Exact

Exact Union RegExp

Repository files (latest commit first)

Filename	Latest commit message	Latest commit date
Transaction Science 07c4519e6c ed25519-dalek 3 across the standard: the dalek and RustCrypto generation together ... Every requirement on ed25519-dalek 2, x25519-dalek 2, curve25519-dalek 4 and signature 2 moves to 3, 3, 5 and 3 (51 in 48 manifests); hpke 0.13 -> 0.15. 52 lockfiles now resolve ed25519-dalek 3.0.0, the released stable. Each 2.x feature set maps to the 3.x set with the same API: 2's defaults (fast, std, zeroize) are 3's defaults plus `alloc`, which 3 no longer turns on and which PKCS#8 encoding and `Signature::to_vec` need; 2's `std` is 3's `alloc`. Key generation. dalek 3's `generate` takes a rand_core 0.10 RNG, and every site passed a rand_core 0.6 one. dalek 2's `generate` was fill 32 bytes, then `from_bytes`; that body is written out at the 50 call sites with the RNG each already used, so no RNG or dependency changes. op-screening and joule-code-identity take the 32 bytes from getrandom directly. - jouleclaw-hpke on hpke 0.15 and rand 0.10. Measured across versions first: a seed names the same public key under 0.13 + rand 0.9 and 0.15 + rand 0.10; the stored secret is now the clamped scalar of the same X25519 key, and a secret stored by either version opens what the other sealed. A test pins the seeded key. The suite is asked for by name (hkdfsha2 is no longer a default, ML-KEM is and is not used here). - wai-rs's corpus signs with hazmat::raw_sign over dalek's own Sha512. HPKE's X25519 and the claim corpus's curve arithmetic are unchanged on 5. - smart-byte: iroh 1.0.0-rc.0 pinned ed25519-dalek =3.0.0-pre.7; iroh 1 and iroh-blobs 0.104 take the stable 3. iroh-blobs is not called yet. - sandbox-tee's amd-kds: c28dd2e46 moved sandbox to sha2 0.11 with rsa 0.9 (digest 0.10) still under it, and the feature has not compiled since; the nightly-only job that builds it had not run. rsa 0.10 (digest 0.11, signature 3), and the test chain built field by field and signed over its DER, so x509-cert's builder and the signature traits it ties RSA to are not needed. 52 tests pass with amd-kds,intel-pcs. Left where a third party has not released on the new curve: c2pa (wai-rs), openmls and frost-ed25519 (joule-code, jcp-threshold), bulletproofs and snow (smart-byte; smart-byte-zk keeps curve 4.1 because its scalars go into bulletproofs, and says so), and the libp2p 0.56 and hpke-rs 0.7 lines. Each is handed bytes, never a key type of ours. Checked: all 53 workspaces `cargo check --all-targets` with every feature, or the defaults where every feature fails for a reason the log shows is not this (CUDA, livekit, an old wasm-bindgen); wai-rs as CI builds it and with codecs and neural on macOS; gstreamer on Linux; the desktop app after the Studio. Tests: jcp-rs 755 (one RAPL timing test failed once at load 229 on the arm64 host, passes alone on main there and in the full run), wai-rs lib 1252, joule-code 694 + 88, smart-byte 183, op-screening 50, sandbox-tee 52, joulecontract 17, jouleclaw-hpke 12. The wasmQ modules of record rebuild byte for byte on x86_64 Linux (main's lockfile first, then this one): no dalek code is in them, so only the lockfile digest moves. The views page's script and styles are unchanged; its engine moved with its dependencies. lockfiles current, advisories: nothing new. Co-Authored-By: Claude Opus 5.5 <[email protected]>		2026-11-12 00:15:14 -05:00
..
crates	joule-ui-core lints clean, and a test holds apply_deltas to stopping at the first error	2026-11-07 03:08:05 -05:00
Cargo.lock	ed25519-dalek 3 across the standard: the dalek and RustCrypto generation together	2026-11-12 00:15:14 -05:00
Cargo.toml	ed25519-dalek 3 across the standard: the dalek and RustCrypto generation together	2026-11-12 00:15:14 -05:00
LICENSE	licences: ship the Apache-2.0 text the code already declares	2026-10-09 17:22:32 -04:00