define PR in your recipes, even for the "r0" release, so that when editing the recipe it is clear that the PR number needs to be updated. You should always increment PR when modifying a recipe. Sometimes this can be avoided if the change will have no effect on the actual packages generated by the recipe, such as updating the SRC_URI to point to a new host. If in any doubt then you should increase the PR regardless of what has been changed. The PR value should never be decremented. If you accidentally submit a large PR value for example then it should be left at the value and just increased for new releases, not reset back to a lower version. When a recipe is being processed some variables are automatically set based on the recipe file name and can be used for other purposes from within the recipe itself. These include: PN The package name. Determined from the recipe filename - everything up until the first underscore is considered to be the package name. For the strace_4.5.14.bb recipe the PN variable would be set to "strace". PV The package version. Determined from the recipe filename - everything between the first underscore and the final .bb is considered to be the package version. For the strace_4.5.14.bb recipe the PV variable would be set to "4.5.14". PR The package release. This is explicitly set in the recipe. It defaults to "r0" if not set. P The package name and versions separated by a hyphen.P = "${PN}-${PV}" For the strace_4.5.14.bb recipe the P variable would be set to "strace-4.5.14". PF The package name, version and release separated by hyphens.PF = "${PN}-${PV}-${PR}" For the strace_4.5.14.bb recipe, with PR set to "r1" in the recipe, the PF variable would be set to "strace-4.5.14-r1". While some of these variables are not commonly used in recipes (they are used internally though) both PN and PV are used a lot. In the following example we are instructing the packaging system to include an additional directory in the package. We use PN to refer to the name of the package rather than spelling out the package name:FILES_${PN} += "${sysconfdir}/myconf" In the next example we are specifying the URL for the package source, by using PV in place of the actual version number it is possible to duplicate, or rename, the recipe for a new version without having to edit the URL:SRC_URI = "ftp://ftp.vim.org/pub/vim/unix/vim-${PV}.tar.bz2"
Variables One of the most confusing parts of bitbake recipes for new users is the large amount of variables that appear to be available to change and/or control the behaviour of some aspect of the recipe. Some variables, such as those derived from the file name are reasonably obvious, others are not at all obvious. There are several places where these variables are derived from and/or used: A large number of variables are defined in the bitbake configuration file conf/bitbake.conf - it's often a good idea to look through that file when trying to determine what a particular variable means. Machine and distribution configuration files in conf/machine and conf/distro will sometimes define some variables specific to the machine and/or distribution. You should look at the appropriate files for your targets to see if anything is being defined that effects the recipes you are building. Bitbake itself will define some variables. The FILE variable that defines the name of the bitbake recipe being processed is set by bitbake itself for example. Refer to the bitbake manual for more information on the variables that bitbake sets. The classes that are used via the inherit keyword define and/or use the majority of the remaining variables. A class is like a library that contains parts of a bitbake recipe that is used by multiple recipes. To make them usable in more situations they often include a large number of variables to control how the class operates. Another important aspect is that there are three different types of things that binaries and libraries are used for and they often have different variables for each. These include: target Refers to things built for the target and are expected to be run on the target device itself. native Refers to things built to run natively on the build host itself. cross Refers to things built to run natively on the build host itself, but produce output which is suitable for the target device. Cross versions of packages usually only exist for things like compilers and assemblers - i.e. things which are used to produce binary applications themselves.
Header Practically all recipes start with a header section which describes various aspects of the package that is being built. This information is typically used directly by the package format (such as ipkg or deb) as meta data used to describe the package. Variables used in the header include: DESCRIPTION Describes what the software does. Hopefully this gives enough information to a user to know if it's the right application for them. The default description is: "Version ${PV}-${PR} of package ${PN}". HOMEPAGE The URL of the home page of the application where new releases and more information can be found. The default homepage is "unknown". SECTION The section is used to categorize the application into a specific group. Often used by GUI based installers to help users when searching for software. See for a list of the available sections. The default section is "base". PRIORITY The default priority is "optional". LICENSE The license for the application. If it is not one of the standard licenses then the license itself must be included (where?). As well as being used in the package meta-data the license is also used by the src_distribute class. The default license is "unknown".
Sources: Downloading, patching and additional files A recipe's purpose is to describe how to take a software package and build it for your target device. The location of the source file (or files) is specified via the in the recipe. This can describe several types of URIs, the most common are: http and https Specifies files to be downloaded. A copy is stored locally so that future builds will not download the source again. cvs, svn and git Specifies that the files are to be retrieved using the specified version control system. files Plain files which are included locally. These can be used for adding documentation, init scripts or any other files that need to be added to build the package under openembedded. patches Plain files which are treated as patches and automatically applied. If an http, https or file URI refers to a compressed file, an archive file or a compressed archive file, such as .tar.gz or .zip, then the files will be uncompressed and extracted from the archive automatically. Archive files will be extracted from within the working directory, ${WORKDIR} and plain files will be copied into the same directory. Patches will be applied from within the unpacked source directory, ${S}. (Details on these directories is provided in the next section.) The following example from the havp recipe shows a typical SRC_URI definition:SRC_URI = "http://www.server-side.de/download/havp-${PV}.tar.gz \ file://sysconfdir-is-etc.patch \ file://havp.init \ file://doc.configure.txt \ file://volatiles.05_havp" This describes several files http://www.server-side.de/download/havp-${PV}.tar.gz This is the URI of the havp source code. Note the use of the ${PV} variable to specify the version. This is done to enable the recipe to be renamed for a new version without the need to edit the recipe itself. Because this is a .tar.gz compressed archive the file will be decompressed and extracted in the working dir ${WORKDIR}. file://sysconfdir-is-etc.patch This is a local file that is used to patch the extracted source code. If a filename ends in .patch or .diff then this is treated as patch file and applied with striplevel 1. The patch will be applied from the unpacked source directory, ${S}. In this case ${S} will be ${WORKDIR}/havp-0.82, and luckily the havp-0.82.tar.gz file extracts itself into that directory (so no need to explicitly change ${S}). file://havp.init file://doc.configure.txt file://volatiles.05_havp" These are plain files which are just copied into the working directory ${WORKDIR}. These are then used during the install task in the recipe to provide init scripts, documentation and volatiles configuration information for the package. Full details on the SRC_URI variable and all the support URIs are available in the section of the reference chapter.
Directories: What goes where A large part of the work of a recipe is involved with specifying where files are found and where they have to go. It's important for example that programs do not try and use files from /usr/include or /usr/lib since they are for the host system, not the target. Similarly you don't want programs installed into /usr/bin since that may overwrite your host system programs with versions that don't work on the host! The following are some of the directories commonly referred to in recipes and will be described in more detail in the rest of this section: Working directory: WORKDIR This working directory for a recipe is where archive files will be extracted, plain files will be placed, subdirectories for logs, installed files etc will be created. Unpacked source code directory: S This is where patches are applied and where the program is expected to be compiled. Destination directory: D The destination directory. This is where your package should be installed into. The packaging system will then take the files from directories under here and package them up for installation on the target. Installation directories: bindir, docdir, ... There are a set of variables available to describe all of the paths on the target that you may want to use. Recipes should use these variables rather than hard coding any specific paths. Staging directories: STAGING_LIBDIR, STAGING_INCDIR, ... Staging directories are a special area for headers, libraries and other files that are generated by one recipe that may be needed by another recipe. A library package for example needs to make the library and headers available to other recipes so that they can link against them. File path directories: FILE, FILE_DIRNAME, FILESDIR, FILESPATH These directories are used to control where files are found. Understanding these can help you separate patches for different versions or releases of your recipes and/or use the same patch over multiple versions etc.
WORKDIR: The working directory The working directory is where the source code is extracted, plain files (not patches) are copied and where the logs and installation files are created. A typical reason for needing to reference the work directory is for the handling of non patch files. If we take a look at the recipe for quagga we can see example non patch files for configuration and init scripts:SRC_URI = "http://www.quagga.net/download/quagga-${PV}.tar.gz \ file://fix-for-lib-inpath.patch \ file://quagga.init \ file://quagga.default \ file://watchquagga.init \ file://watchquagga.default"The recipe has two init files and two configuration files, which are not patches, but are actually files that it wants to include in the generated packages. Bitbake will copy these files into the working directory. So to access them during the install task we refer to them via the WORKDIR variable:do_install () { # Install init script and default settings install -m 0755 -d ${D}${sysconfdir}/default ${D}${sysconfdir}/init.d ${D}${sysconfdir}/quagga install -m 0644 ${WORKDIR}/quagga.default ${D}${sysconfdir}/default/quagga install -m 0644 ${WORKDIR}/watchquagga.default ${D}${sysconfdir}/default/watchquagga install -m 0755 ${WORKDIR}/quagga.init ${D}${sysconfdir}/init.d/quagga install -m 0755 ${WORKDIR}/watchquagga.init ${D}${sysconfdir}/init.d/watchquagga ...
