[Niqnil]

Why

cargo audit warns that chacha20 0.10.0 in Cargo.lock is yanked. 0.10.0 and 0.10.1 were yanked because their SSE2 backend, used by the RNG and the 64-bit-counter variants, called an SSE4.1 intrinsic (_mm_extract_epi32). That is undefined behaviour on an x86 CPU without SSE4.1. 0.10.2 (RustCrypto/stream-ciphers#580) stores the counter row with a plain SSE2 store instead.

chacha20 reaches the workspace only through rand 0.10.3, which is used by rustrade-execution and by tungstenite, in turn used by rustrade-integration. The practical risk is low, since every x86 CPU since about 2008 has SSE4.1 and modern ones take the AVX2 path. Still, a yanked crate shouldn't stay in the lockfile.

Change

• cargo update -p chacha20 --precise 0.10.2. Cargo.lock changes by this one package (version and checksum), and its dependency list is unchanged.
• No Cargo.toml change: chacha20 is not a direct dependency.
• No CHANGELOG entry, since the lockfile doesn't reach library users. A fresh resolve already skips yanked versions.

Supply-chain screen

• Downloaded both .crate files. Their SHA-256 sums match the old and new Cargo.lock checksums.
• Diffed src/ from 0.10.0 to 0.10.2:
  • the SSE2 fix;
  • an aarch64 NEON backend widened to 8-block batches;
  • crate-level docs moved to the README;
  • re-export reshuffles: Key and KeyIvInit are now re-exported at the crate root, and ChaCha20LegacyCore and Nonce are added;
  • a doc-comment fix and a constant naming in rng.rs.
• No build.rs, no proc-macro, no network, process or env access, and no new dependencies. Not a tier-2 crate.

Verification

• cargo test --workspace --all-features --lib --tests: 2836 passed.
• The pre-commit clippy shows only the expected existing warning, and both rustdoc passes are clean.
• Local cargo audit no longer reports the yanked crate. The remaining advisories are the ones already ignored in ci.yml and deny.toml.