Build Scripts

Some packages need to compile third-party non-Rust code, for example C libraries. Other packages need to link to C libraries which can either be located on the system or possibly need to be built from source. Others still need facilities for functionality such as code generation before building (think parser generators).

Cargo does not aim to replace other tools that are well-optimized for these tasks, but it does integrate with them with custom build scripts. Placing a file named build.rs in the root of a package will cause Cargo to compile that script and execute it just before building the package.

// Example custom build script.
fn main() {
    // Tell Cargo that if the given file changes, to rerun this build script.
    println!("cargo::rerun-if-changed=src/hello.c");
    // Use the `cc` crate to build a C file and statically link it.
    cc::Build::new()
        .file("src/hello.c")
        .compile("hello");
}

Some example use cases of build scripts are:

  • Building a bundled C library.
  • Finding a C library on the host system.
  • Generating a Rust module from a specification.
  • Performing any platform-specific configuration needed for the crate.

The sections below describe how build scripts work, and the examples chapter shows a variety of examples on how to write scripts.

Note: The package.build manifest key can be used to change the name of the build script, or disable it entirely.

Life Cycle of a Build Script

Just before a package is built, Cargo will compile a build script into an executable (if it has not already been built). It will then run the script, which may perform any number of tasks. The script may communicate with Cargo by printing specially formatted commands prefixed with cargo:: to stdout.

The build script will be rebuilt if any of its source files or dependencies change.

By default, Cargo will re-run the build script if any of the files in the package changes. Typically it is best to use the rerun-if commands, described in the change detection section below, to narrow the focus of what triggers a build script to run again.

Once the build script successfully finishes executing, the rest of the package will be compiled. Scripts should exit with a non-zero exit code to halt the build if there is an error, in which case the build script’s output will be displayed on the terminal.

Inputs to the Build Script

When the build script is run, there are a number of inputs to the build script, all passed in the form of environment variables.

In addition to environment variables, the build script’s current directory is the source directory of the build script’s package.

Outputs of the Build Script

Build scripts may save any output files or intermediate artifacts in the directory specified in the OUT_DIR environment variable. Scripts should not modify any files outside of that directory.

Build scripts communicate with Cargo by printing to stdout. Cargo will interpret each line that starts with cargo:: as an instruction that will influence compilation of the package. All other lines are ignored.

The order of cargo:: instructions printed by the build script may affect the order of arguments that cargo passes to rustc. In turn, the order of arguments passed to rustc may affect the order of arguments passed to the linker. Therefore, you will want to pay attention to the order of the build script’s instructions. For example, if object foo needs to link against library bar, you may want to make sure that library bar’s cargo::rustc-link-lib instruction appears after instructions to link object foo.

The output of the script is hidden from the terminal during normal compilation. If you would like to see the output directly in your terminal, invoke Cargo as “very verbose” with the -vv flag. This only happens when the build script is run. If Cargo determines nothing has changed, it will not re-run the script, see change detection below for more.

All the lines printed to stdout by a build script are written to a file like target/debug/build/<pkg>/output (the precise location may depend on your configuration). The stderr output is also saved in that same directory.

The following is a summary of the instructions that Cargo recognizes, with each one detailed below.

MSRV: 1.77 is required for cargo::KEY=VALUE syntax. To support older versions, use the cargo:KEY=VALUE syntax.

The rustc-link-arg instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building supported targets (benchmarks, binaries, cdylib crates, examples, and tests). Its usage is highly platform specific. It is useful to set the shared library version or linker script.

The rustc-link-arg-bin instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building the binary target with name BIN. Its usage is highly platform specific. It is useful to set a linker script or other linker options.

The rustc-link-arg-bins instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building a binary target. Its usage is highly platform specific. It is useful to set a linker script or other linker options.

The rustc-link-lib instruction tells Cargo to link the given library using the compiler’s -l flag. This is typically used to link a native library using FFI.

The LIB string is passed directly to rustc, so it supports any syntax that -l does.
Currently the fully supported syntax for LIB is [KIND[:MODIFIERS]=]NAME[:RENAME].

The -l flag is only passed to the library target of the package, unless there is no library target, in which case it is passed to all targets. This is done because all other targets have an implicit dependency on the library target, and the given library to link should only be included once. This means that if a package has both a library and a binary target, the library has access to the symbols from the given lib, and the binary should access them through the library target’s public API.

The optional KIND may be one of dylib, static, or framework. See the rustc book for more detail.

The rustc-link-arg-tests instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building a tests target.

The rustc-link-arg-examples instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building an examples target.

The rustc-link-arg-benches instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building a benchmark target.

The rustc-link-search instruction tells Cargo to pass the -L flag to the compiler to add a directory to the library search path.

The optional KIND may be one of dependency, crate, native, framework, or all. See the rustc book for more detail.