S: The unpacked source code directory Bitbake expects to find the extracted source for a package in a directory called <packagename>-<version> in the WORKDIR directory. This is the directory it will change into before patching, compiling and installing the package. For example, we have a package called widgets_1.2.bb which we are extracting from the widgets-1.2.tar.gz file. Bitbake expects the source to end up in a directory called widgets-1.2 within the work directory. If the source does not end up in this directory then bitbake needs to be told this by explicitly setting S. If widgets-1.2.tar.gz actually extracts into a directory called widgets, without the version number, instead of widgets-1.2 then the S variable will be wrong and patching and/or compiling will fail. Therefore we need to override the default value of S to specify the directory the source was actually extracted into:SRC_URI = "http://www.example.com/software/widgets-${PN}.tar.gz" S = "${WORKDIR}/widgets"
D: The destination directory The destination directory is where the completed application and all of its files are installed into in preparation for packaging. Typically an installation would place files in directories such as /etc and /usr/bin by default. Since those directories are used by the host system we do not want the packages to install into those locations. Instead they need to install into the directories below the destination directory. So instead of installing into /usr/bin the package needs to install into ${D}/usr/bin. The following example from arpwatch shows the make install command being passed a ${D} as the DESTDIR variable to control where the makefile installs everything:do_install() { ... oe_runmake install DESTDIR=${D} The following example from quagga shows the use of the destination directory to install the configuration files and init scripts for the package:do_install () { # Install init script and default settings install -m 0755 -d ${D}${sysconfdir}/default ${D}${sysconfdir}/init.d ${D}${sysconfdir}/quagga install -m 0644 ${WORKDIR}/quagga.default ${D}${sysconfdir}/default/quagga install -m 0755 ${WORKDIR}/quagga.init ${D}${sysconfdir}/init.d/quagga You should not use directories such as /etc and /usr/bin directly in your recipes. You should use the variables that define these locations. The full list of these variables can be found in the section of the reference chapter.
Staging directories Staging is used to make libraries, headers and binaries available from the build of one recipe for use by another recipe. Building a library for example requires that packages be created containing the libraries and headers for development on the target as well as making them available on the host for building other packages that need the libraries and headers. Making the libraries, headers and binaries available for use by other recipes on the host is called staging and is performed by the stage task in the recipe. Any recipes that contain items that are required to build other packages should have a stage task to make sure the items are all correctly placed into the staging area. The following example from clamav shows the clamav library and header being placed into the staging area:do_stage () { oe_libinstall -a -so libclamav ${STAGING_LIBDIR} install -m 0644 libclamav/clamav.h ${STAGING_INCDIR} } The following from the p3scan recipe shows the path to the clamav library and header being passed to the configure script. Without this the configure script would either fail to find the library, or worse still search the host system's directories for the library. Passing in the location results in it searching the correct location and finding the clamav library and headers:EXTRA_OECONF = "--with-clamav=${STAGING_LIBDIR}/.. \ --with-openssl=${STAGING_LIBDIR}/.. \ --disable-ripmime"While the staging directories are automatically added by OpenEmbedded to the compiler and linking commands it is sometimes necessary, as in the p3scan example above, to explicitly specify the location of the staging directories. Typically this is needed for autoconf scripts that search in multiple places for the libraries and headers. Many of the helper classes, such as pkgconfig and autotools add appropriate commands to the stage task for you. Check with the individual class descriptions in the reference section to determine what each class is staging automatically for you. A full list of staging directories can be found in the section in the reference chapter.
FILESPATH/FILESDIR: Finding local files The file related variables are used by bitbake to determine where to look for patches and local files. Typically you will not need to modify these, but it is useful to be aware of the default values. In particular when searching for patches and/or files (file:// URIs), the default search path is: ${FILE_DIRNAME}/${PF} This is the package name, version and release, such as "strace-4.5.14-r1". This is very rarely used since the patches would only be found for the one exact release of the recipe. ${FILE_DIRNAME}/${P} This is the package name and version, such as "strace-4.5.14". This is by far the most common place to place version specific patches. ${FILE_DIRNAME}/${PN} This is the package name only, such as "strace". This is not commonly used. ${FILE_DIRNAME}/files This is just the directory called "files". This is commonly used for patches and files that apply to all versions of the package. ${FILE_DIRNAME}/ This is just the base directory of the recipe. This is very rarely used since it would just clutter the main directory. Each of the paths is relative to ${FILE_DIRNAME} which is the directory in which the recipe that is being processed is located. The full set of variables that control the file locations are: FILE The path to the .bb file which is currently being processed. FILE_DIRNAME The path to the directory which contains the FILE which is currently being processed.FILE_DIRNAME = "${@os.path.dirname(bb.data.getVar('FILE', d))}" FILESPATH The default set of directories which are available to use for the file:// URIs. Each directory is searched, in the specified order, in an attempt to find the file specified by each file:// URI: FILESPATH = "${FILE_DIRNAME}/${PF}:${FILE_DIRNAME}/${P}:\ ${FILE_DIRNAME}/${PN}:${FILE_DIRNAME}/files:${FILE_DIRNAME}" FILESDIR The default directory to search for file:// URIs. Only used if the file is not found in FILESPATH. This can be used to easily add one additional directory to the search path without having to modify the default FILESPATH setting. By default this is just the first directory from FILESPATH. FILESDIR = "${@bb.which(bb.data.getVar('FILESPATH', d, 1), '.')}" Sometimes recipes will modify the FILESPATH or FILESDIR variables to change the default search path for patches and files. The most common situation in which this is done is when one recipe includes another one in which the default values will be based on the name of the package doing the including, not the included package. Typically the included package will expect the files to be located in a directory based on its own name. As an example the m4-native recipe includes the m4 recipe. This is fine, except that the m4 recipe expects its files and patches to be located in a directory called m4, while the native file name results in them being searched for in m4-native. So the m4-native recipe sets the FILESDIR variable to the value of the actual m4 directory (where m4 itself has its files stored): include m4_${PV}.bb inherit native FILESDIR = "${@os.path.dirname(bb.data.getVar('FILE',d,1))}/m4"
Basic examples By now you should know enough about the bitbake recipes to be able to create a basic recipe. We'll cover a simple single file recipe and then a more advanced example that uses the autotools helper class (to be described later) to build an autoconf based package.
Hello world Now it's time for our first recipe. This is going to be one of the simplest possible recipes: all code is included and there's only one file to compile and one readme file. While this isn't all that common, it's a useful example because it doesn't depend on any of the helper classes which can sometime hide a lot of what is going on. First we'll create the myhelloworld.c file and a readme file. We'll place this in the files subdirectory, which is one of the places that is searched for file:// URIs:$ mkdir recipes/myhelloworld $ mkdir recipes/myhelloworld/files $ cat > recipes/myhelloworld/files/myhelloworld.c #include <stdio.h> int main(int argc, char** argv) { printf("Hello world!\n"); return 0; } ^D $ cat > recipes/myhelloworld/files/README.txt Readme file for myhelloworld. ^D Now we have a directory for our recipe, recipes/myhelloworld, and we've created a files subdirectory in there to store our local files. We've created two local files, the C source code for our helloworld program and a readme file. Now we need to create the bitbake recipe. First we need the header section, which will contain a description of the package and the release number. We'll leave the other header variables out for now:DESCRIPTION = "My hello world program" PR = "r0" Next we need to tell it which files we want to be included in the recipe, which we do via file:// URIs and the SRC_URI variable:SRC_URI = "file://myhelloworld.c \ file://README.txt" Note the use of the \ to continue a file and the use of file:// local URIs, rather than other types such as http://. Now we need to provide a compile task which tells bitbake how to compile this program. We do this by defining a do_compile function in the recipe and providing the appropriate commands: do_compile() { ${CC} ${CFLAGS} ${LDFLAGS} ${WORKDIR}/myhelloworld.c -o myhelloworld } Note the: use of the pre-defined compiler variables, ${CC}, ${CFLAGS} and ${LDFLAGS}. These are set up automatically to contain the settings required to cross-compile the program for the target. use of ${WORKDIR} to find the source file. As mentioned previously all files are copied into the working directory and can be referenced via the ${WORKDIR} variable. And finally we want to install the program and readme file into the destination directory so that it'll be packaged up correctly. This is done via the install task, so we need to define a do_install function in the recipe to describe how to