GithubHelp home page GithubHelp logo

alexxnica / deb-ostree-builder Goto Github PK

View Code? Open in Web Editor NEW

This project forked from dbnicholson/deb-ostree-builder

0.0 1.0 0.0 41 KB

Stripped down Endless ostree builder for debian

Shell 36.76% Python 63.24%

deb-ostree-builder's Introduction

Endless OSTree Builder (EOB)

This program assembles the Endless OS (EOS) from prebuilt packages and content. Its main functions are:

  1. Assemble packages into ostree
  2. Publish the ostree repo to a remote server

Design

EOB is designed to be simple. It is written in bash script and has just enough flexibility to meet our needs. The simplicity allows us to have a complete in-house understanding of the build system, enabling smooth organic growth as our requirements evolve. The build master(s) who maintain this system are not afraid of encoding our requirements in simple bash script.

When added complexity is minimal, we prefer calling into lower level tools directly rather than utilizing abstraction layers (e.g. we call debootstrap instead of using live-tools). This means we have a thorough understanding of the build system and helps to achieve our secondary goals of speed and flexibility.

The build process is divided into several stages, detailed below. An invocation can run some or all of these stages.

check_update stage

This stage does not perform ostree building, but is used to determine if an ostree build is required. If it exits successfully, no ostree build is needed.

OS stage

This stage creates the OS in a clean directory tree, populating it with apt packages.

ostree stage

This stage makes appropriate modifications to the output of the previous stage and commits it to a locally stored ostree repository. The ostree repository is created if it does not already exist.

publish stage

This stage publishes the local ostree repository to the remote ostree server.

error stage

This stage is only run in the event of an error and simply cleans up for a subsequent build.

Setup

Known to work on Debian Wheezy, Ubuntu 13.04 and Ubuntu 13.10. Required packages:

  • rsync
  • ostree
  • python3-apt
  • attr
  • x86: grub2
  • arm: mkimage, device-tree-compiler

ostree signing

EOB signs the ostree commits it makes with GPG. A private keyring must be installed in /etc/deb-ostree-builder/gnupg and the key ID must be set in the configuration.

Configuration

The ostree builder configuration is built up from a series of INI files. The configuration files are stored in the config directory of the builder. The order of configuration files read in is:

  • Default settings - config/defaults.ini
  • Product settings - config/product/$product.ini
  • Branch settings - config/branch/$branch.ini
  • Architecture settings - config/arch/$arch.ini
  • Platform settings - config/platform/$platform.ini
  • System config settings - /etc/deb-ostree-builder/config.ini
  • Local build settings - config/local.ini

None of these files are required to be present, but the defaults.ini file contains many settings that are expected throughout the core of the build.

New configuration options should be added and documented in defaults.ini. See the existing file for options that are available to customize. Settings in the default build section are usually set in the OSTreeBuilder class as they're static across all builds.

Format

The format of the configuration files is INI as mentioned above. However, a form of interpolation is used to allow referring to other options. For instance, an option foo can use the value from an option bar by using ${bar} in its value. If bar was in a different section, it can be referred to by prepending the other section in the form of ${other:bar}. The build section is the default section. Any interpolation without an explicit section can fallback to a value in the build section. For example, if bar doesn't exist in the current section, it will also be looked for in the build section.

The INI file parsing is done using the configparser python module. The interpolation feature is provided by its ExtendedInterpolation class. See the python documentation for a more detailed discussion of this feature.

The system and local configuration files are not typically used. They can allow for a permanent or temporary override for a particular host or build.

Merged options

In some cases, an option needs to represent a set of values rather than a single setting. Adding or removing items from the list is not possible with the features in the configuration parser.

To allow some method of building these lists, the builder will take multiple options of the form $prefix_add_* and $prefix_del_* and merge them together into one option named $prefix. Values in the various $prefix_add_* options are added to a set, and then values in the various $prefix_del_* options are removed from the set. If the option $prefix already exists, it is not changed. This allows a configuration file to override all of the various add and del options from other files to provide the list exactly in the form it wants.

The currently supported merged options are:

  • cache:hooks
  • seed:hooks
  • os:hooks
  • os:packages
  • ostree:extra_refs
  • publish:hooks
  • error:hooks

See the defaults.ini file for a description of these options.

Accessing options

The build core accesses these settings via environment variables. The variables take the form of EOB_$SECTION_$OPTION. The build section is special and these settings are exported in the form EOB_$OPTION without the section in the variable name.

Execution

To run EOB, use the deb-ostree-builder script, optionally with a branch name:

./deb-ostree-builder [options] master

If no branch name is specified, master is used. If you want to only run certain stages, modify the buildscript file accordingly before starting the program.

Options available: --product : specify product to build (debian, debianplatform, debiansdk) --arch : specify architecture to build (i386, armhf) --platform : specify a sub-architecture to build --force : perform a build even if the update check says it's not needed --dry-run : perform a build, but do not publish the results

Customization

The core of EOB is just a wrapper. The real content of the output is defined by customization scripts found under hooks/. These scripts have access to environment variables and library functions allowing them to integrate correctly with the core.

The scripts to run are organized under hooks/GROUP where GROUP is a group of hooks run by a particular stage. The hooks to run are managed in the configuration with merged hooks keys under each group. For instance, the os group hooks to run are defined in the os:hooks configuration key. This allows easy customization for different OS variants. These are merged options as described above, so they can be added to or pruned by specific products.

If a script has an executable bit, it is executed directly. Otherwise it is executed through bash and will have access to the library functions.

If a script filename finishes with ".chroot" then it is executed within the chroot environment, as if it is running on the final system. Otherwise, the script is executed under the regular host environment. It is preferred to avoid chrooted scripts when it is easy to run the operation outside of the chroot environment.

Scripts are executed in lexical order and the convention is to prefix them with a two-digit number to make the order explicit. Each script should be succinct - we prefer to have a decent number of small-ish scripts, rather than having a small number of huge bash rambles.

check_update customization

The check_update stage calls the cache customization hooks. The intention is to determine facts about the current build and compare them to cached facts from the previous build. Facts are stored in the build specific cache directory, determined from the function eob_cachedir. Cache files should be named using the eob_cachefile function.

The check_update stage determines if an update is needed by seeing if the modification times for any files in the cache directory have been updated. Therefore, the hook should only update its cache file if there's a difference from the previous build.

os customization

At the start of the os stage, the customization hooks under seed are run. At this stage the ${EOB_ROOTDIR} is totally empty. Place anything here that you want to be present at the time of initial bootstrap, which follows.

After the initial bootstrap, the customization hooks under os are run. These scripts are responsible for making any configuration changes to the system (discouraged), installing packages, etc. The OS packages are installed by scripts at index 50.

ostree customization

The ostree stage currently has no customization hooks.

publish customization

Keeping with the design that the core is simple and the meat is kept under customization, the publish stage does nothing more than call into customization hooks kept in publish. This stage should publish the local ostree repo to the remote server.

error customization

Like the publish stage, the error stage simply calls the customization hooks kept in error. These hooks should clean up for subsequent builds.

deb-ostree-builder's People

Contributors

dbnicholson avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.