Config Model

A config is a yaml file that defines everything about a given configuration. This includes:

meta data (e.g. its name, description, etc)
the components that should be built
how those components are built
dependencies between components
what artifacts are produced
how to configure the fvp
how to load and run the artifacts on the fvp

A config is declarative; the user declares how things relate and how things should be done, and the tool extracts the information to decide exactly what should be done and in what order in order to complete a task. All the data is contained within the config and drives the tool. This way, shrinkwrap is highly extensible. The user specifies the config(s) that should be used when invoking shrinkwrap.

Merging Configs

A config is laid out as a hierarchical data structure, using nested dictionaries. This suits it very well to being split into partial configs that are merged together into a single, final config. This allows maximal reuse of the config fragments and improves maintainability. Each config can optionally define a set of foundational layers which it then builds upon. Furthermore, the user can optionally specify a custom overlay config on the command line. Layers are merged in order according to the following rules:

For each leaf key in the union of the hierarchical dictionaries:

If the upper value is null or not present, then the lower value is taken
If both the upper and lower values are lists, then the final value is the lower list with the upper list appended to its end.
In all other cases the upper value is taken

You can use the process command to merge configs and see the resulting output to get a better feel for how it works. See Commands.

Merging Example

lower config

people:
  Iris:
    age: 2
    likes:
      - Peppa Pig
      - Bananas

upper config

people:
  Iris:
    age: 3
    likes:
      - Peas
  James:
    age: 6
    likes:
      - FIFA

merged result

people:
  Iris:
    age: 3
    likes:
      - Peppa Pig
      - Bananas
      - Peas
  James:
    age: 6
    likes:
      - FIFA

Macros

Macros are placeholders that can be specified in various parts of a config yaml file, which are substituted (“resolved”) with information that is only known at build-time or run-time. There are specific rules about which macros can be used in which parts of the config, and about the order in which they get substituted.

Macros take the following form:

${<type>:[<name>]}

where:

type is a required namespace for the macro family

name is an optional name for the macro within its namespace. For some macro types, there are a fixed set of names. For others, the names are defined by the config itself.

You can use the process command to resolve macros and see the resulting output to get a better feel for how they work. See Commands.

Defined Macros

macro	scope	description
`${param:sourcedir}`	build.<component>.{params, prebuild, build, postbuild, artifacts}	Directory in which the component’s source code is located.
`${param:builddir}`	build.<component>.{params, prebuild, build, postbuild, artifacts}	Directory in which the component should be built, if the component’s build system supports separation of source and build trees.
`${param:configdir}`	build.<component>.{params, prebuild, build, postbuild, artifacts}	Directory containing the config store. This MUST only be used for resolving files that already exist in the store.
`${param:jobs}`	build.<component>.{params, prebuild, build, postbuild, artifacts}	Maximum number of low level parallel jobs specified on the command line. To be passed to (e.g.) make as `-j${param:jobs}`.
`${btvar:<name>}`	build.<component>.{sourcedir, builddir, repo, toolchain, params, prebuild, build, postbuild, artifacts}, run.rtvars	Build-time variables. The variable names, along with default values are declared in buildex.btvars, and the user may override the value on the command line.
`${param:join_equal}`	build.<component>.{prebuild, build, postbuild}	String containing all of the component’s parameters (from its params dictionary), concatenated as `key=value` pairs.
`${param:join_space}`	build.<component>.{prebuild, build, postbuild}	String containing all of the component’s parameters (from its params dictionary), concatenated as `key value` pairs.
`${artifact:<name>}`	build.<component>.{params, prebuild, build, postbuild, artifacts}, build.btvars	Build path of an artifact declared by another component. Usage of these macros determine the component build dependency graph.
`${artifact:<name>}`	run.rtvars	Package path of an artifact.
`${rtvar:<name>}`	run.params	Run-time variables. The variable names, along with default values are declared in run.rtvars, and the user may override the value on the command line.