install the package:do_install() { install -m 0755 -d ${D}${bindir} ${D}${docdir}/myhelloworld install -m 0755 ${S}/myhelloworld ${D}${bindir} install -m 0644 ${WORKDIR}/README.txt ${D}${docdir}/myhelloworld } Note the: use of the install command to create directories and install the files, not cp. way directories are created before we attempt to install any files into them. The install command takes care of any subdirectories that are missing, so we only need to create the full path to the directory - no need to create the subdirectories. way we install everything into the destination directory via the use of the ${D} variable. way we use variables to refer to the target directories, such as ${bindir} and ${docdir}. use of ${WORKDIR} to get access to the README.txt file, which was provided via a file:// URI. We'll consider this release 0 and version 0.1 of a program called helloworld. So we'll name the recipe myhelloworld_0.1.bb:$ cat > recipes/myhelloworld/myhelloworld_0.1.bb DESCRIPTION = "Hello world program" PR = "r0" SRC_URI = "file://myhelloworld.c \ file://README.txt" do_compile() { ${CC} ${CFLAGS} ${LDFLAGS} ${WORKDIR}/myhelloworld.c -o myhelloworld } do_install() { install -m 0755 -d ${D}${bindir} ${D}${docdir}/myhelloworld install -m 0644 ${S}/myhelloworld ${D}${bindir} install -m 0644 ${WORKDIR}/README.txt ${D}${docdir}/myhelloworld } ^DNow we are ready to build our package, hopefully it'll all work since it's such a simple example:$ bitbake -b recipes/myhelloworld/myhelloworld_0.1.bb NOTE: package myhelloworld-0.1: started NOTE: package myhelloworld-0.1-r0: task do_fetch: started NOTE: package myhelloworld-0.1-r0: task do_fetch: completed NOTE: package myhelloworld-0.1-r0: task do_unpack: started NOTE: Unpacking /home/lenehan/devel/oe/local-recipes/myhelloworld/files/helloworld.c to /home/lenehan/devel/oe/build/titan-glibc-25/tmp/work/myhelloworld-0.1-r0/ NOTE: Unpacking /home/lenehan/devel/oe/local-recipes/myhelloworld/files/README.txt to /home/lenehan/devel/oe/build/titan-glibc-25/tmp/work/myhelloworld-0.1-r0/ NOTE: package myhelloworld-0.1-r0: task do_unpack: completed NOTE: package myhelloworld-0.1-r0: task do_patch: started NOTE: package myhelloworld-0.1-r0: task do_patch: completed NOTE: package myhelloworld-0.1-r0: task do_configure: started NOTE: package myhelloworld-0.1-r0: task do_configure: completed NOTE: package myhelloworld-0.1-r0: task do_compile: started NOTE: package myhelloworld-0.1-r0: task do_compile: completed NOTE: package myhelloworld-0.1-r0: task do_install: started NOTE: package myhelloworld-0.1-r0: task do_install: completed NOTE: package myhelloworld-0.1-r0: task do_package: started NOTE: package myhelloworld-0.1-r0: task do_package: completed NOTE: package myhelloworld-0.1-r0: task do_package_write: started NOTE: Not creating empty archive for myhelloworld-dbg-0.1-r0 Packaged contents of myhelloworld into /home/lenehan/devel/oe/build/titan-glibc-25/tmp/deploy/ipk/sh4/myhelloworld_0.1-r0_sh4.ipk Packaged contents of myhelloworld-doc into /home/lenehan/devel/oe/build/titan-glibc-25/tmp/deploy/ipk/sh4/myhelloworld-doc_0.1-r0_sh4.ipk NOTE: Not creating empty archive for myhelloworld-dev-0.1-r0 NOTE: Not creating empty archive for myhelloworld-locale-0.1-r0 NOTE: package myhelloworld-0.1-r0: task do_package_write: completed NOTE: package myhelloworld-0.1-r0: task do_populate_sysroot: started NOTE: package myhelloworld-0.1-r0: task do_populate_sysroot: completed NOTE: package myhelloworld-0.1-r0: task do_build: started NOTE: package myhelloworld-0.1-r0: task do_build: completed NOTE: package myhelloworld-0.1: completed Build statistics: Attempted builds: 1 $ The package was successfully built, the output consists of two .ipkg files, which are ready to be installed on the target. One contains the binary and the other contains the readme file:$ ls -l tmp/deploy/ipk/*/myhelloworld* -rw-r--r-- 1 lenehan lenehan 3040 Jan 12 14:46 tmp/deploy/ipk/sh4/myhelloworld_0.1-r0_sh4.ipk -rw-r--r-- 1 lenehan lenehan 768 Jan 12 14:46 tmp/deploy/ipk/sh4/myhelloworld-doc_0.1-r0_sh4.ipk $ It's worthwhile looking at the working directory to see where various files ended up:$ find tmp/work/myhelloworld-0.1-r0 tmp/work/myhelloworld-0.1-r0 tmp/work/myhelloworld-0.1-r0/myhelloworld-0.1 tmp/work/myhelloworld-0.1-r0/myhelloworld-0.1/patches tmp/work/myhelloworld-0.1-r0/myhelloworld-0.1/myhelloworld tmp/work/myhelloworld-0.1-r0/temp tmp/work/myhelloworld-0.1-r0/temp/run.do_configure.21840 tmp/work/myhelloworld-0.1-r0/temp/log.do_stage.21840 tmp/work/myhelloworld-0.1-r0/temp/log.do_install.21840 tmp/work/myhelloworld-0.1-r0/temp/log.do_compile.21840 tmp/work/myhelloworld-0.1-r0/temp/run.do_stage.21840 tmp/work/myhelloworld-0.1-r0/temp/log.do_configure.21840 tmp/work/myhelloworld-0.1-r0/temp/run.do_install.21840 tmp/work/myhelloworld-0.1-r0/temp/run.do_compile.21840 tmp/work/myhelloworld-0.1-r0/install tmp/work/myhelloworld-0.1-r0/install/myhelloworld-locale tmp/work/myhelloworld-0.1-r0/install/myhelloworld-dbg tmp/work/myhelloworld-0.1-r0/install/myhelloworld-dev tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc/usr tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc/usr/share tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc/usr/share/doc tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc/usr/share/doc/myhelloworld tmp/work/myhelloworld-0.1-r0/install/myhelloworld-doc/usr/share/doc/myhelloworld/README.txt tmp/work/myhelloworld-0.1-r0/install/myhelloworld tmp/work/myhelloworld-0.1-r0/install/myhelloworld/usr tmp/work/myhelloworld-0.1-r0/install/myhelloworld/usr/bin tmp/work/myhelloworld-0.1-r0/install/myhelloworld/usr/bin/myhelloworld tmp/work/myhelloworld-0.1-r0/image tmp/work/myhelloworld-0.1-r0/image/usr tmp/work/myhelloworld-0.1-r0/image/usr/bin tmp/work/myhelloworld-0.1-r0/image/usr/share tmp/work/myhelloworld-0.1-r0/image/usr/share/doc tmp/work/myhelloworld-0.1-r0/image/usr/share/doc/myhelloworld tmp/work/myhelloworld-0.1-r0/myhelloworld.c tmp/work/myhelloworld-0.1-r0/README.txt $Things to note here are: The two source files are in tmp/work/myhelloworld-0.1-r0, which is the working directory as specified via the ${WORKDIR} variable; There's logs of the various tasks in tmp/work/myhelloworld-0.1-r0/temp which you can look at for more details on what was done in each task; There's an image directory at tmp/work/myhelloworld-0.1-r0/image which contains just the directories that were to be packaged up. This is actually the destination directory, as specified via the ${D} variable. The two files that we installed were originally in here, but during packaging they were moved into the install area into a subdirectory specific to the package that was being created (remember we have a main package and a -doc package being created. The program was actually compiled in the tmp/work/myhelloworld-0.1-r0/myhelloworld-0.1 directory, this is the source directory as specified via the ${S} variable. There's an install directory at tmp/work/myhelloworld-0.1-r0/install which contains the packages that were being generated and the files that go in the package. So we can see that the myhelloworld-doc package contains the single file /usr/share/doc/myhelloworld/README.txt, the myhelloworld package contains the single file /usr/bin/myhelloworld and the -dev, -dbg and -local packages are all empty. At this stage it's good to verify that we really did produce a binary for the target and not for our host system. We can check that with the file command:$ file tmp/work/myhelloworld-0.1-r0/install/myhelloworld/usr/bin/myhelloworld tmp/work/myhelloworld-0.1-r0/install/myhelloworld/usr/bin/myhelloworld: ELF 32-bit LSB executable, Hitachi SH, version 1 (SYSV), for GNU/Linux 2.4.0, dynamically linked (uses shared libs), for GNU/Linux 2.4.0, not stripped $ file /bin/ls /bin/ls: ELF 64-bit LSB executable, AMD x86-64, version 1 (SYSV), for GNU/Linux 2.4.0, dynamically linked (uses shared libs), for GNU/Linux 2.4.0, stripped $This shows us that the helloworld program is for an SH processor (obviously this will change depending on what your target system is), while checking the /bin/ls program on the host shows us that the host system is an AMD X86-64 system. That's exactly what we wanted.
An autotools package Now for an example of a package that uses autotools. These are programs that you need to run a configure script for, passing various parameters, and then make. To make these work when cross-compiling you need to provides a lot of variables to the configure script. But all the hard work has already been done for you. There's an which takes care of most of the complexity of building an autotools based package. Let's take a look at the tuxnes recipe which is an example of a very simple autotools based recipe:$ cat recipes/tuxnes/tuxnes_0.75.bb DESCRIPTION = "Tuxnes Nintendo (8bit) Emulator" HOMEPAGE = "http://prdownloads.sourceforge.net/tuxnes/tuxnes-0.75.tar.gz" LICENSE = "GPLv2" SECTION = "x/games" PRIORITY = "optional" PR = "r1" SRC_URI = "http://heanet.dl.sourceforge.net/sourceforge/tuxnes/tuxnes-0.75.tar.gz" inherit autotools This is a really simple recipe. There's the standard header that describes the package. Then the SRC_URI, which in this case is an http URL that causes the source code to be downloaded from the specified URI. And finally there's an "inherit autotools" command which loads the autotools class. The autotools class will take care of generating the required configure, compile and install tasks. So in this case there's nothing else to do - that's all there is to it. It would be nice if it was always this simple. Unfortunately there's usually a lot more involved for various reasons including the need to: Pass parameters to configure to enable and disable features; Pass parameters to configure to specify where to find libraries and headers; Make modifications to prevent searching for headers and libraries in the normal locations (since they belong to the host system, not the target); Make modifications to prevent the configure script from trying to compile and run programs - any programs it compiles will be for the target and not the host and so cannot be run. Manually implement staging scripts; Deal with lots of other more complex issues; Some of these items are covered in more detail in the advanced autoconf section.
