Add typing: asn, exceptions, hashes, pwdbased, utils.

[roberthdevries]

No description provided.

[Trooper-X]

This also adds the ruff and ty dependencies.
Would be nice if this MR gets merged.

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: REQUEST_CHANGES
Findings: 7 total — 7 posted, 0 skipped
6 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [High] AES-SIV single-block associated-data length uses char count of original input, not encoded byte length — wolfcrypt/ciphers.py:397-400
• [Medium] sign_with_seed no longer accepts bytearray/memoryview seeds (regression) — wolfcrypt/ciphers.py:2513-2546
• [Medium] make_key_from_seed now silently UTF-8-encodes a str seed instead of rejecting it — wolfcrypt/ciphers.py:2360-2367
• [Low] ChaCha init renamed size to _size, breaking the documented backward-compatible keyword — wolfcrypt/ciphers.py:544
• [Low] HKDF helpers annotate hash_cls as instance type instead of class type — wolfcrypt/hkdf.py:33,78,105
• [Low] Random no longer nulls native_object on init failure; del frees an uninitialized RNG — wolfcrypt/random.py:37-52
• [Low] RsaPublic.init made key a required positional argument — wolfcrypt/ciphers.py:771-774

Review generated by Skoll

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: REQUEST_CHANGES
Findings: 7 total — 7 posted, 0 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [High] New undeclared runtime dependency on typing_extensions — wolfcrypt/ciphers.py:28, wolfcrypt/hashes.py:28
• [Medium] Removed _ffi.from_buffer() drops bytearray/memoryview support for seed/rand inputs — wolfcrypt/ciphers.py:2383, 2548, 2559, 2067, 2110
• [Medium] **_Cipher.new() dropped kwargs, breaking PEP 272 extra keyword arguments — wolfcrypt/ciphers.py:187-199
• [Medium] HKDF functions annotate hash_cls as instance type instead of class type — wolfcrypt/hkdf.py:33, 78, 105
• [Medium] asn.py leaves function arguments unannotated while enabling ANN ruff rules — wolfcrypt/asn.py:81, 99
• [Low] test_mldsa now relies on cffi's low-level TypeError instead of an explicit guard — tests/test_mldsa.py:186
• [Low] RsaPublic.init made key a required positional argument — wolfcrypt/ciphers.py:781

Review generated by Skoll

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: COMMENT
Findings: 3 total — 3 posted, 0 skipped
3 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] ML-DSA seed handling: bytearray/memoryview now rejected, and the two seed methods validate inconsistently — wolfcrypt/ciphers.py:2371-2392 (make_key_from_seed), 2516-2540 (sign_with_seed)
• [Low] *ML-KEM _with_random helpers no longer accept bytearray/memoryview for rand — wolfcrypt/ciphers.py:2059-2077 (encapsulate_with_random), 2103-2119 (make_key_with_random)
• [Low] hkdf.py forces a runtime import of _Hmac for a type annotation (no from future import annotations) — wolfcrypt/hkdf.py:30-33

Review generated by Skoll

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: COMMENT
Findings: 4 total — 4 posted, 0 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] ML-DSA seed validation now rejects bytearray/memoryview (regression) — wolfcrypt/ciphers.py:2383,2537
• [Medium] Advertised list/tuple seed support is untested and likely fails at the cffi boundary — wolfcrypt/ciphers.py:2372,2391
• [Low] setup.py install_requires not synced with new typing-extensions runtime dependency — setup.py:62-63
• [Low] Hard runtime import of private cffi symbol _cffi_backend.Lib only to satisfy a cast — wolfcrypt/__init__.py:20-22,56

Review generated by Skoll

[dgarske]

See #125 (review)
If you disagree just make note. We are getting close on this and thank you for your efforts

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: COMMENT
Findings: 4 total — 4 posted, 0 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] typing-extensions dependency has no minimum version (override requires =4.4.0) — pyproject.toml:27
• [Low] bytes(seed) silently accepts an int, dropping the friendly type check for ML-DSA seeds — wolfcrypt/ciphers.py:2383,2536
• [Info] New module wolfcrypt/types.py shadows the stdlib types module name — wolfcrypt/types.py:1
• [Info] Inconsistent # ty:ignore comment lacks the space used everywhere else — wolfcrypt/__init__.py:53

Review generated by Skoll

[dgarske]

Please also resolve merge conflicts.

