Phase 4 of 4 · Study weeks 13-16
Supply chain, interop and the decision to adopt
This phase makes the last checkpoint possible: a project you can ship, call from another language and defend to a team.
Topics every week
- Week 138 h
- Cargo workspaces and features
rust-toolchain.tomland MSRV- Semver and API guidelines
- Editions
Hand inA workspace with a pinned toolchain and a stated MSRV
- Week 148 h
cargo-denyand advisories- Auditing dependencies
- A dependency budget
build.rsand proc macros as risk
Hand inA policy file where a seeded bad dependency fails CI
- Week 158 h
- A C ABI with
cbindgen - Python bindings with PyO3
- Calling C++ with
cxx - Ownership across a boundary
Hand inA Python or C client that reads and writes the store
- A C ABI with
- Week 168 h
- Release builds and SBOMs
- Upgrade policy for a six-week release cycle
- Where Rust helps and where it does not
- Review
Hand inCheckpoint 4, the adoption memo and the upgrade policy
Practice projects
- Practice 1, the dependency policy (Week 14): write a
deny.tomlfor licences and advisories. Done when: a seeded vulnerable crate and a seeded unlicensed crate each fail CI, and removing them passes. - Practice 2, the boundary (Week 15): expose
getandputto Python. Done when: a Python test passes, memory is freed on both sides and the ownership rule for each pointer or buffer is written down. - Practice 3, the upgrade rehearsal (Week 16): build with the previous and the current stable compiler. Done when: a table compares tests, build time and binary size, and says whether to move.
- Optional, a C++ caller: call the store from C++ with
cxx. Done when: a test passes and the build runs in CI.
GitHub repos to study
Existence and last commit were checked on 22 Nov 2026 by shallow clone. Every repository below had a commit in the last 12 months.
| Repo | What to read | Week |
|---|---|---|
| EmbarkStudios/cargo-deny | The check types and the config file | 14 |
| rustsec/rustsec | The advisory database and its format | 14 |
| PyO3/pyo3 | How ownership and the interpreter lock are handled | 15 |
| mozilla/cbindgen | How a C header is generated from Rust types | 15 |
| dtolnay/cxx | A checked boundary between Rust and C++ | 15 |
Milestone checkpoint
Run Checkpoint 4. Build from a pinned toolchain, add the dependency policy and the seeded bad crates, write the Python or C client and produce the SBOM. Write the adoption memo with the numbers from the checkpoints: what Rust prevented, what it cost in hours and where it did not help. Hand in the CI output (a failing and a passing run), the client test, ADR-7 (the unsafe and dependency policy) and the memo.
Check yourself
Answer each question in your head first, then open it to check.
What does an advisory database tell you that a lockfile does not?
Which versions of which crates have known vulnerabilities, so you can match your lockfile against them.Why are
build.rsfiles and procedural macros a supply-chain risk?They run arbitrary code at build time, with your permissions, before any of your tests run.What goes wrong when a pointer crosses a C boundary?
Nobody knows who frees it, how long it lives and what the layout is. Write the rule for each pointer and test both sides.Why does Rust have no long-term support release, and what do you do about it?
Releases arrive every six weeks and old ones get no fixes. Upgrade on a schedule, pin the toolchain and keep a stated minimum supported version for libraries.When would you tell a team not to adopt Rust?
When the team has no one to review the code, when the component is small and rarely changes, or when a library you need exists only in another language. Put numbers on the cost first.