Release v0.6.0

CHANGELOG.md

@@ -5,26 +5,23 @@ All notable changes to eth.zig will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [Unreleased]
-
-### Fixed
-- DEX `getAmountIn`/`getAmountOut` are correct standard Uniswap V2; the `getAmountIn inverse` round-trip test used unrealistic tiny-reserve parameters where flooring a small output legitimately breaks the within-2-units invariant. Corrected the test to balanced reserves and un-quarantined it (#83). The full unit suite now runs with zero skips.
-- BIP-39 seed derivation (`mnemonic.toSeed`) was correct; its Trezor passphrase test had a wrong expected vector. Corrected the test to the canonical vector (verified against `std.crypto` PBKDF2) and un-quarantined it (#82).
-- `mulDiv` now returns the correct result when the intermediate product overflows u256 with a full-width divisor, e.g. `mulDiv(MAX, MAX, MAX)` (#81). The limb-native Knuth-D divider mis-handled this boundary; the rare overflow path now computes exactly in native u512 (the benchmarked u128 fast path is unchanged).
-- Pure-Zig Keccak-256 fallback (`keccak_optimized.zig`) now produces correct digests (#80): replaced the broken lane-complementing chi optimization with the standard chi step (which lowers to ANDN anyway). Verified against pinned vectors and cross-validated with stdlib across all block sizes.
-- **Unit tests now actually run.** The test step aggregated module tests via `_ = eth.module` field access, which only forces semantic analysis -- it never collected the per-module `test` blocks, so assertions compiled but never executed (`zig build test` was green even with a deliberately failing test). The test artifact is now rooted at `src/root.zig`, whose test block direct-imports every module file. This surfaced real, previously-hidden defects, now fixed: `abi_json` still used the 0.15 `ArrayList` API (broken on 0.16); `rlp.bytesToUint` failed to compile for small integer types; `mnemonic.entropyToMnemonic` had a comptime-resolution bug and a u8 overflow for 24-word phrases; `parseInputs`/`parseParam` formed an inferred-error-set dependency loop; `rlp` struct reflection used the pre-0.17 `Type.Struct.fields` layout (now via `std.meta.fieldNames`, portable across 0.16 and 0.17-dev). Fixed several stale test expectations (`formatHash`/log-address length typos, a static-tuple ABI length, a fixed-size hex length error). Vendored XKCP/secp256k1 C is now built with `-fno-sanitize=undefined` so their intentional unaligned loads do not trap the Debug UBSan runtime.
-- Resolved the four pre-existing issues the revived test suite exposed: two were real code bugs (broken pure-Zig Keccak fallback #80, `mulDiv` at the u256 boundary #81) and two were bad tests (a BIP-39 passphrase seed vector #82, a DEX round-trip assumption #83). All are fixed and the full unit suite now runs with zero skips (775/775).
-
-### Changed
-- Benchmark comparison refreshed against current alloy crates (alloy-primitives 1.6.0, alloy-sol-types 1.6.0, alloy-dyn-abi 1.6.0, alloy-consensus 2.0.5, alloy-signer 2.0.5, alloy-rlp 0.3.15); bench/alloy-bench migrated from alloy-primitives 0.8 / alloy 0.6. New score: eth.zig wins 18/26 (alloy improved Keccak on larger inputs, hex encoding, RLP u256 decoding, and UniswapV4-style swap math)
+## [0.6.0] - 2026-06-10
 
 ### Added
 - `nonce_manager` module: atomic nonce manager for concurrent senders (#75). `NonceManager.init` seeds lazily from the `pending` transaction count (no RPC in `init`); `next()` is a lock-free atomic fetch-and-add so multiple threads never receive the same nonce; `peek()` reads the next nonce without advancing; `resync()`/`reset()` re-fetch from chain after dropped txs; and `onFailure(nonce)` returns a nonce to the pool only when it is the most recently issued one (a middle nonce cannot be safely reused once a higher one is in flight). Also adds `provider.getTransactionCountAt(address, tag)` so the nonce count can be read at any block tag (e.g. `.pending`)
 - MEV-Share backrunner example (`examples/08_mev_share_backrunner.zig`): a teaching bot that matches the MEV-Share SSE event stream against comptime-computed Uniswap V2/V3 swap selectors and composes a backrun bundle (user tx hash + signed EIP-1559 backrun tx), with a safe `DRY_RUN` default that prints the `mev_sendBundle` params instead of submitting (#38)
 - `mev_share` module: MEV-Share client mirroring `mev-share-client-ts` -- `sendTransaction` (eth_sendPrivateTransaction via `flashbots.Relay`), `simulateBundle` (mev_simBundle with `SimBundleOpts`/`SimBundleResult`), blocking SSE event stream subscription (`MevShareClient.on` with `PendingEvent`/`PendingTransaction`/`PendingBundle` and pure `parseEventData`), and `getEventHistory` (GET /api/v1/history) (#34)
 - `eth_sendPrivateTransaction` (MEV-Share private transactions) on `flashbots.Relay`: submit a single private transaction with hint preferences (calldata, contract_address, logs, function_selector), builder selection, fast mode, and max inclusion block (#40)
 
