Architecture

noyalib-serde-yaml is deliberately the smallest crate in the family. Its whole source is one line:

#![allow(unused)]
fn main() {
pub use noyalib::compat::serde_yaml::*;
}

Everything else is contract.

What the re-export gives you

noyalib::compat::serde_yaml is the core library’s behavioural serde_yaml 0.9 shim: the same function names, types, and error phrasing, pinned by an 18-case contract suite captured live from serde_yaml 0.9.34 and kept in the core repository (tests/serde_yaml_contract.rs). Publishing that module under a package name lets Cargo’s package = rename substitute it for serde_yaml without a single source change downstream.

Why a separate crate

A dependency rename can only target a crate on the registry with the right name. The core cannot be renamed, so a one-line crate is the smallest artefact that makes the migration a Cargo.toml edit. Keeping it empty of logic means there is nothing here to drift from the core.

Features

The crate enables exactly std and compat-serde-yaml on the core, with default features off, so it pulls nothing the shim does not need.

Testing

tests/drop_in.rs exercises the surface as a renamed dependency would see it. docs/CONTRACT.md records the behavioural guarantees. Two libFuzzer targets in fuzz/ feed arbitrary input through the shim’s from_str and a to_string round trip, and CI replays the seed corpus on every push. The README’s Rust block is compiled in CI against this crate under the rename.

Lockstep

The crate pins noyalib at the identical =0.0.X and releases with it (core ADR-0005). The exact pin is the compatibility contract: the behaviour a user gets is the behaviour that version of the core tested.