Dependencies: What's needed to build and/or run the package? Dependencies should be familiar to anyone who has used an .rpm and .deb based desktop distribution. A dependency is something that a package requires either to run the package (a run-time dependency) or to build the package (a build-time or compile-time, dependency). There are two variables provided to allow the specifications of dependencies: DEPENDS Specifies build-time dependencies, via a list of bitbake recipes to build prior to building the recipe. These are programs (flex-native) or libraries (libpcre) that are required in order to build the package. RDEPENDS Specifies run-time dependencies, via a list of packages to install prior to installing the current package. These are programs or libraries that are required in order to run the program. Note that libraries which are dynamically linked to an application will be automatically detected and added to RDEPENDS and therefore do not need to be explicitly declared. If a library was dynamically loaded then it would need to be explicitly listed. If we take openssh for an example, it requires zlib and openssl in order to both build and run. In the recipe we have:DEPENDS = "zlib openssl"This tells bitbake that it will need to build and stage zlib and openssl prior to trying to build openssh, since openssh requires both of them. Note that there is no RDEPENDS even though openssh requires both of them to run. The run time dependencies on libz1 (the name of the package containing the zlib library) and libssl0 (the name of the package containing the ssl library) are automatically determined and added via the auto shared libs dependency code.
Methods: Built-in methods to make your life easier There are several helper functions defined by the base class, which is included by default for all recipes. Many of these are used a lot in both recipes and other classes. The most commonly seen, and most useful functions, include: oe_runmake This function is used to run make. However unlike calling make yourself this will pass the EXTRA_OEMAKE settings to make, will display a note about the make command and will check for any errors generated via the call to make. You should never have any reason to call make directly and should also use oe_runmake when you need to run make. oe_runconf (autotools only) This function is used to run the configure script of a package that is using the autotools class. This takes care of passing all of the correct parameters for cross-compiling and for installing into the appropriate target directory. It also passes the value of the EXTRA_OECONF variable to the configure script. For many situations setting EXTRA_OECONF is sufficient and you'll have no need to define your own configure task in which you call oe_runconf manually. If you need to write your own configure task for an autotools package you can use oe_runconf to manually call the configure process when it is required. The following example from net-snmp shows oe_runconf being called manually so that the parameter for specifying the endianess can be computed and passed in to the configure script:do_configure() { # Additional flag based on target endianess (see siteinfo.bbclass) ENDIANESS="${@base_conditional('SITEINFO_ENDIANNESS', 'le', '--with-endianness=little', '--with-endianness=big', d)}" oenote Determined endianess as: $ENDIANESS oe_runconf $ENDIANESS } oe_libinstall This function is used to install .so, .a and associated libtool .la libraries. It will determine the appropriate libraries to install and take care of any modifications that may be required for .la files. This function supports the following options: -C <dir> Change into the specified directory before attempting to install a library. Used when the libraries are in subdirectories of the main package. -s Require the presence of a .so library as one of the libraries that is installed. -a Require the presence of a .a library as one of the libraries that is installed. The following example from gdbm shows the installation of .so, .a (and associated .la) libraries into the staging library area:do_stage () { oe_libinstall -so -a libgdbm ${STAGING_LIBDIR} install -m 0644 ${S}/gdbm.h ${STAGING_INCDIR}/ } oenote Used to display informational messages to the user. The following example from net-snmp uses oenote to tell the user which endianess it determined was appropriate for the target device:do_configure() { # Additional flag based on target endianess (see siteinfo.bbclass) ENDIANESS="${@base_conditional('SITEINFO_ENDIANNESS', 'le', '--with-endianness=little', '--with-endianness=big', d)}" oenote Determined endianess as: $ENDIANESS oe_runconf $ENDIANESS } oewarn Used to display a warning message to the user, warning of something that may be problematic or unexpected. oedebug Used to display debugging related information. These messages will only be visible when bitbake is run with the -D flag to enable debug output. oefatal Used to display a fatal error message to the user, and then abort the bitbake run. The following example from linux-libc-headers shows the use of oefatal to tell the user when it cannot find the kernel source code for the specified target architecture:do_configure () { case ${TARGET_ARCH} in alpha*) ARCH=alpha ;; arm*) ARCH=arm ;; cris*) ARCH=cris ;; hppa*) ARCH=parisc ;; i*86*) ARCH=i386 ;; ia64*) ARCH=ia64 ;; mips*) ARCH=mips ;; m68k*) ARCH=m68k ;; powerpc*) ARCH=ppc ;; s390*) ARCH=s390 ;; sh*) ARCH=sh ;; sparc64*) ARCH=sparc64 ;; sparc*) ARCH=sparc ;; x86_64*) ARCH=x86_64 ;; esac if test ! -e include/asm-$ARCH; then oefatal unable to create asm symlink in kernel headers fi ... base_conditional (python) The base conditional python function is used to set a variable to one of two values based on the definition of a third variable. The general usage is:${@base_conditional('<variable-name>', '<value>', '<true-result>', '<false-result>', d)}"where: variable-name This is the name of a variable to check. value This is the value to compare the variable against. true-result If the variable equals the value then this is what is returned by the function. false-result If the variable does not equal the value then this is what is returned by the function. The ${@...} syntax is used to call python functions from within a recipe or class. This is described in more detail in the section. The following example from the openssl recipe shows the addition of either -DL_ENDIAN or -DB_ENDIAN depending on the value of SITEINFO_ENDIANNESS which is set to le for little endian targets and to be for big endian targets:do_compile () { ... # Additional flag based on target endianess (see siteinfo.bbclass) CFLAG="${CFLAG} ${@base_conditional('SITEINFO_ENDIANNESS', 'le', '-DL_ENDIAN', '-DB_ENDIAN', d)}" ...
Packaging: Defining packages and their contents A bitbake recipe is a set of instructions for creating one, or more, packages for installation on the target device. Typically these are .ipkg or .deb packages (although bitbake itself isn't associated with any particular packaging format). By default several packages are produced automatically without any special action required on the part of the recipe author. The following example of the packaging output from the helloworld example above shows this packaging in action:[NOTE: package helloworld-0.1-r0: task do_package_write: started NOTE: Not creating empty archive for helloworld-dbg-0.1-r0 Packaged contents of helloworld into /home/lenehan/devel/oe/build/titan-glibc-25/tmp/deploy/ipk/sh4/helloworld_0.1-r0_sh4.ipk Packaged contents of helloworld-doc into /home/lenehan/devel/oe/build/titan-glibc-25/tmp/deploy/ipk/sh4/helloworld-doc_0.1-r0_sh4.ipk NOTE: Not creating empty archive for helloworld-dev-0.1-r0 NOTE: Not creating empty archive for helloworld-locale-0.1-r0 NOTE: package helloworld-0.1-r0: task do_package_write: completedWe can see from above that the packaging did the following: Created a main package, helloworld_0.1-r0_sh4.ipk. This package contains the helloworld binary /usr/bin/helloworld. Created a documentation package, helloworld-doc_0.1-r0_sh4.ipk. This package contains the readme file /usr/share/doc/helloworld/README.txt. Considered creating a debug package, helloworld-dbg-0.1-r0_sh4.ipk, a development package helloworld-dev-0.1-r0_sh4.ipk and a locale package helloworld-locale-0.1-r0_sh4.ipk. It didn't create the packages due to the fact that it couldn't find any files that would actually go in the packages. There are several things happening here which are important to understand: There is a default set of packages that are considered for creation. This set of packages is controlled via the PACKAGES variable. For each package there is a default set of files and/or directories that are considered to belong to those packages. The documentation packages for example include anything found /usr/share/doc. The set of files and directories are controlled via the FILES_<package-name> variables. By default packages that contain no files are not created and no error is generated. The decision to create empty packages or not is controlled via the ALLOW_EMPTY variable.
Philosophy Separate packaging, where possible, is of high importance in OpenEmbedded. Many of the target devices have limited storage space and RAM and giving distributions and users the option of not installing a part of the package they don't need allows them to reduce the amount of storage space required. As an example almost no distributions will include documentation or development libraries since they are not required for the day to day operation of the device. In particular if your package provides multiple binaries, and it would be common to only use one or the other, then you should consider separating them into separate packages. By default several groups of files are automatically separated out, including: dev Any files required for development. This includes header files, static libraries, the shared library symlinks required only for linking etc. These would only ever need to be installed by someone attempting to compile applications on the target device. While this does happen it is very uncommon and so these files are automatically moved into a separate package doc Any documentation related files, including man pages. These are files which are of informational purposes only. For many embedded devices there is no way for the user to see any of the documentation anyway, and documentation can consume a lot of space. By separating these out they don't take any space by default but distributions and/or users may choose to install them if they need some documentation on a specific package. locale Locale information provides translation information for packages. Many users do not require these translations, and many devices will only want to provide them for user visible components, such as UI related items, and not for system binaries. By separating these out it is left up to the distribution or users to decide if they are required or not.
