set/frozenset values fell through to repr() (hash-randomized member order), and objects
with the default <X at 0x..> repr embedded a per-process memory address - both made
fingerprint_data return a different digest across restarts despite its 'deterministic'
contract (a fix-wave d446f50 turned a loud TypeError into a silent unstable hash).
_fingerprint_normalize now canonically sorts set/frozenset members, and the default=
handler raises TypeError on an identity-based repr instead of hashing an address.
Signed-off-by: disqualifier <dev@disqualifier.me>
the docstring said a list-nested blob is "not covered by the depth-limit raise
below", but the cutoff check uses _has_encrypted_field, which does walk lists -
so a list-nested blob reached via the cutoff dict scan DOES raise, while the
same blob one level shallower (hit during normal traversal instead) is what's
actually skipped silently. doc-only, no behavior change.
Signed-off-by: disqualifier <dev@disqualifier.me>
hoists the byte-identical OAEP(MGF1(SHA256), SHA256, label=None) construction out
of encrypt_aes_key_with_rsa and decrypt_aes_key_with_rsa into a module-level
_OAEP_PADDING constant so a future scheme change lands in one place instead of
two in lockstep. also restores encrypt_aes_key_with_rsa's Args/Returns/Raises
docstring block, matching its four siblings, after Wave-1 over-stripped it to
bare prose.
Signed-off-by: disqualifier <dev@disqualifier.me>
both the bounded pass and the unbounded _has_encrypted_field fallback
descended only through dict values, so a blob nested inside a list at
any depth was invisible and the function returned False. reencrypt()
already skips list-nested blobs (documented gotcha), so after rotation
such a blob was stranded under the old key while this audit reported
the record clean - a rotation-data-loss trap once the old wrapped-key
record is deleted. both traversal passes now walk list/tuple items in
addition to dict values; the blob-detection predicate is unchanged.
bump 0.1.7 -> 0.1.8
Signed-off-by: disqualifier <dev@disqualifier.me>
EC-5: _load_private_key branches on whether a password was given so the normalized
ValueError matches the actual cryptography TypeError case (was always claiming
"encrypted but no password" even when a password was given for an unencrypted key).
EC-6: encrypt_aes_key_with_rsa/decrypt_aes_key_with_rsa now raise a clear ValueError
via _require_rsa for a non-RSA key (e.g. Ed25519/EC), instead of crashing raw with
AttributeError at wrap/unwrap — this lib is RSA-envelope only.
EC-7: initialize() requires exactly 32 bytes (isinstance bytes, len==32), rejecting a
16/24-byte key (silent AES-128/192 downgrade) or a str instead of failing late and
opaquely at first encrypt.
EC-8: fingerprint_data gains a default= handler (datetime/date/time, bytes/bytearray,
and a type-tagged repr fallback) plus a key-type-tagging pre-pass so datetime/bytes/
ObjectId-like values no longer TypeError and int-vs-str dict keys no longer collide
to the same fingerprint. Never logs the data being fingerprinted.
Also compresses the essay-length docstrings (module + several methods) to cut
narration while keeping the load-bearing footgun notes (RSA-only, AES-256 key length,
never-log-key-material) intact — zero behavior change, re-verified after.
Signed-off-by: disqualifier <dev@disqualifier.me>
self_test(is_file=False) only forwarded is_file to the public-key wrap;
decrypt_aes_key_with_rsa had no is_file parameter and always open()'d its
argument, so a PEM string was opened as a filename, misdiagnosing a good
keypair as non-pairing and leaking the private key PEM into the
FileNotFoundError traceback. decrypt_aes_key_with_rsa now takes is_file
(default True, preserving current callers), and self_test threads it
through to both key loads.
encrypt_data json.dumps a dict without checking key types, silently
stringifying int/float/bool/None keys (e.g. snowflake-int-keyed dicts),
so a decrypt round-trip silently lost the original key. encrypt_data now
raises TypeError on a non-str key instead of coercing it.
Signed-off-by: disqualifier <dev@disqualifier.me>
Two silent-data-loss paths in the record-level functions: (1) reencrypt's
traversal_level cutoff silently left blobs nested deeper than the default
under the old key with no signal, contradicting its own documented fail-loud
rotation contract, and is_encrypted_record shared the cutoff so a
post-rotation audit couldn't detect the leftover; (2) all three record
functions inspected only record.values(), never the record itself, so a bare
{secure, iv, data} blob used as the whole document (the README's file-storage
pattern) was invisible to is_encrypted_record and passed through reencrypt
unchanged under the old key.
reencrypt now raises when a blob sits deeper than traversal_level instead of
silently truncating, and detects/handles the record-itself-is-a-blob case;
is_encrypted_record falls back to an unbounded-depth scan past
traversal_level so it reliably flags leftovers regardless of nesting depth;
decrypt_record likewise handles a record that is itself a blob. Bumped to
v0.1.4.
Signed-off-by: disqualifier <dev@disqualifier.me>
EC-1: factor _fingerprint_of(public_key) so encrypt_aes_key_with_rsa fingerprints the
already-loaded key instead of re-opening/parsing the file. encrypt_data rejects a
non-dict/str with a clear TypeError (was: opaque .encode() AttributeError); decrypt_data
raises ValueError on a malformed blob (was: raw KeyError); a wrong-password PEM load gives
a clearer message; reencrypt's dict-only traversal (list-nested blobs skipped) documented.
Signed-off-by: disqualifier <dev@disqualifier.me>
- encrypted OpenSSH private key with no password now raises ValueError (not a raw
TypeError from load_ssh_private_key), matching the PEM path and the docstring (L14)
- a non-PEM/non-SSH public key raises a clear ValueError instead of cryptography's
UnsupportedAlgorithm, consistent with the private-key paths (L15)
- decrypt_data only treats a json-OBJECT plaintext as a dict, so json-shaped strings
('123','true','[1,2]') round-trip as strings; existing dict blobs unaffected (L16)
- both key loads route through shared _load_private_key/_load_public_key helpers
- document reencrypt's fail-loud (vs decrypt_record's per-field swallow) asymmetry (nit).
Signed-off-by: disqualifier <dev@disqualifier.me>
get_rsa_key_fingerprint(is_private=True) only loaded PEM private keys, so an OpenSSH-format private key raised — unlike decrypt_aes_key_with_rsa, which already had the fallback. mirrored it: on a PEM load failure, an OPENSSH-marked key is loaded via load_ssh_private_key. also normalized the encrypted-key-without-password case: cryptography raises TypeError there, which now becomes a clear ValueError('private key is encrypted but no password was provided') in both methods instead of leaking the raw TypeError.
Signed-off-by: disqualifier <dev@disqualifier.me>
both used record.copy() (shallow), leaving unencrypted mutable fields shared between the input and the returned dict, violating the documented 'input is not mutated' contract. switched to copy.deepcopy.
Signed-off-by: disqualifier <dev@disqualifier.me>
get_rsa_key_fingerprint(is_private=True) called load_pem_private_key(password=None),
so an encrypted private key raised a raw TypeError. add an optional password param
forwarded to the load; unencrypted keys ignore it.
verified: encrypted private key fingerprints with its password and matches the
public key's fingerprint; missing password still raises.
Signed-off-by: disqualifier <dev@disqualifier.me>
dsql