libwally-core

This commit adds a new public-key compression function to a cryptographic library and updates the existing decompression function so both can accept already-compressed or already-uncompressed keys as input. It also adds tests and language bindings. There is no security vulnerability here; it is a normal feature addition with defensive input validation.

This commit tightens signature validation in a cryptographic library. It now rejects malformed DER-encoded signatures that are longer than allowed, or where one of the signature numbers (R or S) is zero. Previously, the underlying secp256k1 library would accept these invalid forms silently, which could lead to later failures or unexpected behavior when the signatures are used in Bitcoin/Elements transactions.

This commit adds support in libwally-core for a new Elements/Liquid confidential-wallet policy format. It lets wallet policies include a special '@B' blinding key placeholder, kept separate from ordinary derivation keys. The change is a feature addition with embedded validation rules; it does not appear to fix a known vulnerability, but it does touch sensitive descriptor-parsing code that handles cryptographic keys. There is no vendor statement that this is a security fix, and no independent researcher is credited.

This commit adds a safety check when loading BIP32 extended keys. Before, the code accepted any 33-byte public key bytes without verifying they form a valid public key. Now it calls a public-key validation function and rejects invalid keys. The change is defensive: it prevents malformed or attacker-chosen public keys from being treated as legitimate BIP32 keys later. The commit also adds test vectors for many other invalid key types, though only the public-key validation is newly enforced in this patch.

This commit only adds new build options and a test harness for fuzzing (automated security testing). It does not change any existing library code that handles PSBTs or other data, so it cannot by itself introduce a security vulnerability or fix one. It is purely an infrastructure addition to help future security testing.

This commit fixes a memory-management bug in the library's Partially Signed Bitcoin Transaction (PSBT) handling. When the code failed to allocate memory for new transaction inputs or outputs, it could leave behind partially initialized data and might not clean up old input/output records properly. The patch makes sure initialization only happens if memory was actually allocated, and that any old or partially built data is fully freed before returning an error. In practice this is mainly a reliability/hardening fix against out-of-memory conditions rather than a directly exploitable attack path.

This commit fixes a parsing bug in how the library reads Bitcoin-style witness data from a transaction or PSBT. Before the fix, a malformed input could claim there were millions of witnesses while only providing a few bytes. The code would then try to read far beyond the available data, likely causing a crash or memory corruption. The fix adds a simple size check so the parser rejects obviously impossible inputs early.

This commit fixes a bug in how the library reads compact size numbers from PSBT (Partially Signed Bitcoin Transaction) data. Before the fix, the code could read past the end of a short input buffer when checking how many bytes a variable-length integer needs. The fix adds a check that at least one byte is available before asking how long the integer is. This is a memory-safety bug that could cause crashes or undefined behavior when parsing malformed PSBT data.

This commit replaces a simple length check on Taproot control blocks with a dedicated verification function. The change is likely a hardening or correctness improvement rather than a clear-cut security fix. The old check only ensured the control block had a valid-looking size; the new function may also validate internal structure, such as the parity bit and merkle proof path. Without the implementation of the new verification function, we cannot confirm whether it closes a real vulnerability or just reduces duplicated logic.