Default packages and files The default package settings are defined in conf/bitbake.conf and are suitable for a lot of recipes without any changes. The following list shows the default values for the packaging related variables: PACKAGES This variable lists the names of each of the packages that are to be generated.PACKAGES = "${PN}-dbg ${PN} ${PN}-doc ${PN}-dev ${PN}-locale"Note that the order of packages is important: the packages are processed in the listed order. So if two packages specify the same file then the first package listed in packages will get the file. This is important when packages use wildcards to specify their contents. For example if the main package, ${PN}, contains /usr/bin/* (i.e. all files in /usr/bin), but you want /usr/bin/tprogram in a separate package, ${PN}-tpackage, you would need to either ensure that ${PN}-tpackage is listed prior to ${PN} in PACKAGES or that FILES_${PN} was modified to not contain the wildcard that matches /usr/bin/tprogram. Note that the -dbg package contains the debugging information that has been extracted from binaries and libraries prior to them being stripped. This package should always be the first package in the package list to ensure that the debugging information is correctly extracted and moved to the package prior to any other packaging decisions being made. FILES_${PN} The base package, this includes everything needed to actually run the application on the target system.FILES_${PN} = "\ ${bindir}/* \ ${sbindir}/* \ ${libexecdir}/* \ ${libdir}/lib*.so.* \ ${sysconfdir} \ ${sharedstatedir} \ ${localstatedir} \ /bin/* \ /sbin/* \ /lib/*.so* \ ${datadir}/${PN} \ ${libdir}/${PN}/* \ ${datadir}/pixmaps \ ${datadir}/applications \ ${datadir}/idl \ ${datadir}/omf \ ${datadir}/sounds \ ${libdir}/bonobo/servers" FILES_${PN}-dbg The debugging information extracted from non-stripped versions of libraries and executable's. OpenEmbedded automatically extracts the debugging information into files in .debug directories and then strips the original files.FILES_${PN}-dbg = "\ ${bindir}/.debug \ ${sbindir}/.debug \ ${libexecdir}/.debug \ ${libdir}/.debug \ /bin/.debug \ /sbin/.debug \ /lib/.debug \ ${libdir}/${PN}/.debug" FILES_${PN}-doc Documentation related files. All documentation is separated into its own package so that it does not need to be installed unless explicitly required.FILES_${PN}-doc = "\ ${docdir} \ ${mandir} \ ${infodir} \ ${datadir}/gtk-doc \ ${datadir}/gnome/help" FILES_${PN}-dev Development related files. Any headers, libraries and support files needed for development work on the target.FILES_${PN}-dev = "\ ${includedir} \ ${libdir}/lib*.so \ ${libdir}/*.la \ ${libdir}/*.a \ ${libdir}/*.o \ ${libdir}/pkgconfig \ /lib/*.a \ /lib/*.o \ ${datadir}/aclocal" FILES_${PN}-locale Locale related files.FILES_${PN}-locale = "${datadir}/locale"
Wildcards Wildcards used in the FILES variables are processed via the python function fnmatch. The following items are of note about this function: /<dir>/*: This will match all files and directories in the dir - it will not match other directories. /<dir>/a*: This will only match files, and not directories. /dir: will include the directory dir in the package, which in turn will include all files in the directory and all subdirectories. Note that the order of packages effects the files that will be matched via wildcards. Consider the case where we have three binaries in the /usr/bin directory and we want the test program in a separate package:/usr/bin/programa /usr/bin/programb /usr/bin/testSo we define a new package and instruct bitbake to include /usr/bin/test in it. FILES-${PN}-test = "${bindir}/test" PACKAGES += "FILES-${PN}-test" When the package is regenerated no ${PN}-test package will be created. The reason for this is that the PACKAGES line now looks like this:{PN}-dbg ${PN} ${PN}-doc ${PN}-dev ${PN}-locale ${PN}-testNote how ${PN} is listed prior to ${PN}-test, and if we look at the definition of FILES-${PN} it contains the ${bindir}/* wildcard. Since ${PN} is first it'll match that wildcard and be moved into the ${PN} package prior to processing of the ${PN}-test package. To achieve what we are trying to accomplish we have two options: Modify the definition of ${PN} so that the wildcard does not match the test program. We could do this for example:FILES-${PN} = "${bindir}/p*"So now this will only match things in the bindir that start with p, and therefore not match our test program. Note that FILES-${PN} contains a lot more entries and we'd need to add any of the others that refer to files that are to be included in the package. In this case we have no other files, so it's safe to do this simple declaration. Modify the order of packages so that the ${PN}-test package is listed first. The most obvious way to do this would be to prepend our new package name to the packages list instead of appending it:PACKAGES =+ "FILES-${PN}-test"In some cases this would work fine, however there is a problem with this for packages that include binaries. The package will now be listed before the -dbg package and often this will result in the .debug directories being included in the package. In this case we are explicitly listing only a single file (and not using wildcards) and therefore it would be ok. In general it's more common to have to redefine the entire package list to include your new package plus any of the default packages that you require:PACKAGES = "${PN}-dbg ${PN}-test ${PN} ${PN}-doc ${PN}-dev ${PN}-locale"
Checking the packages During recipe development it's useful to be able to check on exactly what files went into each package, which files were not packaged and which packages contain no files. One of the easiest methods is to run find on the install directory. In the install directory there is one subdirectory created per package, and the files are moved into the install directory as they are matched to a specific package. The following shows the packages and files for the helloworld example:$ find tmp/work/helloworld-0.1-r0/install tmp/work/helloworld-0.1-r0/install tmp/work/helloworld-0.1-r0/install/helloworld-locale tmp/work/helloworld-0.1-r0/install/helloworld-dbg tmp/work/helloworld-0.1-r0/install/helloworld-dev tmp/work/helloworld-0.1-r0/install/helloworld-doc tmp/work/helloworld-0.1-r0/install/helloworld-doc/usr tmp/work/helloworld-0.1-r0/install/helloworld-doc/usr/share tmp/work/helloworld-0.1-r0/install/helloworld-doc/usr/share/doc tmp/work/helloworld-0.1-r0/install/helloworld-doc/usr/share/doc/helloworld tmp/work/helloworld-0.1-r0/install/helloworld-doc/usr/share/doc/helloworld/README.txt tmp/work/helloworld-0.1-r0/install/helloworld tmp/work/helloworld-0.1-r0/install/helloworld/usr tmp/work/helloworld-0.1-r0/install/helloworld/usr/bin tmp/work/helloworld-0.1-r0/install/helloworld/usr/bin/helloworld $The above shows that the -local, -dbg and -dev packages are all empty, and the -doc and base package contain a single file each. Using the "-type f" option to find to show just files will make this clearer as well. In addition to the install directory the image directory (which corresponds to the destination directory, D) will contain any files that were not packaged:$ find tmp/work/helloworld-0.1-r0/image tmp/work/helloworld-0.1-r0/image tmp/work/helloworld-0.1-r0/image/usr tmp/work/helloworld-0.1-r0/image/usr/bin tmp/work/helloworld-0.1-r0/image/usr/share tmp/work/helloworld-0.1-r0/image/usr/share/doc tmp/work/helloworld-0.1-r0/image/usr/share/doc/helloworld $In this case all files were packaged and so there are no left over files. Using find with "-type f" makes this much clearer:$ find tmp/work/helloworld-0.1-r0/image -type f $ Messages regarding missing files are also displayed by bitbake during the package task:NOTE: package helloworld-0.1-r0: task do_package: started NOTE: the following files were installed but not shipped in any package: NOTE: /usualdir/README.txt NOTE: package helloworld-0.1-r0: task do_package: completedExcept in very unusual circumstances there should be no unpackaged files left behind by a recipe.
Excluding files There's no actual support for explicitly excluding files from packaging. You could just leave them out of any package, but then you'll get warnings (or errors if requesting full package checking) during packaging which is not desirable. It also doesn't let other people know that you've deliberately avoided packaging the file or files. In order to exclude a file totally you should avoid installing it in the first place during the install task. In some cases it may be easier to let the package install the file and then explicitly remove the file at the end of the install task. The following example from the samba recipe shows the removal of several files that get installed via the default install task generated by the . By using do_install_append these commands are run after the autotools generated install task: do_install_append() { ... rm -f ${D}${bindir}/*.old rm -f ${D}${sbindir}/*.old ... }
Debian naming A special debian library name policy can be applied for packages that contain a single shared library. When enabled packages will be renamed to match the debian policy for such packages. Debian naming is enabled by including the debian class via either local.conf or your distribution's configuration file:INHERIT += "debian" The policy works by looking at the shared library name and version and will automatically rename the package to <libname><lib-major-version>. For example if the package name (PN) is foo and the package ships a file named libfoo.so.1.2.3 then the package will be renamed to libfoo1 to follow the debian policy. If we look at the lzo_1.08.bb recipe, currently at release 14, it generates a package containing a single shared library :$ find tmp/work/lzo-1.08-r14/install/ tmp/work/lzo-1.08-r14/install/lzo tmp/work/lzo-1.08-r14/install/lzo/usr tmp/work/lzo-1.08-r14/install/lzo/usr/lib tmp/work/lzo-1.08-r14/install/lzo/usr/lib/liblzo.so.1 tmp/work/lzo-1.08-r14/install/lzo/usr/lib/liblzo.so.1.0.0Without debian naming this package would have been called lzo_1.08-r14_sh4.ipk (and the corresponding dev and dbg packages would have been called lzo-dbg_1.08-r14_sh4.ipk and lzo-dev_1.08-r14_sh4.ipk). However with debian naming enabled the package is renamed based on the name of the shared library, which is liblzo.so.1.0.0 in this case. So the name lzo is replaced with liblzo1:$ find tmp/deploy/ipk/ -name '*lzo*' tmp/deploy/ipk/sh4/liblzo1_1.08-r14_sh4.ipk tmp/deploy/ipk/sh4/liblzo-dev_1.08-r14_sh4.ipk tmp/deploy/ipk/sh4/liblzo-dbg_1.08-r14_sh4.ipk Some variables are available which effect the operation of the debian renaming class: LEAD_SONAME If the package actually contains multiple shared libraries then one will be selected automatically and a warning will be generated. This variable is a regular expression which is used to select which shared library from those available is to be used for debian renaming. DEBIAN_NOAUTONAME_<pkgname> If this variable is set to 1 for a package then debian renaming will not be applied for the package. AUTO_LIBNAME_PKGS If set this variable specifies the prefix of packages which will be subject to debian renaming. This can be used to prevent all of the packages being renamed via the renaming policy.
Empty packages By default empty packages are ignored. Occasionally you may wish to actually create empty packages, typically done when you want a virtual package which will install other packages via dependencies without actually installing anything itself. The ALLOW_EMPTY variable is used to control the creation of empty packages: ALLOW_EMPTY Controls if empty packages will be created or not. By default this is "0" and empty packages are not created. Setting this to "1" will permit the creation of empty packages (packages containing no files).
Tasks: Playing with tasks Bitbake steps through a series of tasks when building a recipe. Sometimes you need to explicitly define what a class does, such as providing a do_install function to implement the install task in a recipe and sometimes they are provided for you by common classes, such as the autotools class providing the default implementations of configure, compile and install tasks. There are several methods available to modify the tasks that are being run: Overriding the default task implementation By defining your own implementation of a task you'll override any default or class provided implementations. For example, you can define your own implementation of the compile task to override any default implementation:do_compile() { oe_runmake DESTDIR=${D} } If you wish to totally prevent the task from running you need to define your own empty implementation. This is typically done via the definition of the task using a single colon:do_configure() { : } Appending or prepending to the task Sometimes you want the default implementation, but you require additional functionality. This can done by appending or pre-pending additional functionality onto the task. The following example from units shows an example of installing an addition file which for some reason was not installed via the autotools normal install task:do_install_append() { install -d ${D}${datadir} install -m 0655 units.dat ${D}${datadir} } The following example from the cherokee recipe shows an example of adding functionality prior to the default install task. In this case it compiles a program that is used during installation natively so that it will work on the host. Without this the autotools default install task would fail since it'd try to run the program on the host which was compiled for the target:do_install_prepend () { # It only needs this app during the install, so compile it natively $BUILD_CC -DHAVE_SYS_STAT_H -o cherokee_replace cherokee_replace.c } Defining a new task Another option is to define a totally new task, and then register that with bitbake so that it runs in between two of the existing tasks. The following example shows a situation in which a cvs tree needs to be copied over the top of an extracted tar.gz archive, and this needs to be done before any local patches are applied. So a new task is defined to perform this action, and then that task is registered to run between the existing unpack and patch tasks:do_unpack_extra(){ cp -pPR ${WORKDIR}/linux/* ${S} } addtask unpack_extra after do_unpack before do_patch The task to add does not have the do_ prepended to it, however the tasks to insert it after and before do have the do_ prepended. No errors will be generated if this is wrong, the additional task simply won't be executed. Using overrides Overrides (described fully elsewhere) allow for various functionality to be performed conditionally based on the target machine, distribution, architecture etc. While not commonly used it is possible to use overrides when defining tasks. The following example from udev shows an additional file being installed for the specified machine only by performing an append to the install task for the h2200 machine only:do_install_append_h2200() { install -m 0644 ${WORKDIR}/50-hostap_cs.rules ${D}${sysconfdir}/udev/rules.d/50-hostap_cs.rules }
Classes: The separation of common functionality Often a certain pattern is followed in more than one recipe, or maybe some complex python based functionality is required to achieve the desired end result. This is achieved through the use of classes, which can be found in the classes subdirectory at the top-level of an OE checkout. Being aware of the available classes and understanding their functionality is important because classes: Save developers time by performing actions that they would otherwise need to perform themselves; Perform a lot of actions in the background making a lot of recipes difficult to understand unless you are aware of classes and how they work; A lot of detail on how things work can be learnt from looking at how classes are implemented. A class is used via the inherit method. The following is an example for the curl recipe showing that it uses three classes:inherit autotools pkgconfig binconfigIn this case it is utilizing the services of three separate classes: autotools The is used by programs that use the GNU configuration tools and takes care of the configuration and compilation of the software; pkgconfig The is used to stage the .pc files which are used by the pkg-config program to provide information about the package to other software that wants to link to this software; binconfig The is used to stage the <name>-config files which are used to provide information about the package to other software that wants to link to this software; Each class is implemented via the file in the classes subdirectory named <classname>.bbclass and these can be examined for further details on a particular class, although sometimes it's not easy to understand everything that's happening. Many of the classes are covered in detail in various sections in this user manual.
Staging: Making includes and libraries available for building Staging is the process of making files, such as include files and libraries, available for use by other recipes. This is different to installing because installing is about making things available for packaging and then eventually for use on the target device. Staging on the other hand is about making things available on the host system for use in building later applications. Taking bzip2 as an example you can see that it stages a header file and its library files:do_stage () { install -m 0644 bzlib.h ${STAGING_INCDIR}/ oe_libinstall -a -so libbz2 ${STAGING_LIBDIR} } The oe_libinstall method used in the bzip2 recipe is described in the section, and it takes care of installing libraries (into the staging area in this case). The staging variables are automatically defined to the correct staging location, in this case the main staging variables are used: STAGING_INCDIR The directory into which staged header files should be installed. This is the equivalent of the standard /usr/include directory. STAGING_LIBDIR The directory into which staged library files should be installed. This is the equivalent of the standard /usr/lib directory. Additional staging related variables are covered in the section in . Looking in the staging area under tmp you can see the result of the bzip2 recipes staging task:$ find tmp/staging -name '*bzlib*' tmp/staging/sh4-linux/include/bzlib.h $ find tmp/staging -name '*libbz*' tmp/staging/sh4-linux/lib/libbz2.so tmp/staging/sh4-linux/lib/libbz2.so.1.0 tmp/staging/sh4-linux/lib/libbz2.so.1 tmp/staging/sh4-linux/lib/libbz2.so.1.0.2 tmp/staging/sh4-linux/lib/libbz2.a As well as being used during the stage task the staging related variables are used when building other packages. Looking at the gnupg recipe we see two bzip2 related items:DEPENDS = "zlib bzip2" ... EXTRA_OECONF = "--disable-ldap \ --with-zlib=${STAGING_LIBDIR}/.. \ --with-bzip2=${STAGING_LIBDIR}/.. \ --disable-selinux-support" Bzip2 is referred to in two places in the recipe: DEPENDS Remember that DEPENDS defines the list of build time dependencies. In this case the staged headers and libraries from bzip2 are required to build gnupg, and therefore we need to make sure the bzip2 recipe has run and staged the headers and libraries. By adding the DEPENDS on bzip2 this ensures that this happens. EXTRA_OECONF This variable is used by the to provide options to the configure script of the package. In the gnupg case it needs to be told where the bzip2 headers and libraries are, and this is done via the --with-bzip2 option. In this case it points to the directory which includes the lib and include subdirectories. Since OE doesn't define a variable for one level above the include and lib directories .. is used to indicate one directory up. Without this, gnupg would search the host system headers and libraries instead of those we have provided in the staging area for the target. Remember that staging is used to make things, such as headers and libraries, available for use by other recipes later on. While headers and libraries are the most common items requiring staging, other items such as the pkgconfig files need to be staged as well. For native packages, the binaries also need to be staged.
Autoconf: All about autotools This section is to be completed: About building autoconf packages EXTRA_OECONF Problems with /usr/include, /usr/lib Configuring to search in the staging area -L${STAGING_LIBDIR} vs ${TARGET_LDFLAGS} Site files
Installation scripts: Running scripts during package installation and/or removal Packaging systems such as .ipkg and .deb support pre and post installation and pre and post removal scripts which are run during package installation and/or package removal on the target system. These scripts can be defined in your recipes to enable actions to be performed at the appropriate time. Common uses include starting new daemons on installation, stopping daemons during uninstall, creating new user and/or group entries during install, registering and unregistering alternative implementations of commands and registering the need for volatiles. The following scripts are supported: preinst The preinst script is run prior to installing the contents of the package. During preinst the contents of the package are not available to be used as part of the script. The preinst scripts are not commonly used. postinst The postinst script is run after the installation of the package has completed. During postinst the contents of the package are available to be used. This is often used for the creation of volatile directories, registration of daemons, starting of daemons and fixing up of SUID binaries. prerm The prerm is run prior to the removal of the contents of a package. During prerm the contents of the package are still available for use by the script. postrm The postrm script is run after the completion of the removal of the contents of a package. During postrm the contents of the package no longer exist and therefore are not available for use by the script. Postrm is most commonly used for update alternatives (to tell the alternatives system that this alternative is not available and another should be selected). Scripts are registered by defining a function for: pkg_<scriptname>_<packagename> The following example from ndisc6 shows postinst scripts being registered for three of the packages that ndisc6 creates:# Enable SUID bit for applications that need it pkg_postinst_${PN}-rltraceroute6 () { chmod 4555 ${bindir}/rltraceroute6 } pkg_postinst_${PN}-ndisc6 () { chmod 4555 ${bindir}/ndisc6 } pkg_postinst_${PN}-rdisc6 () { chmod 4555 ${bindir}/rdisc6 } These scripts will be run via /bin/sh on the target device, which is typically the busybox sh but could also be bash or some other sh compatible shell. As always you should not use any bash extensions in your scripts and stick to basic sh syntax. Note that several classes will also register scripts, and that any script you declare will have the script for the classes appended by these classes. The following classes all generate additional script contents: update-rc.d This class is used by daemons to register their init scripts with the init code. Details are provided in the section. module This class is used by linux kernel modules. It's responsible for calling depmod and update-modules during kernel module installation and removal. kernel This class is used by the linux kernel itself. There is a lot of housekeeping required both when installing and removing a kernel and this class is responsible for generating the required scripts. qpf This class is used when installing and/or removing qpf fonts. It registers scripts to update the font paths and font cache information to ensure that the font information is kept up to date as fonts are installed and removed. update-alternatives This class is used by packages that contain binaries which may also be provided by other packages. It tells the system that another alternative is available for consideration. The alternatives system will create a symlink to the correct alternative from one or more available on the system. Details are provided in the section. gtk-icon-cache This class is used by packages that add new gtk icons. It's responsible for updating the icon cache when packages are installed and removed. gconf package The base class used by packaging classes such as those for .ipkg and .deb. The package class may create scripts used to update the dynamic linker's ld cache. The following example from p3scan shows a postinst script which ensures that the required user and group entries exist, and registers the need for volatiles (directories and/or files under /var). In addition to explicitly declaring a postinst script it uses the update-rc.d class which will result in an additional entry being added to the postinst script to register the init scripts and start the daemon (via call to update-rc.d as described in the section).inherit autotools update-rc.d ... # Add havp's user and groups pkg_postinst_${PN} () { grep -q mail: /etc/group || addgroup --system havp grep -q mail: /etc/passwd || \ adduser --disabled-password --home=${localstatedir}/mail --system \ --ingroup mail --no-create-home -g "Mail" mail /etc/init.d/populate-volatile.sh update } Several scripts in existing recipes will be of the following form:if [ x"$D" = "x" ]; then ... fi This is testing if the installation directory, D, is defined and if it is no actions are performed. The installation directory will not be defined under normal circumstances. The primary use of this test is to permit the application to be installed during root filesystem generation. In that situation the scripts cannot be run since the root filesystem is generated on the host system and not on the target. Any required script actions would need to be performed via an alternative method if the package is to be installed in the initial root filesystem (such as including any required users and groups in the default passwd and group files for example.)
