RedBear-OS/local/docs/VENDORED-LIBC-FORK.md

Vendored libc Fork — Redox waitid Surface

Status: TEMPORARY. Vendored for as long as it is required; delete once upstream libc carries the Redox bindings. Location: local/sources/libc/ Base: crates.io libc 0.2.189, unmodified except src/unix/redox/mod.rs. Version label: 0.2.189+rb<branch> — Cat 2 convention. +rb is build metadata, which semver ignores when matching, so ^0.2 requirements still resolve to the fork. A -rb suffix would read as a pre-release and fail. Established: 2026-08-03.

Related: LOCAL-FORK-SUPREMACY-POLICY.md, FORK-BUMP-PATCHING-POLICY.md, local/AGENTS.md § "Local fork dependency rule".

---

Why this fork exists

relibc implements the POSIX waitid surface in full, but the libc crate's Redox bindings never exposed it — and still do not as of 0.2.189. Any crate that calls waitid therefore cannot build for x86_64-unknown-redox:

error[E0531]: cannot find unit struct, unit variant or constant
              `CLD_EXITED` in crate `libc`
error[E0531]: cannot find unit struct, unit variant or constant
              `P_ALL` in crate `libc`

That breaks nix, and through it ctrlc and rustc's bootstrap tooling — which is what blocked rust-native, and with it the whole native toolchain (gcc-native, llvm-native, rust-native) that config/redbear-full.toml requires.

This is a binding gap, not a missing implementation. Every symbol added here is already real in relibc:

Symbol	relibc source	Evidence
waitid()	src/header/sys_wait/mod.rs:153	nm libc.a → T waitid
idtype_t	src/header/sys_wait/mod.rs:17	pub type idtype_t = c_int
P_ALL / P_PID / P_PGID	src/header/sys_wait/mod.rs:20-24	0 / 1 / 2
CLD_* (6)	include/signal.h:457-467	CLD_EXITED 1 … CLD_CONTINUED 6

Values and types are taken from relibc, not from Linux. Redox's idtype_t is c_int where glibc uses an enum, and copying the Linux definitions would have produced a silently wrong ABI.

What is actually changed

One file: src/unix/redox/mod.rs. Every other target is byte-for-byte upstream 0.2.189, so host builds and all non-Redox targets are unaffected.

How it is wired

[patch.crates-io] in the rust workspace, carried durably as local/patches/rust/02-libc-redox-fork.patch:

libc = { path = "../../../../local/sources/libc" }

+rb build metadata is what makes this safe: it leaves the upstream semver (0.2.189) intact, so ^0.2 requirements still resolve. A -rb suffix would be read as a pre-release and fail to satisfy them — see local/AGENTS.md § "Version conventions".

Registered in local/fork-upstream-map.toml as diverged. The fork is vendored from the crates.io package, whose file set differs from the upstream git tag by construction — the package omits .github/ and adds .cargo_vcs_info.json — so a tag content-diff reports differences that mean nothing. snapshot mode was tried first and produced exactly that noise. src/unix/redox/mod.rs is also on the expected-differ list in verify-fork-versions.sh for when the mode changes.

Retiring this fork

Upstream-reportable: the gap belongs in the libc crate. When upstream exposes the Redox waitid surface:

1. Confirm CLD_*, P_*, idtype_t and waitid are all present in src/unix/redox/mod.rs for the version being resolved.
2. Drop 02-libc-redox-fork.patch from recipes/dev/rust/recipe.toml.
3. Remove the libc row from local/fork-upstream-map.toml and both libc) arms from verify-fork-versions.sh.
4. Delete local/sources/libc/.
5. Rebuild rust-native to confirm nix still compiles.

Until then this fork stays. Bumping it follows FORK-BUMP-PATCHING-POLICY.md: re-vendor the newer upstream and re-apply the Redox block — never carry the old base forward.