These paths are also added to the dynamic library search path environment variable if they are within the OUT_DIR. Depending on this behavior is discouraged since this makes it difficult to use the resulting binary. In general, it is best to avoid creating dynamic libraries in a build script (using existing system libraries is fine).

cargo::rustc-flags=FLAGS

The rustc-flags instruction tells Cargo to pass the given space-separated flags to the compiler. This only allows the -l and -L flags, and is equivalent to using rustc-link-lib and rustc-link-search.

cargo::rustc-cfg=KEY[="VALUE"]

The rustc-cfg instruction tells Cargo to pass the given value to the --cfg flag to the compiler. This may be used for compile-time detection of features to enable conditional compilation. Custom cfgs must either be expected using the cargo::rustc-check-cfg instruction or usage will need to allow the unexpected_cfgs lint to avoid unexpected cfgs warnings.

Note that this does not affect Cargo’s dependency resolution. This cannot be used to enable an optional dependency, or enable other Cargo features.

Be aware that Cargo features use the form feature="foo". cfg values passed with this flag are not restricted to that form, and may provide just a single identifier, or any arbitrary key/value pair. For example, emitting cargo::rustc-cfg=abc will then allow code to use #[cfg(abc)] (note the lack of feature=). Or an arbitrary key/value pair may be used with an = symbol like cargo::rustc-cfg=my_component="foo". The key should be a Rust identifier, the value should be a string.

cargo::rustc-check-cfg=CHECK_CFG

Add to the list of expected config names and values that is used when checking the reachable cfg expressions with the unexpected_cfgs lint.

The syntax of CHECK_CFG mirrors the rustc --check-cfg flag, see Checking conditional configurations for more details.

The instruction can be used like this:

#![allow(unused)]
fn main() {
// build.rs
println!("cargo::rustc-check-cfg=cfg(foo, values(\"bar\"))");
if foo_bar_condition {
    println!("cargo::rustc-cfg=foo=\"bar\"");
}
}

Note that all possible cfgs should be defined, regardless of which cfgs are currently enabled. This includes all possible values of a given cfg name.

It is recommended to group the cargo::rustc-check-cfg and cargo::rustc-cfg instructions as closely as possible in order to avoid typos, missing check-cfg, stale cfgs…

See also the conditional compilation example.

MSRV: Respected as of 1.80

cargo::rustc-env=VAR=VALUE

The rustc-env instruction tells Cargo to set the given environment variable when compiling the package. The value can be then retrieved by the env! macro in the compiled crate. This is useful for embedding additional metadata in crate’s code, such as the hash of git HEAD or the unique identifier of a continuous integration server.

See also the environment variables automatically included by Cargo.

Note: These environment variables are also set when running an executable with cargo run or cargo test. However, this usage is discouraged since it ties the executable to Cargo’s execution environment. Normally, these environment variables should only be checked at compile-time with the env! macro.

The rustc-cdylib-link-arg instruction tells Cargo to pass the -C link-arg=FLAG option to the compiler, but only when building a cdylib library target. Its usage is highly platform specific. It is useful to set the shared library version or the runtime-path.

cargo::error=MESSAGE

The error instruction tells Cargo to display an error after the build script has finished running, and then fail the build.

Note: Build script libraries should carefully consider if they want to use cargo::error versus returning a Result. It may be better to return a Result, and allow the caller to decide if the error is fatal or not. The caller can then decide whether or not to display the Err variant using cargo::error.

MSRV: Respected as of 1.84

cargo::warning=MESSAGE

The warning instruction tells Cargo to display a warning after the build script has finished running. Warnings are only shown for path dependencies (that is, those you’re working on locally), so for example warnings printed out in crates.io crates are not emitted by default, unless the build fails. The -vv “very verbose” flag may be used to have Cargo display warnings for all crates.

Build Dependencies

Build scripts are also allowed to have dependencies on other Cargo-based crates. Dependencies are declared through the build-dependencies section of the manifest.

[build-dependencies]
cc = "1.0.46"

The build script does not have access to the dependencies listed in the dependencies or dev-dependencies section (they’re not built yet!). Also, build dependencies are not available to the package itself unless also explicitly added in the [dependencies] table.

It is recommended to carefully consider each dependency you add, weighing against the impact on compile time, licensing, maintenance, etc. Cargo will attempt to reuse a dependency if it is shared between build dependencies and normal dependencies. However, this is not always possible, for example when cross-compiling, so keep that in consideration of the impact on compile time.

Change Detection

When rebuilding a package, Cargo does not necessarily know if the build script needs to be run again. By default, it takes a conservative approach of always re-running the build script if any file within the package is changed (or the list of files controlled by the exclude and include fields). For most cases, this is not a good choice, so it is recommended that every build script emit at least one of the rerun-if instructions (described below). If these are emitted, then Cargo will only re-run the script if the given value has changed. If Cargo is re-running the build scripts of your own crate or a dependency and you don’t know why, see “Why is Cargo rebuilding my code?” in the FAQ.

cargo::rerun-if-changed=PATH

The rerun-if-changed instruction tells Cargo to re-run the build script if the file at the given path has changed. Currently, Cargo only uses the filesystem last-modified “mtime” timestamp to determine if the file has changed. It compares against an internal cached timestamp of when the build script last ran.

