Record the 1.0.x changes and the 0.8.1 release in the CHANGELOG

[zandbelt]

Summary

Follow-up to your comment on #198. Documentation only, no code.

The README already covers ML-DSA

#198 updated it: the JWS alg table has the ML-DSA-44, ML-DSA-65, ML-DSA-87 row with build option CJOSE_ENABLE_ML_DSA, OpenSSL >= 3.5, the JWK kty table has the AKP row with the same requirement, the build options table has CJOSE_ENABLE_ML_DSA, and a paragraph says the algorithms are pure ML-DSA with the empty context string and that HashML-DSA is not offered.

The one place that did not mention it is Prerequisites → Libraries, which said OpenSSL >= 3.0.0 and nothing more. That line now names the 3.5 requirement and points at the build option. It is the only README change here.

The CHANGELOG had a larger gap than ML-DSA

The 1.0.0 (unreleased) section held only the Breaking list that #187 wrote. Nothing merged after it had an entry — not ML-DSA, and not the nine PRs before it:

#185	PBES2-HS256+A128KW, PBES2-HS384+A192KW, PBES2-HS512+A256KW
#186	X25519 and X448 ECDH-ES, the epk private-member refusal, the EC d refusal
#188	A128GCMKW, A192GCMKW, A256GCMKW; the crit refusal; JWE header disjointness
#189, #190, #192, #193	the RSA JWK import refusals
#191	the NULL cjose_err crash in ECDH-ES+A*KW decryption
#196	EVP_Q_mac for HMAC
#198	ML-DSA-44/65/87 and the AKP key type

All of them are now recorded under Update and Fix. The entries reference pull requests, which is what the Breaking list of that section already does, rather than the commit links the released sections use. Nothing in the released sections is touched — the diff is 51 added lines and no deletions.

Classification. The crit refusal is under Breaking: it refuses a JWE or JWS that 0.8.0 accepted. Three more rules refuse input or calls that 0.8.0 accepted without changing the API, so they stay under Fix and a short Compatibility paragraph names them — the disjointness of the header locations, the refusal of a header parameter the algorithm generates, and the refusal of a valueless private member. That mirrors how the 0.8.1 notes handled the same rules.

0.8.1 was missing from this file entirely

You released 0.8.1 from the 0.8.x branch on 2026-09-14, and its section only ever existed on that branch. On main the file jumped from the unreleased 1.0.0 straight to 0.8.0, so a released version was invisible in the release notes. Its section is copied over unchanged, including the autotools test-build fix, which describes what 0.8.1 shipped rather than what main has.

Comparing the two also turned up an entry the 1.0.0 list would otherwise have missed: the EC d refusal, which 0.8.1 lists separately and which main got in #186 rather than with the RSA members in #189.

Testing

No code changes, so nothing to build or run. Verified instead that every entry is true of the tree at 8d1afeb: each function, macro, build option, header parameter and error code exists with that spelling, each behavioural claim matches the end state rather than an intermediate commit of a multi-commit PR (#188 scoped crit per algorithm family before refusing it outright; #185 changed how p2s is handled), and each PR attribution matches the commit that introduced the behaviour.

🤖 Generated with Claude Code

[TheStormN]

Thanks!