Configuration files Configuration files that are installed as part of a package require special handling. Without special handling as soon as the user upgrades to a new version of the package then changes they have made to the configuration files will be lost. In order to prevent this from happening you need to tell the packaging system which files are configuration files. Such files will result in the user being asked how they want to handle any configuration file changes (if any), as shown in this example:Downloading http://nynaeve.twibble.org/ipkg-titan-glibc//./p3scan_2.9.05d-r1_sh4.ipk Configuration file '/etc/p3scan/p3scan.conf' ==> File on system created by you or by a script. ==> File also in package provided by package maintainer. What would you like to do about it ? Your options are: Y or I : install the package maintainer's version N or O : keep your currently-installed version D : show the differences between the versions (if diff is installed) The default action is to keep your current version. *** p3scan.conf (Y/I/N/O/D) [default=N] ?To declare a file as a configuration file you need to define the CONFFILES_<pkgname> variable as a whitespace separated list of configuration files. The following example from clamav shows two files being marked as configuration files:CONFFILES_${PN}-daemon = "${sysconfdir}/clamd.conf \ ${sysconfdir}/default/clamav-daemon"Note the use of ${PN}-daemon as the package name. The ${PN} variable will expand to clamav and therefore these conf files are declared as being in the clamav-daemon package.
Package relationships Explicit relationships between packages are support by packaging formats such as ipkg and deb. These relationships include describing conflicting packages and recommended packages. The following variables control the package relationships in the recipes: RRECOMMENDS Used to specify other packages that are recommended to be installed when this package is installed. Generally this means while the recommended packages are not required they provide some sort of functionality which users would usually want. RCONFLICTS Used to specify other packages that conflict with this package. Two packages that conflict cannot be installed at the same time. RREPLACES Used to specify that the current package replaces an older package with a different name. During package installation the package that is being replaced will be removed since it is no longer needed when this package is installed. RSUGGESTS Used to provide a list of suggested packages to install. These are packages that are related to and useful for the current package but which are not actually required to use the package. RPROVIDES Used to explicitly specify what a package provides at runtime. For example hotplug support is provided by several packages, such as udev and linux-hotplug. Both declare that they runtime provide "hotplug". So any packages that require "hotplug" to work simply declare that it RDEPENDS on "hotplug". It's up to the distribution to specify which actual implementation of "virtual/hotplug" is used. PROVIDES Used to explicitly specify what a package provides at build time. This is typically used when two or more packages can provide the same functionality. For example there are several different X servers in OpenEmbedded, and each declared as providing "virtual/xserver". Therefore a package that depends on an X server to build can simply declare that it DEPENDS on "virtual/xserver". It's up to the distribution to specify which actual implementation of "virtual/xserver" is used.
Fakeroot: Dealing with the need for "root" Sometimes packages require root permissions in order to perform some action, such as changing user or group owners or creating device nodes. Since OpenEmbedded will not keep the user and group information it's usually preferable to remove that from the makefiles. For device nodes it's usually preferable to create them from the initial device node lists or via udev configuration. However if you can't get by without root permissions then you can use to simulate a root environment, without the need to really give root access. Using is done by prefixing the task:fakeroot do_install() {Since this requires fakeroot you also need to add a dependency on fakeroot-native:DEPENDS = "fakeroot-native"See the fuse recipe for an example. Further information on , including a description of how it works, is provided in the reference section: .
Native: Packages for the build host This section is to be completed. What native packages are Using require with the non-native package
Development: Strategies for developing recipes This section is to be completed. How to go about developing recipes How to handle incrementally creating patches How to deal with site file issues Strategies for autotools issues
Advanced versioning: How to deal with rc and pre versions Special care needs to be taken when specify the version number for rc and pre versions of packages. Consider the case where we have an existing 1.5 version and there's a new 1.6-rc1 release that you want to add. 1.5: Existing version; 1.6-rc1: New version. If the new package is given the version number 1.6-rc1 then everything will work fine initially. However when the final release happens it will be called 1.6. If you now create a 1.6 version of the package you'll find that the packages are sorted into the following order: 1.5 1.6 1.6-rc1 This results in the packaging system, such as ipkg, considering the released version to be older than the rc version. In OpenEmbedded the correct naming of pre and rc versions is to use the previous version number followed by a + followed by the new version number. So the 1.6-rc1 release would be given the version number: 1.5+1.6-rc1 These would result in the eventually ordering being: 1.5 1.5+1.6-rc1 1.6 This is the correct order and the packaging system will now work as expected.
Require/include: Reusing recipe contents In many packages where you are maintaining multiple versions you'll often end up with several recipes which are either identical, or have only minor differences between them. The require and/or include directive can be used to include common content from one file into other. You should always look for a way to factor out common functionality into an include file when adding new versions of a recipe. Both require and include perform the same function - including the contents of another file into this recipe. The difference is that require will generate an error if the file is not found while include will not. For this reason include should not be used in new recipes. For example the clamav recipe looks like this:require clamav.inc PR = "r0"Note that all of the functionality of the recipe is provided in the clamav.inc file, only the release number is defined in the recipe. Each of the recipes includes the same clamav.inc file to save having to duplicate any functionality. This also means that as new versions are released it's a simple matter of copying the recipe and resetting the release number back to zero. The following example from iproute2 shows the recipe adding additional patches that are not specified by the common included file. These are patches only needed for newer release and by only adding them in this recipe it permits the common code to be used for both old and new recipes:PR = "r1" SRC_URI += "file://iproute2-2.6.15_no_strip.diff;striplevel=0 \ file://new-flex-fix.patch" require iproute2.inc DATE = "060323" The following example from cherokee shows a similar method of including additional patches for this version only. However it also shows another technique in which the configure task is defined in the recipe for this version, thus replacing the configure task that is provided by the common include:PR = "r7" SRC_URI_append = "file://configure.patch \ file://Makefile.in.patch \ file://Makefile.cget.patch \ file://util.patch" require cherokee.inc do_configure() { gnu-configize oe_runconf sed -i 's:-L\$:-L${STAGING_LIBDIR} -L\$:' ${S}/*libtool }
Python: Advanced functionality with python Recipes permit the use of python code in order to perform complex operations which are not possible with the normal recipe syntax and variables. Python can be used in both variable assignments and in the implementation of tasks. For variable assignments python code is indicated via the use of ${@...}, as shown in the following example:TAG = ${@bb.data.getVar('PV',d,1).replace('.', '_')} The above example retrieves the PV variable from the bitbake data object, then replaces any dots with underscores. Therefore if the PV was 0.9.0 then TAG will be set to 0_9_0. Some of the more common python code in use in existing recipes is shown in the following list: bb.data.getVar(<var>,d,1) Retrieve the data for the specified variable from the bitbake database for the current recipe. <variable>.replace(<key>, <replacement>) Find each instance of the key and replace it with the replacement value. This can also be used to remove part of a string by specifying '' (two single quotes) as the replacement. The following example would remove the '-frename-registers' option from the CFLAGS variable:CFLAGS := "${@'${CFLAGS}'.replace('-frename-registers', '')}" os.path.dirname(<filename>) Return the directory only part of a filename. This is most commonly seen in existing recipes when setting the FILESDIR variable (as described in the section). By obtaining the name of the recipe file itself, FILE, and then using os.path.dirname to strip the filename part:FILESDIR = "${@os.path.dirname(bb.data.getVar('FILE',d,1))}/make-${PV}"Note however that this is no longer required as FILE_DIRNAME is automatically set to the dirname of the FILE variable and therefore this would be written in new recipes as:FILESDIR = "$FILE_DIRNAME/make-${PV}" <variable>.split(<key>)[<index>] Splits the variable around the specified key. Use [<index>] to select one of the matching items from the array generated by the split command. The following example from the recipe genext2fs_1.3+1.4rc1.bb would take the PV of 1.3+1.4rc1 and split it around the + sign, resulting in an array containing 1.3 and 1.4rc1. It then uses the index of [1] to select the second item from the list (the first item is at index 0). Therefore TRIMMEDV would be set to 1.4rc1 for this recipe: TRIMMEDV = "${@bb.data.getVar('PV', d, 1).split('+')[1]}" As well as directly calling built-in python functions, those functions defined by the existing classes may also be called. A set of common functions is provided by the base class in classes/base.bbclass: base_conditional This function is used to set a variable to one of two values based on the definition of a third variable. The general usage is:${@base_conditional('<variable-name>', '<value>', '<true-result>', <false-result>', d)}"where: variable-name This is the name of a variable to check. value This is the value