auto-build.conf - Test-AutoBuild configuration file
The auto-build.conf file contains the configuration information for a build instance.
It is used to define general metadata such as the host name of build machine,
the administrator / developer email addresses,
email alerts; the parameters for accessing source control repositories; the modules to be checked out from source control and built.
The configuration file is processed by the Config::Record(3pm) module, so refer to that manual page for formal syntax rules. Prior to loading it is also run through the Template(3pm) module - the Template::Manual::Syntax(3pm) manual page details the syntax for the pre-processor. The pre-processor is most useful if there are a large number of modules to be defined, each with pretty much identical configuration parameters.
The example configuration file contains inline comments summarizing important points, however, the full details are as follows:
The fully qualified path to the build engine definition.
There is only one build engine setup currently provided,
so this can be left on its default setting of /etc/auto-build.d/engine/host-main.conf At a later date extra engines will be provided for doing builds with chroot,
or virtual machines.
The fully qualified path to the directory containing template files for the HTML status pages,
RSS feed and email alerts.
The template files are all processed using the Template(3pm) module,
so refer to the Template::Manual::Syntax(3pm) manual page for details on the syntax.
The path to the directory within which the build instance will operate.
This directory should previously have been populated with pre-requisite sub-dirs by running the auto-build-make-root(1) script.
The RPM install perl-Test-AutoBuild-account installs the correct directory structure under /var/lib/builder,
so if the build host only needs to run a sigle instance of the builder,
this parameter can be left unchanged.
A short descriptive name for this build configuration. This label is used as the page title for HTML status pages, the subject in the email alerts, and the title of the RSS feed.
The email address for the person responsible for administering the build instance.
The name of the person responsible for administering the build instance.
The email address for the development team whose software is being built. This would typically be the address of the main developers mailing list.
The name of the development team whose software is being built, or alternatively the name of the developers mailing list.
When the build host has multiple hostname aliases configured (for virtual hosting), specifies the preferred hostname for accessing the FTP / HTTP sites. This is used for generating links within the email alerts and RSS feeds.
The URL path prefix under which the HTML status pages are accessible.
If using the /etc/auto-build.d/http/aliased.conf HTTP config file,
then this would be set to /builder.
If the user.conf HTTP config file is used,
then this would be set to /~builder.
Finally,
if the vhost.conf HTTP config file is used,
then this is simply set to the empty string.
The method used for acquiring a lock file to prevent multiple build instances running at the same time against the same buildRoot directory.
This is the best option, but is only implemented for Linux, SunOS and FreeBSD. It correctly protects from multiple build instances running against a single build root, even if the build root is on NFS and accessed from multiple hosts.
This is more portable than fcntl,
working on any UNIX platform,
but it does not protect a NFS hosted build root from being accessed by multiple hosts.
This uses a simple file creation/deletion test. The downside is that if the build engine or host machine crashes, it will be neccessary to manually delete the lock file ($buildRoot/.build.mutex) before the builder will run again.
There are various features available in the build engine workflow which can be turned on/off as desired.
They are all contained within a nested features block.
If the cache is enabled, then modules will only be built if their sources have changed since the previous build cycle. This provides much faster cycle times if there are many modules in the build config, only a few of which ever change
Whether to try and checkout the latest sources from version control If checkout is disabled, the build will run with whatever sources #were checked out on the previous cycle.
Whether to run the 'createrepo' tool against the RPMs on the HTTP distribution site. The XML metadata files created are used by package update tools such as Yum for resolving dependancies between RPMs.
Whether to run the 'yum-arch' tool against the RPMs on the HTTP distribution site. This tool is now deprecated in favour of the 'createrepo' tool, so this module is only provided for backwards compatability with old YUM versions.
Whether to generate an APT index of RPMs / Debian packages on the HTTP distribution site.
Whether to send email alerts at the end of each build cycle. The alerts contain details of which modules successfully built, or failed to build.
The parameters for email alerts are grouped within a nested configuration block with the name alert.
For example
alert = {
...parameters..
}
The scope can be either builder in which case a single email is sent containing details of all modules, or it can be module in which case a separate email is sent for each module. This latter option is useful in the case where a single builder instance is building modules maintained by a number of distinct development teams.
The trigger defines the conditions under which an alert will be generated. If the trigger is always, alerts will be sent on every build cycle. If the trigger is fail, alerts will be sent whenever the a module fails to build. If the trigger is first-fail, alerts will be sent the first time a module fails, but no further alerts will be sent until the module has been fixed. This latter option prevents a single coding error, generating a long series of duplicated build alerts.
This specifies the recipient of the email alerts. It can either be an explicit email address, or one of two magic addresses. The first magic address is admin which is substituted by the value of the adminEmail configuration parameter. The second magic address is group which is substituted by the value of the groupEmail configuration parameter.
This specifies the address to use in the From: field of the email.
If the build host machine is not running an SMTP relay process on port 25, then this parameter is required to specify the hostname of the STMP sever to relay alerts via.
If the cCache feature is set to 1, then various files and pieces of metadata are cached across build cycles. Since disk space is a finite resource it is neccessary to limit the size of the cache. The suitable size will depend on the size of particular modules being built. The cache should be sized such that the output of at least 1 complete cycle be saved.
The maximum size of the cache, can be specified in megabytes by postfixing the letter M, or in gigabytes by postfixing the letter G.
The maximum age sets a limit on how long a cache will be considered valid, before a module is forced to be re-built from scratch.
The nested env configuration block provides the ability to specify arbitrary environment variables which will be set whenever the build engine executes a command.
env = {
ENV-NAME = ENV-VALUE
...
}
For large development projects, multiple projects will be checked out for building from a single source repository. Thus, the parameters required to access particular source repositories are defined separately to the modules themselves. The definition is within the nested repositories block.
repositories = {
repo-name = {
...repository params...
}
}
The available repository configuration parameters are:
The arbitrary descriptive name for the repository, used to identify repositories in the HTML status pages.
Specifies the source control tool used to access files within the repository. There are many source control tools supported
The CVS repository type requires that the command cvs be present in the $PATH. The CVSROOT environment variable should be set in the env configuration block to identify the location of the repository. For remote repositories it may also be neccessary to specify the CVS_RSH environment variable. For example
env = {
CVSROOT = :pserver:anonymous@cvs.gna.org:/cvs/testautobuild
}
The Perforce repository type requires that the command p4 be present in the $PATH. The P4CONFIG environment variable should to set to point to a client view. The paths in the view spec will be filled in automatically by the build engine. For example
env = {
P4CONFIG = /var/lib/builder/.p4config
}
The GNU Arch repository type requires that the command tla be present in the $PATH. The 'archive-name' and 'archive-uri' options should be set to identify the remote repository. For example
options = {
archive-name = lord@emf.net--2004
archive-uri = http://arch.quackerhead.com/~lord/archives/lord@emf.net--2004
}
The Mercurial repository type requires that the command hg be present in the $PATH. Additionally, it must be at least version 0.7 or later. The 'base-url' option should be set to identify the base URL under which the repositories are located. For example
options = {
base-url = http://www.selenic.com/
}
The GIT repository type requires that the command git be present in the $PATH. Additionally, it must be at least version 1.5.0.0 or later. The 'base-url' option should be set to identify the base URL under which the repositories are located. For example
options = {
base-url = git://anongit.freedesktop.org/git/
}
The Bazaar repository type requires that the command gbzr be present in the $PATH. Additionally, it must be at least version 0.91 or later. The 'base-url' option should be set to identify the base URL under which the repositories are located. For example
options = {
base-url = http://www.darcs.net/repos/
}
The Monotone repository type requires that the command mtn be present in the $PATH. Additionally, it must be at least version 0.37 or later. The 'server' option should be set to identify the hostname & port number at which the repository is located. For example
options = {
server = guitone.thomaskeller.biz
}
The module path will simply be the name of the branch to checkout
The Darcs repository type requires that the command darcs be present in the $PATH. Additionally, it must be at least version 1.0.0 or later. The 'base-url' option should be set to identify the base URL under which the repositories are located. For example
options = {
base-url = http://www.darcs.net/repos/
}
The Subversion repository type requires that the command svn be present in the $PATH. The 'url' option should be set to identify the base URL at which the repository is accessed.
options = {
url = http://aplaws.redhat.com/svn/aplaws/
}
The SVK repository type requires that the command svk be present in the $PATH. It (currently) requires no special options to be set, rather expecting the modules to specify their full checkout URL. This will, however, likely change in the future.
The local disk repository type bypasses all version control, allowing source files to be simply copied straight from another directory. The 'directory' otion to refer to the base directory containing the source files.
options = {
directory = /var/lib/builder/local-files
}
The nested options configuration block is used to set parameters which are specific to each repository type. Refer to the description of the type parameter for details of what options are valid for a particular type.
The nested env configuration block is used to set environment variables which are specific to each repository type. Refer to the description of the type parameter for details of what variables are relevant for a particular type.
The modules block is where the applications / tools / libraries to be built are defined. The definitions for each module are within the nested modules block:
modules = {
module-name = {
...module params...
}
}
This defines a free text display name for the module, and will be used in email alerts, web status pages & related areas.
Controls whether builds can be cached, if no SCM changes were detected. Defaults to 1, if set to 0, then the module will be built on every build cycle, even if no SCM changes were detected.
Use to temporarily hide a module from the build. If set to 0, the effect will be just as if the module was never included in the config file.
Define the email address of a module-specific build administrator. If this is omitted, it defaults to the global administrator email address.
Define a name for the module-specific administrator. As with the admin-email, if this is omitted, the global administrator name is used as a default.
Define the email address of a mailing list for the module specific development team. If this is omitted, it defaults to the global developers email address.
Define a name for the module-specific development team. If this is omitted it defaults to the global development team name.
The nested source block defines the location from which the source code for the module will be checked out.
source = {
... source location options ...
}
There are two entries required to be presented in the source block:
This defines the name of the repository from which the source will be checked out. This repository must be defined earlier in the configuration file.
This defines the path from which to check out the module. The syntax for the path varies depending on the type of repository, and will typically be specified relative to the root of the repository.
The nested env configuration block provides the ability to specify arbitrary environment variables which will be set whenever the build engine executes a command for this module.
env = {
ENV-NAME = ENV-VALUE
...
}
The nested links configuration block defines a list of 'interesting' hyperlinks pertaining to this module. No functional interpretation is placed on the links, they are simply passed through to the HTML template engine when generating the status pages. It is common to provide links to the project's homepage, and a URL for browsing the source code online.
links = (
{
...first link...
}
{
... second link...
}
)
Within the block for each link, there are two required entries:
The nested groups configuration block defines the list of groups of which the module is a member. The entries in the list are simply names of groups defined earlier in the configuration block. The group membership is primarily used to control generation of the HTML pages.
groups = (
group-name-a
group-name-b
...
)
The nested artifacts configuration block defines a list of build artifacts to publish to the distribution servers. A build artifact can be anything which is generated during the course of a build, test logs, documentation files, code coverage reports. Each entry in the artifacts list is a block with the following keys
The path to the source files for the artifact, relative to the module's build root. The source can be a file, a directory, or a BSD glob.
The destination path at which to publish the artifacts, if the source is a file, this parameter will be treated as a filename. If the source is a glob, or directory, this parameter will be treated as the base directory.
An optional parameter providing an alternate link to the published artifacts. This is useful if the web server does not allow directory browsing, and there is no index.html file.
The label to use when displaying a link to the artifacts.
XXX
engine = /etc/auto-build.d/engine/main.conf
templateDir = /etc/auto-build.d/templates
buildRoot = /var/lib/builder
label = Continous Automatic Builder
adminEmail = admin@example.com
adminName = Build Administrator
groupEmail = dev@example.com
groupName = Build Developers
hostname = example.com
httpPrefix = /~builder
lockMethod = fcntl
features = {
cache = 1
checkout = 1
createrepo_index = 1
yum_index = 1
apt_index = 1
email_alerts = 1
}
alert = {
scope = builder
trigger = first-fail
to = group
from = builder@example.com
smtpServer = mail.example.com
}
maxCacheSize = 100M
maxCacheAge = 7d
groups = {
software = {
label = Software
}
autobuild = {
label = Test-AutoBuild
}
docs = {
label = Documentation
}
}
env = {
USER = builder
}
repositories = {
gna-cvs = {
label = GNA Anonymous CVS Server
type = cvs
env = {
CVSROOT = :pserver:anonymous@cvs.gna.org:/cvs/testautobuild
}
}
}
modules = {
autobuild-dev = {
label = Test AutoBuild Unstable
admin-email = autobuild-admin@example.com
admin-name = Module Administrator
group-email = autobuild-dev@example.com
group-name = Module Developers
source = {
repository = gna-cvs
path = testautobuild
}
env = {
HTMLURLPREFIX = /~builder/artifacts/autobuild-dev/apidocs
}
groups = (
autobuild
software
)
links = (
{
href = http://www.autobuild.org/
label = Homepage
}
{
href = http://cvs.gna.org/viewcvs/testautobuild/testautobuild/
label = Browse CVS Repository
}
)
artifacts = (
{
src = README
dst = README
label = Readme
}
{
src = blib/coverage/*
dst = coverage/
label = Code Test & POD coverage Reports
}
{
src = blib/html/*
dst = apidocs/
path = apidocs/Test/AutoBuild.html
label = Module API documentation
}
)
}
}
isos = {
autobuild-unstable = {
name = autobuild-unstable.iso
label = Test-AutoBuild Unstable
packageTypes = (
rpm
zip
)
modules = (
autobuild-dev
)
}
}
/etc/auto-build.d/templates/* - templates for HTML status pages & email alerts /etc/auto-build.d/engine/*.conf - definitions for build engine runtime setup /etc/auto-build.d/httpd/*.conf - example apache configuration files /etc/auto-build.d/cron/*.conf - example crontab entries for automating builder /var/lib/builder - default build working directory
auto-build(1), Config::Record(3pm), Template(3pm)
Daniel P. Berrange <dan@berrange.com>