#125 (review)

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: COMMENT
Findings: 2 total — 2 posted, 0 skipped
2 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Low] *MlKem _with_random drops buffer-protocol support for rand — wolfcrypt/ciphers.py:2096-2098,2140
• [Low] setup.py dependencies out of sync with pyproject (typing-extensions) — setup.py:62-63

Review generated by Skoll

[dgarske]

Please resolve merge conflicts. Also see #125 (review)

[dgarske]

Skoll Code Review

Scan type: reviewOverall recommendation: COMMENT
Findings: 4 total — 4 posted, 0 skipped
4 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] make_key_with_random drops buffer-protocol handling, inconsistent with encapsulate_with_random — wolfcrypt/ciphers.py:2151
• [Low] New TypeError validation branches (with_random / seed) are not covered by tests — wolfcrypt/ciphers.py:2096-2099
• [Low] Random.native_object changed from settable attribute to read-only property — wolfcrypt/random.py:56-60
• [Info] Abstract properties typed - int are overridden by None class attributes — wolfcrypt/ciphers.py:175-177

Review generated by Skoll

[dgarske]

🟠 [Medium] make_key_with_random drops buffer-protocol handling, inconsistent with encapsulate_with_random

This PR replaces _ffi.from_buffer(rand) with a bare rand in wc_KyberKey_MakeKeyWithRandom(mlkem_priv.native_object, rand, len(rand)), but does NOT add the validation/normalization that its sibling MlKemPublic.encapsulate_with_random gained in the same PR (lines 2096-2101: memoryview(rand) guard + rand = bytes(rand)). The old _ffi.from_buffer(rand) accepted any buffer-protocol object (bytes, bytearray, memoryview); passing such objects straight into cffi now depends on cffi's narrower direct-conversion rules, so a bytearray/memoryview caller that worked before can now raise TypeError. Two public entry points that take a rand buffer for the same crypto module now behave differently. Tests only exercise bytes (ref_rand_for_make_key = h2b(...)), so this asymmetry is not caught.

Fix: Mirror the validation/bytes(rand) conversion used in encapsulate_with_random so the two *_with_random APIs accept the same input types consistently.

[dgarske]

🔵 [Low] New TypeError validation branches (with_random / seed) are not covered by tests

The PR adds a memoryview(rand) buffer-protocol guard that raises a custom TypeError in encapsulate_with_random (and analogous refactors normalize seed in make_key_from_seed/sign_with_seed). The tests exercise only the happy path with bytes; the new TypeError branch (e.g. passing a non-buffer such as an int or str) is not tested, so a regression in that guard would go unnoticed. Combined with the finding above, the divergence between make_key_with_random and encapsulate_with_random is also untested.

Fix: Add a small negative test for the new buffer-protocol validation branches.

[dgarske]

🔵 [Low] Random.native_object changed from settable attribute to read-only property

native_object (a non-underscore, effectively public attribute) is converted from a plain writable attribute to a read-only @property backed by _native_object. Any external code that assigned rng.native_object = ... now fails with AttributeError (this is exactly why the test test_delete_descriptor_binding.py had to drop its r.native_object = None line). The change is otherwise reasonable and the internal __del__/init-failure handling is now safer, but this is a backward-incompatible surface change on a public name that is not called out in the ChangeLog (which only mentions 'Add typing annotations').

Fix: Confirm no external consumer sets Random.native_object; consider a brief ChangeLog note that it is now read-only. No code change required.

[dgarske]

⚪ [Info] Abstract properties typed - int are overridden by None class attributes

In this typing-focused PR, several abstract properties are declared to return int (_Cipher.key_size, _Cipher.block_size, _Cipher._key_sizes; _Hash.digest_size) but concrete subclasses override them with class attributes set to None (Aes.key_size = None, ChaCha.key_size = None, Sha3.digest_size = None). This is fine at runtime (ABC only checks the name exists, and the falsy None is intentionally used as a sentinel in _Cipher.__init__), but it is type-inconsistent with the abstract signatures and will produce warnings under the newly added [tool.ty] config (all = "warn"). Consider annotating these as int | None (or Optional[int]) to match reality.

Fix: Widen the abstract property annotations to int | None where subclasses legitimately use None as a sentinel, to keep the type declarations consistent.

[dgarske]

#125 (review)