Welcome! If you are reading this document, it means you are interested in contributing
to the libc
crate.
libc
has two active branches: main
and libc-0.2
. main
is for active
development of the upcoming v1.0 release, and should be the target of all pull
requests. libc-0.2
is for updates to the currently released version.
If a pull request to main
is a good candidate for inclusion in an 0.2.x
release, include @rustbot label stable-nominated
in a comment to propose this.
Good candidates will usually meet the following:
- The included changes are non-breaking.
- The change applies cleanly to both branches.
- There is a usecase that justifies inclusion in a stable release (all additions should always have a usecase, hopefully).
Once a stable-nominated
PR targeting main
has merged, it can be cherry
picked to the libc-0.2
branch. A maintainer will likely do these cherry
picks in a batch.
Alternatively, you can start this process yourself by creating a new branch
based on libc-0.2
and running git cherry-pick -xe commit-sha-on-main
(git cherry-pick -xe start-sha^..end-sha
if a range of commits is needed). git
will automatically add the "cherry picked from commit" note, but try to add a
backport note so the original PR gets crosslinked:
# ... original commit message ...
(backport <https://github.com/rust-lang/libc/pull/1234>) # add manually
(cherry picked from commit 104b6a4ae31c726814c36318dc718470cc96e167) # added by git
Once the cherry-pick is complete, open a PR targeting libc-0.2
.
See the tracking issue for details.
Want to use an API which currently isn't bound in libc
? It's quite easy to add
one!
The internal structure of this crate is designed to minimize the number of
#[cfg]
attributes in order to easily be able to add new items which apply
to all platforms in the future. As a result, the crate is organized
hierarchically based on platform. Each module has a number of #[cfg]
'd
children, but only one is ever actually compiled. Each module then reexports all
the contents of its children.
This means that for each platform that libc supports, the path from a
leaf module to the root will contain all bindings for the platform in question.
Consequently, this indicates where an API should be added! Adding an API at a
particular level in the hierarchy means that it is supported on all the child
platforms of that level. For example, when adding a Unix API it should be added
to src/unix/mod.rs
, but when adding a Linux-only API it should be added to
src/unix/linux_like/linux/mod.rs
.
If you're not 100% sure at what level of the hierarchy an API should be added at, fear not! This crate has CI support which tests any binding against all platforms supported, so you'll see failures if an API is added at the wrong level or has different signatures across platforms.
New symbol(s) (i.e. functions, constants etc.) should also be added to the
symbols list(s) found in the libc-test/semver
directory. These lists keep
track of what symbols are public in the libc crate and ensures they remain
available between changes to the crate. If the new symbol(s) are available on
all supported Unixes it should be added to unix.txt
list1,
otherwise they should be added to the OS specific list(s).
With that in mind, the steps for adding a new API are:
- Determine where in the module hierarchy your API should be added.
- Add the API, including adding new symbol(s) to the semver lists.
- Send a PR to this repo.
- Wait for CI to pass, fixing errors.
- Wait for a merge!
1: Note that this list has nothing to do with any Unix or Posix
standard, it's just a list shared among all OSs that declare #[cfg(unix)]
.
We have two automated tests running on GitHub Actions:
cd libc-test && cargo test
- Use the
skip_*()
functions inbuild.rs
if you really need a workaround.
- Style checker
Sometimes an upstream adds a breaking change to their API e.g. removing outdated items,
changing the type signature, etc. And we probably should follow that change to build the
libc
crate successfully. It's annoying to do the equivalent of semver-major versioning
for each such change. Instead, we mark the item as deprecated and do the actual change
after a certain period. The steps are:
- Add
#[deprecated(since = "", note="")]
attribute to the item.
- The
since
field should have a next version oflibc
(e.g., if the current version is0.2.1
, it should be0.2.2
). - The
note
field should have a reason to deprecate and a tracking issue to call for comments (e.g., "We consider removing this as the upstream removed it. If you're using it, please comment on #XXX").
- If we don't see any concerns for a while, do the change actually.
When Rust removes a support for a target, the libc crate also may remove the support anytime.
Now that you've done the amazing job of landing your new API or your new platform in this crate, the next step is to get that sweet, sweet usage from crates.io! The only next step is to bump the version of libc and then publish it. If you'd like to get a release out ASAP you can follow these steps:
- Increment the patch version number in
Cargo.toml
andlibc-test/Cargo.toml
. - Send a PR to this repository. It should look like this, but it'd also be nice to fill out the description with a small rationale for the release (any rationale is ok though!).
- Once merged, the release will be tagged and published by one of the libc crate maintainers.