Schema

Top-Level keys

The following is the set of top-level public keys that should be defined by a config. There are some additional private keys that the tool will add (and make visible as part of the process command), but these are subject to change and not documented.

key	type	description
description	string	A human-readable description of what the config contains and does. Displayed by the `inspect` command.
concrete	boolean	true if the config is intended to be directly built and run, or false if it is intended as a fragment to be included in other configs.
build	dictionary	Contains all the components to be built. The key is the component name and the value is a dictionary.
run	dictionary	Contains all the information about how to run the built artifacts on the FVP.

build section

The build section, contains a dictionary of components that must be built. The keys are the component names and the values are themselves dictionaries, each containing the component meta data.

buildex section

When the schema was originally created, we made a mistake. The components should have been under build: components:, allowing room for new build data to be added under build: without being confused for components. In order to retrofit a solution without breaking compatibility, the buildex section is created.

key	type	description
btvars	dictionary	Build-Time variables. Keys are the variable names and values are a dictionary with keys ‘type’ (which must be one of ‘path’ and ‘string’) and ‘value’ (which takes the default value). Build-Time variables can be overridden by the user at the command line.

component section

key	type	description
repo	dictionary	Specifies information about the git repo(s) that must be cloned and checked out. By default, Shrinkwrap syncs the git repo to the specified revision when building. `--no-sync` or `--no-sync-all` can be used to tell Shrinkwrap to build it in whatever state the user left it in. Not required if `sourcedir` is provided.
sourcedir	string	If specified, points to the path on disk where the source repo can be found. Useful for developer use cases where a local repo already exists.
builddir	string	If specified, the location where the component will be built. If not specified, shrinkwrap allocates its own location based on SHRINKWRAP_BUILD.
toolchain	string	Defines the toolchain to be used for compilation. Value is set as CROSS_COMPILE environment variable before invoking any prebuild/build/postbuild commands. When using the standard image with a container runtime, the options are: `aarch64-none-elf-`, `arm-none-eabi-`, `aarch64-linux-gnu-`, or `arm-linux-gnueabihf-`.
stderrfilt	bool	Optional, defaults to false. When true, and –verbose is not specified, filters stderr of the component’s build task so that only lines containing ‘error’ and ‘warning’ are output. Everything else is suppressed. Useful for EDK2 which is extremely chatty.
params	dictionary	Optional set of key:value pairs. When building most components, they require a set of parameters to be passed. By setting them out as a dictionary, it is easy to override and add to them in higher layers. See `${param:join_*}` macros.
prebuild	list	List of shell commands to be executed during component build before the `build` list.
build	list	List of shell commands to be executed during component build.
postbuild	list	List of shell commands to be executed during component build after the `build` list.
artifacts	dictionary	Set of artifacts (files and/or directories) that the component exports. Key is artifact name and value is path to built artifact. Other components can reference them with the `${artifact:<name>}` macros. Used to determine build dependencies.

run section

key	type	description
name	string	Name of the FVP binary, which must be in $PATH.
rtvars	dictionary	Run-Time variables. Keys are the variable names and values are a dictionary with keys ‘type’ (which must be one of ‘path’ and ‘string’) and ‘value’ (which takes the default value). Run-Time variables can be overridden by the user at the command line.
params	dictionary	Dictionary of parameters to be passed to the FVP. Similar to the component’s params, laying these out in a dictionary makes it easy for higher layers to override and add parameters.
prerun	list	List of shell commands to be executed before the FVP is started.
terminals	dictionary	Describes the set of UART terminals available for the FVP. key is the terminal parameter name known to the FVP (e.g. `bp.terminal_0`) See below for format of the value.

terminal section

key	type	description
friendly	string	Label to display against the terminal when muxing to stdout. An empty string disables the prefix for the output.
port_regex	string	Regex to use to find the TCP port of the terminal when parsing the FVP stdout. Must have single capture group.
type	enum-string	Terminal type. See below for options.
no_color	boolean	Optional (defaults to false, only applies to [‘stdout’, ‘stdinout’] types): If true, output from this terminal is not color-coded. If this terminal carries the interactive shell, it is advised to set this to true to prevent interferring with the shell’s escape sequences. –no-color command line option causes this to behave as if set to true.
no_escapes	bool/string	Optional (defaults to false, only applies to [‘stdout’, ‘stdinout’] types): If true, strips any escape sequences from the output stream before forwarding to the terminal. If a string, behaves as if true until the string is found in the output, which sets it to false. Useful to expunge escape sequences from EDK2 during boot.
logfile	string	Optional (defaults to none, only applies to [‘stdout’, ‘stdinout’] types): Specifies path to a log file where all output to the terminal will be duplicated.

Terminal types:

stdout: Mux output to stdout. Do not supply any input.
stdinout: Mux output to stdout. Forward stdin to its input. Max of 1 of these types allowed.
telnet: Shrinkwrap will print out a telnet command to run in a separate terminal to get a unique interactive terminal.
xterm: Shrinkwrap will automatically launch xterm to provide a unique interactive terminal. Only works when runtime=null.