README.md

Workspace tool

ws is a lightweight tool for managing a collection of code repositories. It is intended to handle coarse dependencies between projects, building multiple projects in the right order and passing the correct flags to each (such as PKG_CONFIG_PATH and similar). It is not intended to be a full-fledged build system, but instead should merely build each project in the right order and with the right environment glue. Everything it does can be done by-hand, so rather than a replacing existing build tools, it merely automates the tedious task of manually specifying --prefix, setting env vars, and rebuilding projects in the right order when they change.

Inside each workspace, projects are installed inside a localized install directory so that no installations ever leak out of a workspace. Projects dependent on other projects automatically link to the other projects' install directories by setting PKG_CONFIG_PATH, LD_LIBRARY_PATH, and everything else necessary.

Note that ws does not directly handle source code syncing. That job is left to repo and similar tools.

Dependencies

ws depends on the Python 3 PyYAML, which you can get either with sudo apt install python3-yaml or via pip3 install -r requirements.txt from the top of the repository.

Installing

To install ws, you can use the setup.py script at the top level of the repository: python3 setup.py install <add any custom options here>. You can also use pip: pip3 install . from the top of the repository. Finally, if you want the installed ws to directly symlink into your source directory instead of being a one-time copy of the code, use pip3 install -e ., which activates pip "developer mode". This way, code changes immediately take effect without re-running the install step.

ws

The ws script is the main point of interaction with your workspaces. It assumes you have already synced a bunch of code using repo or some other tool and, unless you use special options, it assumes you are currently somewhere inside the root of the source that ws manages. However, you can be anywhere inside that tree and do not have to be at the top of it.

The normal workflow for ws is as follows:

repo init -u MANIFEST-REPO-URL
repo sync
ws init -s repo
ws build

By default, ws init will look for a file called ws-manifest.yaml at the root of the repository containing the git-repo manifest (the one we passed -u into when we called repo init). This file contains dependency and build system information for the projects that ws manages. Note that ws does not have to manage all the same projects that repo manages, but it can. The full format for ws-manifest.yaml is at the bottom of the README.

If you don't use the git-repo tool, you can instead pass in your own ws manifest via ws init -s fs -m. This lets you manage the manifest however you like (e.g. submodules, or manually).

bash-completion

If you like bash-completions and typing things fast, you can do:

. bash-completion/ws

And get auto-completion for ws commands.

ws init

When you run ws init, ws creates a .ws directory in the current working directory. This directory can contain multiple workspaces, but there is always a default workspace, which is the one that gets used if you don't specify an alternate workspace with the -w option. One reason to create multiple workspaces is to manage multiple build configurations, such as separate debug and release builds. However, all workspaces in the same .ws directory will still operate on the same source code (the repositories configured in ws-manifest.yaml).

If you wish to create multiple workspaces, you can use ws init with an argument to do so. For example, ws init new would create a new workspace called new. However, it would not be used by default until you run ws default new. That said, you can also use -w to operate on it (e.g. ws -w new build).

If you specify -m, you can manually point to a ws-manifest.yaml to use. By default, this is relative to a repository containing a git-repo manifest (e.g. if you have a .repo directory after running repo init, then it is relative to .repo/manifests). If you specify -s fs, then it can point anywhere on the filesystem instead.

ws default

ws default is used to change the default workspace (the one used when you don't specify a -w option).

ws build

ws build is the main command you run. If you specify no arguments, it will build every project that repo knows about. If you instead specify a project or list of projects, it will build only those, plus any dependencies of them. Additionally, ws will checksum the source code on a per-repo basis and avoid rebuilding anything that hasn't changed. The checksumming logic uses git for speed and reliability, so source managed by ws has to use git.

ws clean

ws clean cleans the specified projects, or all projects if no arguments are given. By default, it just runs the clean command for the underlying build system (meson, cmake, etc.). If you also use the -f/--force switch, it will instead remove the entire build directory instead of trusting the underlying build system.

ws env

ws env allows you to enter the build environment for a given project. If given no arguments, it gives you an interactive shell inside the build directory for the project. If given arguments, it instead runs the specified command from that directory. In both cases, it sets up the right build enviroment so build commands you might use will work correctly and you can inspect if something seems wrong.

An example use of ws env is to manually build something or to tweak the build configuration of a given project in a way that ws doesn't know how to handle.

ws test

ws test allows you to run unit tests on a project that you built. The tests are configured in the ws manifest file and can be any set of arbitrary commands. The tests will be run from the build directory of the project as if you had run ws env -b PROJECT TEST.

The cwd paremeter to the tests allows tests to run in an alternate directory that the build directory. Any of the template variables listed below can be used for this.

ws config

ws config sets either workspace-wide or per-project configuration settings. The following settings are supported:

Workspace-wide settings:

• type: debug or release. This specifies the workspace build type.

Per-project settings:

• enable: sets whether or not to build the given project. Typically you want to build everything, but you might satisfy a particular dependency from the distro, or manually build and install it outside of the workspace.
• args: sets build arguments for a particular project, which get directly passed to the builder (e.g. cmake or meson). An example would be passing -D KEY=VAL to set a preprocessor variable.

ws manifest

The ws manifest is a YAML file specifying a few things about the projects ws manages:

• What build system they use (currently supports meson, cmake, and setuptools).
• What dependencies they have on other projects managed by ws.
• Any special environment variables they need.
• Any special builder options needed (e.g. -DCMAKE_ type of options). These options are passed straight through into each build system without modification.
• Any other manifests that should be included. Include paths can be absolute or relative. If they are relative, they are interpreted relative to the parent directory of the including manifest. Directories can also be included, in which case every manifest file in the directory file is included.
• Any search paths to search for manifests listed in "include". Can be either absolute or relative. If relative, it's relative to the parent directory of this manifest.

The syntax is as follows:

include:
    - some-other-manifest.yaml
    - some-directory-of-manifests

search-path:
    - ../projects # a directory containing manifests

projects:
    some-project:
        build: meson
        deps:
            - gstreamer
            - ...
        targets:
            - docs
            - install
        env:
            GST_PLUGIN_PATH: ${LIBDIR}/gstreamer-1.0
        tests:
            - some test command here
            - some other test command here
            - cwd: ${SRCDIR}
              cmds:
                  - these commands
                  - will be run from the source directory
                  - instead of the build directory

    gstreamer:
        build: meson
        args:
            - -D gtk_doc=disabled

In this case, some-project builds with meson, and requires gstreamer and some other dependencies. In order to find gstreamer plugins, it needs GST_PLUGIN_PATH set. It uses template syntax to refer to ${LIBDIR}, which will be filled in with the library path for the project.

Here is the complete list of usable template variables:

- ${BUILDDIR}: the project build directory
- ${SRCDIR}: the project source directory (top of the project's git repository)
- ${LIBDIR}: the library path for the project (what `LD_LIBRARY_PATH` will be
  set to for the project's build environment.
- ${PREFIX}: the project's prefix (what you would pass to `--prefix`).