…() fix (#236) ## Summary - **Dependency refresh**: all crates in `Cargo.toml` / `Cargo.lock` bumped to latest, including major-version bumps that required code changes. - **Adopted upstream bug fix** (fjall-rs#286): `BlobTree::clear()` was corrupting the on-disk tree by skipping the version-manifest update, causing recovery to fail on reopen with *"Tried to open a BlobTree, but the existing tree is of type StandardTree"*. Cherry-picked upstream `db394880`, adapted to our diverged `Memtable::new(id, comparator)` signature and `&*config.fs` argument, plus rewrote the new default `tree_type()` impl in idiomatic `if/else` to satisfy our stricter `deny(clippy::all)` (`obfuscated_if_else`, `unnecessary_lazy_evaluations`). ## Commits 1. `build(deps)` — `cargo update` for ~30 semver-compatible bumps; manifest bumps for `structured-zstd 0.0.12 → 0.0.21` (our maintained fork; each pre-1.0 bump is breaking) and `rand 0.9 → 0.10` (dev-dep major bump — adapted `benches/bloom.rs` to `RngExt::random()` and `Rng::fill()` after `RngCore` was removed from rand crate root). 2. `test(blob_tree)` — regression test reproducing upstream #286: insert into kv-separated tree → `clear()` → reopen fails. Written **before** the fix per the test-first protocol; failed cleanly on stock code. 3. `fix` — cherry-pick of upstream `db394880` adapted to our fork. After the fix the regression test passes. ## Upstream sync analysis Post-divergence commits in `fjall-rs/lsm-tree:main` evaluated for adoption: | Upstream | Adopted | Reason | |---|---|---| | `bad4fe0a` (seqno of point-read ingested items) | ❌ | Our refactor already adjusts `global_seqno` at the `get` / `get_with_block` caller level, so the bug doesn't reproduce here. | | `db394880` + `557cd0db` (clear() blob tree corruption #286) | ✅ | Real bug, also present in our fork — adopted with regression test. | | `05c082ff` + `a8db0880` (`Slice::as_slice()` convenience) | ❌ | Pure API addition; `Slice` already implements `Deref<Target=[u8]>`. Can be added separately if needed. | Open upstream issues triaged: none are confirmed bugs (only #289 — ingestion-pinning enhancement). ## Test plan - [x] `cargo check --all-features --all-targets` clean - [x] `cargo clippy --all-features --all-targets -- -D warnings` clean - [x] `cargo nextest run --all-features` — 1271 passed, 6 skipped - [x] `cargo test --doc --all-features` — 41 passed, 2 ignored - [x] New regression test `blob_tree_clear_then_reopen_succeeds` fails on stock fork, passes after the fix Closes #235  ## Summary by CodeRabbit * **New Features** * AbstractTree trait now provides a default tree type selection based on configuration. * **Bug Fixes** * Fixed BlobTree clear operation to properly reset version history, preventing potential corruption when reopening the tree. * **Chores** * Updated dependencies: structured-zstd and rand. * **Tests** * Added regression test for blob tree clear and reopen operations.  [![Review Change Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/structured-world/coordinode-lsm-tree/pull/236?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack)   --------- Co-authored-by: marvin-j97 <marvin.janke.97@gmail.com>

Uh oh!

Uh oh!

Very small databases are always recovered without pinning after ingestion #289

Description

Activity

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions