chore: release v5.0.0#272
Open
sw-release-bot[bot] wants to merge 1 commit into
Open
Conversation
polaz
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
from
8b69d29 to
22132a5
Compare
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
from
22132a5 to
9333d1d
Compare
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
4 times, most recently
from
537ad18 to
5fad0f0
Compare
![sw-release-bot[bot]](https://avatars.githubusercontent.com/u/228206483?s=60&v=4){kind=small}
Author
There was a problem hiding this comment.
⚠️ Performance Alert ⚠️
Possible performance regression was detected for benchmark 'lsm-tree db_bench'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.15.
| Benchmark suite | Current: 1c24a37 | Previous: ce61479 | Ratio |
|---|---|---|---|
fillseq |
929652.2170850233 ops/sec (normalized) |
1203358.892593456 ops/sec (normalized) |
1.29 |
fillrandom |
485572.12090749916 ops/sec (normalized) |
684004.055959021 ops/sec (normalized) |
1.41 |
readrandom |
249420.72751907416 ops/sec (normalized) |
293185.8790669865 ops/sec (normalized) |
1.18 |
overwrite |
520188.67348067 ops/sec (normalized) |
651165.60047447 ops/sec (normalized) |
1.25 |
mergerandom |
396090.88280023885 ops/sec (normalized) |
466073.7463258474 ops/sec (normalized) |
1.18 |
This comment was automatically generated by workflow using github-action-benchmark.
CC: @polaz
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
13 times, most recently
from
1f6bceb to
ca002fa
Compare
5 tasks
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
2 times, most recently
from
4d67a8e to
6cb13ca
Compare
polaz
added a commit
that referenced
this pull request
May 23, 2026
…#133 phase 2) (#310) ## Summary Foundation layer for `O_DIRECT` cache-bypass I/O. Two pieces: 1. **`AlignedBuf`** (`src/fs/aligned_buf.rs`) — heap-allocated byte buffer with caller-specified alignment. `Vec<u8>` defaults to `align_of::<u8>() = 1`, which violates `O_DIRECT`'s typical 4 KiB userspace-buffer alignment (EINVAL on unaligned write/read). `AlignedBuf` allocates via `Layout::from_size_align`, rounds capacity up to a multiple of alignment, and exposes a `Vec`-like surface (`len`/`capacity`/`as_slice`/`as_capacity_mut`/`set_len`/`clear`) + raw ptr accessors for kernel handoff. 2. **`FsOpenOptions::direct_io`** — new `bool` field + builder method. On Linux/Android (arches where `asm-generic/fcntl.h`'s `O_DIRECT = 0o40000` is the authoritative value: x86, x86_64, aarch64, riscv32/64, loongarch64, s390x) the flag becomes a `custom_flags(O_DIRECT)` on the std `OpenOptions` builder, in both `StdFs::open` and `IoUringFs::open`. `O_DIRECT` is declared as a named constant rather than pulling in `libc` — matches the existing `EXDEV` / `flock` pattern in `std_fs.rs`. Architectures with a divergent `O_DIRECT` bit (arm `0o200000`, mips `0o100000`, parisc, sparc) are not gated on purpose: emitting the wrong bit silently would be worse than honouring the documented "`direct_io` may be ignored" contract. macOS / Windows / other Unix targets fall through to a cached open for the same reason — `F_NOCACHE` / `FILE_FLAG_NO_BUFFERING` each need their own opt-in plumbing. ## Design choices (explicit so review does not re-raise) - **`FsOpenOptions` is `#[non_exhaustive]`.** This is breaking (struct-literal callers + exhaustive pattern matches stop compiling) but only relative to the same release that introduces the new `direct_io` field — which is already breaking for the same callers. Bundling both changes in one release confines the break to a single semver-major bump (v5.0.0, queued in PR #272) and lets every future field land as semver-minor. Builder methods (`.read()`, `.write()`, … `.direct_io()`) cover every field, so non-struct-literal callers are unaffected. - **`fs::direct_io` submodule IS gated behind `#[cfg(feature = "std")]`.