+### Changed
+- Benchmark comparison refreshed against current alloy crates (alloy-primitives 1.6.0, alloy-sol-types 1.6.0, alloy-dyn-abi 1.6.0, alloy-consensus 2.0.5, alloy-signer 2.0.5, alloy-rlp 0.3.15); bench/alloy-bench migrated from alloy-primitives 0.8 / alloy 0.6. New score: eth.zig wins 18/26 (alloy improved Keccak on larger inputs, hex encoding, RLP u256 decoding, and UniswapV4-style swap math)
+
 ### Fixed
+- **Unit tests now actually run.** The test step aggregated module tests via `_ = eth.module` field access, which only forces semantic analysis -- it never collected the per-module `test` blocks, so assertions compiled but never executed (`zig build test` was green even with a deliberately failing test). The test artifact is now rooted at `src/root.zig`, whose test block direct-imports every module file. This surfaced real, previously-hidden defects, now fixed: `abi_json` still used the 0.15 `ArrayList` API (broken on 0.16); `rlp.bytesToUint` failed to compile for small integer types; `mnemonic.entropyToMnemonic` had a comptime-resolution bug and a u8 overflow for 24-word phrases; `parseInputs`/`parseParam` formed an inferred-error-set dependency loop; `rlp` struct reflection used the pre-0.17 `Type.Struct.fields` layout (now via `std.meta.fieldNames`, portable across 0.16 and 0.17-dev). Fixed several stale test expectations (`formatHash`/log-address length typos, a static-tuple ABI length, a fixed-size hex length error). Vendored XKCP/secp256k1 C is now built with `-fno-sanitize=undefined` so their intentional unaligned loads do not trap the Debug UBSan runtime.
+- Pure-Zig Keccak-256 fallback (`keccak_optimized.zig`) now produces correct digests (#80): replaced the broken lane-complementing chi optimization with the standard chi step (which lowers to ANDN anyway). Verified against pinned vectors and cross-validated with stdlib across all block sizes.
+- `mulDiv` now returns the correct result when the intermediate product overflows u256 with a full-width divisor, e.g. `mulDiv(MAX, MAX, MAX)` (#81). The limb-native Knuth-D divider mis-handled this boundary; the rare overflow path now computes exactly in native u512 (the benchmarked u128 fast path is unchanged).
+- BIP-39 seed derivation (`mnemonic.toSeed`) was correct; its Trezor passphrase test had a wrong expected vector. Corrected the test to the canonical vector (verified against `std.crypto` PBKDF2) and un-quarantined it (#82).
+- DEX `getAmountIn`/`getAmountOut` are correct standard Uniswap V2; the `getAmountIn inverse` round-trip test used unrealistic tiny-reserve parameters where flooring a small output legitimately breaks the within-2-units invariant. Corrected the test to balanced reserves and un-quarantined it (#83). The full unit suite now runs with zero skips.
 - `sse_transport.SseParser.feedLine`: replaced the removed `std.mem.trimRight` with `std.mem.trimEnd` and gave the line-parse result an explicit struct type. These were latent Zig 0.16 migration misses that only surfaced when `feedLine` was semantically analyzed (it is not referenced by the unit-test root), and broke any consumer of `MevShareClient.on` -- surfaced while building the MEV-Share backrunner example (#38)
 
 ## [0.5.0] - 2026-06-10

README.md

@@ -214,15 +214,15 @@ Built something with eth.zig? Open a PR to add it here.
 **One-liner:**
 
 ```bash
-zig fetch --save git+https://github.com/StrobeLabs/eth.zig.git#v0.5.0
+zig fetch --save git+https://github.com/StrobeLabs/eth.zig.git#v0.6.0
 ```
 
 **Or add manually** to your `build.zig.zon`:
 
 ```zig
 .dependencies = .{
     .eth = .{
-        .url = "git+https://github.com/StrobeLabs/eth.zig.git#v0.5.0",
+        .url = "git+https://github.com/StrobeLabs/eth.zig.git#v0.6.0",
         .hash = "...", // run `zig build` and it will tell you the expected hash
     },
 },

SECURITY.md

@@ -4,8 +4,8 @@
 
 | Version | Supported |
 |---------|-----------|
-| 0.5.x   | Yes       |
-| < 0.5   | No        |
+| 0.6.x   | Yes       |
+| < 0.6   | No        |
 
 ## Reporting a Vulnerability
 

build.zig.zon

@@ -1,6 +1,6 @@
 .{
     .name = .eth_zig,
-    .version = "0.5.0",
+    .version = "0.6.0",
     .fingerprint = 0xd0f21900fa26f179,
     .minimum_zig_version = "0.16.0",
     .dependencies = .{},

docs/content/docs/installation.mdx

@@ -13,15 +13,15 @@ description: How to add eth.zig to your Zig project.
 **One-liner:**
 
 ```bash
-zig fetch --save git+https://github.com/StrobeLabs/eth.zig.git#v0.5.0
+zig fetch --save git+https://github.com/StrobeLabs/eth.zig.git#v0.6.0
 ```
 
 **Or add manually** to your `build.zig.zon`:
 
 ```zig
 .dependencies = .{
     .eth = .{
-        .url = "git+https://github.com/StrobeLabs/eth.zig.git#v0.5.0",
+        .url = "git+https://github.com/StrobeLabs/eth.zig.git#v0.6.0",
         .hash = "...", // run `zig build` and it will tell you the expected hash
     },
 },