to compare the variable against. true-result If the variable equals the value then this is what is returned by the function. false-result If the variable does not equal the value then this is what is returned by the function. The following example from the openssl recipe shows the addition of either -DL_ENDIAN or -DB_ENDIAN depending on the value of SITEINFO_ENDIANNESS which is set to le for little endian targets and to be for big endian targets:do_compile () { ... # Additional flag based on target endianess (see siteinfo.bbclass) CFLAG="${CFLAG} ${@base_conditional('SITEINFO_ENDIANNESS', 'le', '-DL_ENDIAN', '-DB_ENDIAN', d)}" ... base_contains Similar to base_conditional except that it is checking for the value being an element of an array. The general usage is:${@base_contains('<array-name>', '<value>', '<true-result>', <false-result>', d)}" where: array-name This is the name of the array to search. value This is the value to check for in the array. true-result If the value is found in the array then this is what is returned by the function. false-result If the value is not found in the array then this is what is returned by the function. The following example from the task-angstrom-x11 recipe shows base_contains being used to add a recipe to the runtime dependency list but only for machines which have a touchscreen: RDEPENDS_angstrom-gpe-task-base := "\ ... ${@base_contains("MACHINE_FEATURES", "touchscreen", "libgtkstylus", "",d)} \ ... Tasks may be implemented in python by prefixing the task function with "python ". In general this should not be needed and should be avoided where possible. The following example from the devshell recipe shows how the compile task is implemented in python:python do_compile() { import os import os.path workdir = bb.data.getVar('WORKDIR', d, 1) shellfile = os.path.join(workdir, bb.data.expand("${TARGET_PREFIX}${DISTRO}-${MACHINE}-devshell", d)) f = open(shellfile, "w") # emit variables and shell functions devshell_emit_env(f, d, False, ["die", "oe", "autotools_do_configure"]) f.close() }
Preferences: How to disable packages When bitbake is asked to build a package and multiple versions of that package are available then bitbake will normally select the version that has the highest version number (where the version number is defined via the PV variable). For example if we were to ask bitbake to build procps and the following packages are available:$ ls recipes/procps procps-3.1.15/ procps-3.2.1/ procps-3.2.5/ procps-3.2.7/ procps.inc procps_3.1.15.bb procps_3.2.1.bb procps_3.2.5.bb procps_3.2.7.bb $then we would expect it to select version 3.2.7 (the highest version number) to build. Sometimes this is not actually what you want to happen though. Perhaps you have added a new version of the package that does not yet work or maybe the new version has no support for your target yet. Help is at hand since bitbake is not only looking at the version numbers to decided which version to build but it is also looking at the preference for each of those version. The preference is defined via the DEFAULT_PREFERENCE variable contained within the recipe. The default preference (when no DEFAULT_PREFERENCE is specified) is zero. Bitbake will find the highest preference that is available and then for all the packages at the preference level it will select the package with the highest version. In general this means that adding a positive DEFAULT_PREFERENCE will cause the package to be preferred over other versions and a negative DEFAULT_PREFERENCE will cause all other packages to be preferred. Imagine that you are adding procps version 4.0.0, but that it does not yet work. You could delete or rename your new recipe so you can build a working image, but what you really want to do is just ignore the new 4.0.0 version until it works. By adding:DEFAULT_PREFERENCE = "-1"to the recipe this is what will happen. Bitbake will now ignore this version (since all of the existing versions have a preference of 0). Note that you can still call bitbake directly on the recipe:bitbake -b recipes/procps/procps_4.0.0.bbThis enables you to test, and fix the package manually without having bitbake automatically select it normally. By using this feature in conjunction with overrides you can also disable (or select) specific versions based on the override. The following example from glibc shows that this version has been disabled for the sh3 architecture because it doesn't support sh3. This will force bitbake to try and select one of the other available versions of glibc instead:recipes/glibc/glibc_2.3.2+cvs20040726.bb:DEFAULT_PREFERENCE_sh3 = "-99"
Initscripts: How to handle daemons This section is to be completed. update-rc.d class sh syntax stop/stop/restart params sample/standard script? volatiles
Alternatives: How to handle the same command in multiple packages Alternatives are used when the same command is provided by multiple packages. A classic example is busybox, which provides a whole set of commands such as /bin/ls and /bin/find, which are also provided by other packages such as coreutils (/bin/ls) and findutils (/bin/find). A system for handling alternatives is required to allow the user to choose which version of the command they wish to have installed. It should be possible to install either one, or both, or remove one when both are installed etc, and to have no issues with the packages overwriting files from other packages. The most common reason for alternatives is to reduce the size of the binaries. By cutting down on features, built in help, error messages and combining multiple binaries into one large binary it's possible to save considerable space. Often users are not expected to use the commands interactively in embedded appliances and therefore these changes have no visible effect to the user. In some situations users may have interactive access, or they may be more advanced users who want shell access on appliances that normally don't provide it, and in these cases they should be able to install the full functional version if they desire.
Example of alternative commands Most distributions include busybox in place of the full featured version of the commands. The following example shows a typical install in which the find command, which we'll use as an example here, is the busybox version:root@titan:~$ find --version find --version BusyBox v1.2.1 (2006.12.17-05:10+0000) multi-call binary Usage: find [PATH...] [EXPRESSION] root@titan:~$ which find which find /usr/bin/findIf we now install the full version of find:root@titan:~$ ipkg install findutils ipkg install findutils Installing findutils (4.2.29-r0) to root... Downloading http://nynaeve.twibble.org/ipkg-titan-glibc//./findutils_4.2.29-r0_sh4.ipk Configuring findutils update-alternatives: Linking //usr/bin/find to find.findutils update-alternatives: Linking //usr/bin/xargs to xargs.findutils Then we see that the standard version of find changes to the full featured implementation:root@titan:~$ find --version find --version GNU find version 4.2.29 Features enabled: D_TYPE O_NOFOLLOW(enabled) LEAF_OPTIMISATION root@titan:~$ which find which find /usr/bin/find
Using update-alternatives Two methods of using the alternatives system are available: Via the . This is the simplest method, but is not usable in all situations. Via directly calling the update-alternatives command. The provides the simplest method of using alternatives but it only works for a single alternative. For multiple alternatives they need to be manually registered during post install. Full details on both methods is provided in the section of the reference manual.
Volatiles: How to handle the /var directory The /var directory is for storing volatile information, that is information which is constantly changing and which in general may be easily recreated. In embedded applications it is often desirable that such files are not stored on disk or flash for various reasons including: The possibility of a reduced lifetime of the flash; The limited amount of storage space available; To ensure filesystem corruption cannot occur due to a sudden power loss. For these reasons many of the OpenEmbedded distributions use a tmpfs based memory filesystem for /var instead of using a disk or flash based filesystem. The consequence of this is that all contents of the /var directory is lost when the device is powered off or restarted. Therefore special handling of /var is required in all packages. Even if your distribution does not use a tmpfs based /var you need to assume it does when creating packages to ensure the package can be used on those distributions that do use a tmpfs based /var. This special handling is provided via the populate-volatiles.sh script. If your package requires any files, directories or symlinks in /var then it should be using the populate-volatiles facilities.
Declaring volatiles This section is to be completed. how volatiles work default volatiles don't include any /var stuff in packages even if your distro doesn't use /var in tmpfs, others do updating the volatiles cache during install
Logging and log files As a consequence of the volatile and/or small capacity of the /var file system some distributions choose methods of logging other than writing to a file. The most typical is the use of an in-memory circular log buffer which can be read using the logread command. To ensure that each distribution is able to implement logging in a method that is suitable for its goals all packages should be configured by default to log via syslog, and not log directly to a file, if possible. If the distribution and/or end-user requires logging to a file then they can configure syslog and/or your application to implement this.
Summary In summary the following are required when dealing with /var: Configure all logging to use syslog whenever possible. This leaves the decision on where to log up to the individual distributions. Don't include any /var directories, files or symlinks in packages. They would be lost on a reboot and so should not be included in packages. The only directories that you can assume exist are those listed in the default volatiles file: recipes/initscripts/initscripts-1.0/volatiles. For any other directories, files or links that are required in /var you should install your own volatiles list as part of the package.
Miscellaneous This section is to be completed. about optimization about download directories about parallel builds about determining endianess (aka net-snmp, openssl, hping etc style) about PACKAGES_DYNAMIC about LEAD_SONAME about "python () {" - looks like it is always run when a recipe is parsed? see pam/libpam about SRCDATE with svn/cvs? about INHIBIT_DEFAULT_DEPS? about COMPATIBLE_MACHINE and COMPATIBLE_HOST about SUID binaries, and the need for postinst to fix them up about passwd and group (some comment in install scripts section already).