** It uses `std::fs::OpenOptions` and is therefore std-only. The gate is the first concrete step toward honestly feature-gating the std backend per crate policy. However, it does NOT by itself unblock a `no_std + alloc` build: the rest of the `fs::*` backend — the `Fs`/`FsFile` trait *definitions* and all impls (`std_fs`, `io_uring_fs`, even `MemFs`) — still references `std::io::{Read, Write, Seek}` + `std::path::Path` directly, and those have no `core::*` equivalents. Porting the trait surface off `std::io` / `std::path` is a structural prerequisite tracked as a separate prerequisite issue (#311), itself part of the no-std migration epic (#274). The `no-std-check` CI job is `continue-on-error` for exactly this reason — net error count is not changed by this PR. ## What's NOT in this PR The Tree-level `Config::direct_io` knob and compaction-writer integration are deferred to a follow-up. The current SST writer uses `Vec<u8>` buffers and unaligned block sizes; wiring `direct_io` into `Config` without an alignment-aware writer would EINVAL on first write. Foundation lands here so the writer refactor can build on it. ## Test plan - [x] 12 unit tests for `AlignedBuf` (alignment verification, capacity rounding, zero init, rejects non-power-of-two / excessive alignment, zero-capacity dangling sentinel, `set_len` growth + panic, `clear`, `as_capacity_mut` writes, `Send`/`Sync` compile check, pointer stability) - [x] `fs_open_options_default` + `fs_open_options_builders` updated to cover the new field - [x] Full nextest suite: 1542 tests pass (1 slow), 6 skipped - [x] cargo clippy all-features all-targets with -D warnings — clean - [x] Doctest for `AlignedBuf::new_zeroed` example compiles and passes ## Related - #311 — port `Fs`/`FsFile` traits off `std::io` + `std::path` (prerequisite for the rest of `fs::*` to become honestly feature-gateable) - #274 — no-std migration epic (parent) - #272 — release-plz v5.0.0 (semver-major bump that ships the breaking changes here) Part of #133
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
2 times, most recently
from
8f95b2a to
310edff
Compare
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
2 times, most recently
from
a270353 to
40c8fd6
Compare
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
17 times, most recently
from
f303ee5 to
e5ddeca
Compare
8 tasks
Member
Hold — V5 release batch coordinationThis release-plz PR is held in open state pending completion of V5 format batch. RationalePer Q-cycle release strategy decision: V5 will bundle several format-changing features so post-V5 manifests/blocks have a single coherent version step (no V5→V6 rapid succession that would penalize users with chained format migrations). V5 batch contents (in addition to what's already in this PR)
Estimated total wall time: ~2-3 weeks from now. Already in this PR (will ship as part of V5)
Action plan
If batch hits a blockerPer-issue independence allows partial V5 ship — if e.g. #297 unexpectedly blocks, V5 ships with #352 + #298 + #224 only; #297 follows in V5.x (additive) or V6 if it requires format break. Marketing pull-quote (target)"V5: production-grade integrity stack — BuRR filter, Page ECC, AAD encryption, per-KV protection, block-level seqno bounds for CDC, manifest hardening, runtime-toggleable everything with zero-downtime migration via compaction." When this resolves
|
This was referenced May 26, 2026
polaz
added a commit
that referenced
this pull request
May 27, 2026
## Summary V5-1 of the V5 release batch — foundation for live-toggleable configuration that downstream V5 features (#297 manifest hardening, #298 per-KV protection, #224 scan_since_seqno) build on. Caller updates settings via `Tree::update_runtime_config`; subsequent write paths load the new snapshot lockless. Existing on-disk data stays in its original format and reads transparently via per-block self-description. Compaction acts as the live migration mechanism — source blocks rewritten per the current snapshot over subsequent cycles, all data converges without stop-the-world coordination. Closes #352. ## Layered tier separation (no_std-aware) | Module | Tier | Why | |--------|------|-----| | `runtime_config::types` | `alloc` (compiles on no_std + alloc) | Pure data: `RuntimeConfig` + `ChecksumAlgorithm`. Reachable from no_std consumers (block decoders, format constants). | | `runtime_config::handle` | std-bound (`#[cfg(feature = "std")]`) | `ArcSwap`-based snapshot. `load()` = single atomic load; `update(fn)` = clone, mutate, atomic swap. | ## Public API ```rust impl Tree { pub fn runtime_config(&self) -> Arc<RuntimeConfig>; pub fn update_runtime_config<F: FnOnce(&mut RuntimeConfig)>(&self, mutator: F); } ``` ## ChecksumAlgorithm Used by downstream V5 features (#297, #298, #300) for block-level and per-KV checksums. Stable `wire_tag()` / `from_wire_tag()` for on-disk self-description so each block can dispatch verify per its own recorded algorithm. | Variant | Bytes | Speed | Notes | |---------|-------|-------|-------| | `Xxh3_64` (default) | 8 | ~50-100 GB/s SIMD | Default — matches crate-wide XXH3 convention | | `Xxh3Low32` | 4 | same as `Xxh3_64` | Per-entry use case where space matters | | `Crc32c` | 4 | ~30 GB/s HW | Mathematically-proven burst-error detection | ## Tests - **10 unit tests** in `runtime_config::types` + `runtime_config::handle`: - digest sizes, wire-tag roundtrip, atomic swap visibility - snapshot held during update is unchanged (compaction-as-migration prerequisite) - concurrent-reads-during-update consistent state (no torn reads under stress) - multiple back-to-back updates: final-state observable - **4 integration tests** in `tests/runtime_config_integration.rs`: - Tree starts at default - update visible on next load - snapshot outlives update with pre-update state - runtime config resets to default on reopen (by design — caller re-applies) **Workspace: 1664/1664 tests pass** (+24 new vs prior baseline). Lints clean across all-features lib + tests. Doc tests pass (43). ## Test plan - [x] strict lints clean across all-features lib + tests - [x] full nextest workspace: 1664/1664 pass - [x] doc tests pass (43) - [x] runtime_config unit tests pass (10/10) - [x] runtime_config integration tests pass (4/4) - [x] concurrent torn-read stress test passes (8 readers × 5000 reads + writer) ## Dependency `arc-swap = "1.9"` (std-bound; the `experimental-thread-local` no_std mode is out of scope for now). Handle behind `#[cfg(feature = "std")]`; future no_std consumers can use `spin::RwLock<RuntimeConfig>` as alternative documented in code. ## V5 batch coordination This is V5-1. After merge, V5-2 (#297), V5-3 (#298), V5-4 (#224) can land in parallel worktrees consuming this foundation. release-plz PR #272 stays held until full V5 batch lands — then v5.0.0 ships with the complete integrity stack as one coherent release. See #215 ROADMAP for full V5 plan. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Added runtime-configurable checksum algorithms for block and key-value integrity checks. * New lockless, live configuration API to update and read runtime settings without restarts. * Configuration changes are isolated per tree session and reset to defaults upon tree reopen. * **Tests** * Added integration and unit tests validating runtime configuration updates, atomic snapshot semantics, and persistence behavior. <!-- review_stack_entry_start --> [](https://app.coderabbit.ai/change-stack/structured-world/coordinode-lsm-tree/pull/355?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack) <!-- review_stack_entry_end --> <!-- end of auto-generated comment: release notes by coderabbit.ai -->
](https://app.coderabbit.ai/change-stack/structured-world/coordinode-lsm-tree/pull/355?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack){kind=link}
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
from
e5ddeca to
5e00028
Compare
sw-release-bot
Bot
force-pushed
the
release-plz-2026-05-19T18-47-08Z
branch
from
5e00028 to
59e4a4f
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
🤖 New release
coordinode-lsm-tree: 4.5.0 -> 5.0.0Changelog
This PR was generated with release-plz.