If the path points to a directory, it will scan the entire directory for any modifications.

If the build script inherently does not need to re-run under any circumstance, then emitting cargo::rerun-if-changed=build.rs is a simple way to prevent it from being re-run (otherwise, the default if no rerun-if instructions are emitted is to scan the entire package directory for changes). Cargo automatically handles whether or not the script itself needs to be recompiled, and of course the script will be re-run after it has been recompiled. Otherwise, specifying build.rs is redundant and unnecessary.

cargo::rerun-if-env-changed=NAME

The rerun-if-env-changed instruction tells Cargo to re-run the build script if the value of an environment variable of the given name has changed.

Note that the environment variables here are intended for global environment variables like CC and such, it is not possible to use this for environment variables like TARGET that Cargo sets for build scripts. The environment variables in use are those received by cargo invocations, not those received by the executable of the build script.

The package.links key may be set in the Cargo.toml manifest to declare that the package links with the given native library. The purpose of this manifest key is to give Cargo an understanding about the set of native dependencies that a package has, as well as providing a principled system of passing metadata between package build scripts.

[package]
# ...
links = "foo"

This manifest states that the package links to the libfoo native library. When using the links key, the package must have a build script, and the build script should use the rustc-link-lib instruction to link the library.

Primarily, Cargo requires that there is at most one package per links value. In other words, it is forbidden to have two packages link to the same native library. This helps prevent duplicate symbols between crates. Note, however, that there are conventions in place to alleviate this.

Build scripts can generate an arbitrary set of metadata in the form of key-value pairs. This metadata is set with the cargo::metadata=KEY=VALUE instruction.

The metadata is passed to the build scripts of dependent packages. For example, if the package bar depends on foo, then if foo generates key=value as part of its build script metadata, then the build script of bar will have the environment variables DEP_FOO_KEY=value. See the “Using another sys crate” for an example of how this can be used.

Note that metadata is only passed to immediate dependents, not transitive dependents.

MSRV: 1.77 is required for cargo::metadata=KEY=VALUE. To support older versions, use cargo:KEY=VAUE (unsupported directives are assumed to be metadata keys).

*-sys Packages

Some Cargo packages that link to system libraries have a naming convention of having a -sys suffix. Any package named foo-sys should provide two major pieces of functionality:

  • The library crate should link to the native library libfoo. This will often probe the current system for libfoo before resorting to building from source.
  • The library crate should provide declarations for types and functions in libfoo, but not higher-level abstractions.

The set of *-sys packages provides a common set of dependencies for linking to native libraries. There are a number of benefits earned from having this convention of native-library-related packages:

  • Common dependencies on foo-sys alleviates the rule about one package per value of links.
  • Other -sys packages can take advantage of the DEP_NAME_KEY=value environment variables to better integrate with other packages. See the “Using another sys crate” example.
  • A common dependency allows centralizing logic on discovering libfoo itself (or building it from source).
  • These dependencies are easily overridable.

It is common to have a companion package without the -sys suffix that provides a safe, high-level abstractions on top of the sys package. For example, the git2 crate provides a high-level interface to the libgit2-sys crate.

Overriding Build Scripts

If a manifest contains a links key, then Cargo supports overriding the build script specified with a custom library. The purpose of this functionality is to prevent running the build script in question altogether and instead supply the metadata ahead of time.

To override a build script, place the following configuration in any acceptable config.toml file.

[target.x86_64-unknown-linux-gnu.foo]
rustc-link-lib = ["foo"]
rustc-link-search = ["/path/to/foo"]
rustc-flags = "-L /some/path"
rustc-cfg = ['key="value"']
rustc-env = {key = "value"}
rustc-cdylib-link-arg = ["…"]
metadata_key1 = "value"
metadata_key2 = "value"

With this configuration, if a package declares that it links to foo then the build script will not be compiled or run, and the metadata specified will be used instead.

The warning, rerun-if-changed, and rerun-if-env-changed keys should not be used and will be ignored.

Jobserver

Cargo and rustc use the jobserver protocol, developed for GNU make, to coordinate concurrency across processes. It is essentially a semaphore that controls the number of jobs running concurrently. The concurrency may be set with the --jobs flag, which defaults to the number of logical CPUs.

Each build script inherits one job slot from Cargo, and should endeavor to only use one CPU while it runs. If the script wants to use more CPUs in parallel, it should use the jobserver crate to coordinate with Cargo.

As an example, the cc crate may enable the optional parallel feature which will use the jobserver protocol to attempt to build multiple C files at the same time.