Skip to content

release 0.9.0-beta.9: bsv-rs =0.9.0-beta.9 (the security prerelease), with the room made before 0.9.0 - #52

Merged
Calgooon merged 3 commits into
betafrom
repin/bsv-rs-0.9.0-beta.9
Nov 17, 2026
Merged

Calgooon merged 3 commits into
betafrom
repin/bsv-rs-0.9.0-beta.9

Conversation

@Calgooon

Copy link
Copy Markdown
Contributor

The pin of bsv-rs moves from =0.9.0-beta.8 to =0.9.0-beta.9 and the crate cuts 0.9.0-beta.9, so the toolbox's and the CLI's beta lines can follow onto bsv-rs's security prerelease: they pin bsv-rs exactly and a build holds one bsv-rs, so they reach it only with a tracker that pins it. This prerelease is the one the make-room change of PR 49 (7bbce4f, five types #[non_exhaustive], the labels, Verdict::new) rides: its ## [Unreleased] text moves word for word under ## [0.9.0-beta.9] - 2026-11-17, and ## [Unreleased] stays above, empty.

The commits

  • c1c51b1 deps: pin bsv-rs =0.9.0-beta.9: Cargo.toml's requirement and Cargo.lock by cargo update -p bsv-rs --precise 0.9.0-beta.9 alone.
  • fa607cc release: 0.9.0-beta.9: the version in Cargo.toml and the crate's own lock row; CHANGELOG.md.

The lock

Two rows move, each for its own reason:

  • bsv-rs 0.9.0-beta.8 -> 0.9.0-beta.9, checksum 864decc6... -> 4da4ae5e...; its dependency list is unchanged (the pin).
  • bsv-tracker 0.9.0-beta.8 -> 0.9.0-beta.9 (the release).

No other row moves.

The package against the tag

  • bsv-rs-0.9.0-beta.9/.cargo_vcs_info.json in the registry: "sha1": "a6c8676b0ea655cbc53878c46cc6f8a003fd8519"
  • git ls-remote --tags https://github.com/Calhooon/bsv-rs 'v0.9.0-beta.9*': a6c8676b0ea655cbc53878c46cc6f8a003fd8519 refs/tags/v0.9.0-beta.9^{}

They match.

No answer moved

  • This crate's uses of bsv-rs, git grep -n 'bsv_rs::' -- src tests: five lines in four files (src/compact.rs:34, src/evidence.rs:3, src/evidence.rs:115 (a doctest), tests/common/mod.rs:2, tests/transitions.rs:11), each naming bsv_rs::transaction::MerklePath and/or MerklePathLeaf, nothing else.
  • diff -rq of bsv-rs's src between the two registry packages: src/transaction/beef.rs, src/transaction/beef/link.rs, the new src/transaction/beef/link_limits.rs and src/transaction/CLAUDE.md, nothing else. src/transaction/merkle_path.rs is byte for byte the same (cmp). The only mentions of MerklePath in the diff are the linker reading the size of a MerklePathLeaf for its new heap charge.
  • cargo test --locked -- --list, base (7bbce4f) against tip, with each test's target: 87 and 87, diff empty. Each passes at both (87 passed, 0 failed); tests/compact_golden.rs (PR 49's golden) untouched, 11 passed.
  • No source line of this crate changes: git diff origin/beta --stat touches CHANGELOG.md, Cargo.lock and Cargo.toml alone.

The advisory, and whether this crate reaches its door

GHSA-33fm-jcm5-4m3m (High, CVSS 3.1 7.5), read with gh api repos/Calhooon/bsv-rs/security-advisories/GHSA-33fm-jcm5-4m3m: the prerelease range it names runs through 0.9.0-beta.8, patched in 0.9.0-beta.9. In its words, cut: "the #749 link bound (GHSA-gvmh-f5gf-5vjj) meters a BEEF's serialized wire bytes but not the heap each cloned stub re-materializes - the proof a stub carries and the struct backings of its inputs and outputs - so an untrusted BEEF within the lookup resolver's size bound can force a memory-exhaustion denial of service; the fix keeps that wire budget and adds a second, heap budget beside it". The door is linking a BEEF: Transaction::from_beef (the advisory), and Beef::find_transaction_for_signing (bsv-rs's changelog).

This crate does not reach that door. Through bsv-rs it builds a merkle path, reduces it to a root and reads its height; it links no BEEF: git grep -n -iE 'bsv_rs|beef|from_beef|find_transaction' -- src tests, less the five MerklePath lines above, is empty. The one default of 0.9.0-beta.9 that refuses earlier (BeefLinkLimits::max_link_heap_bytes) is on that door. bsv-rs's changelog states no breaking change since 0.9.0-beta.8.

What a host changes at this prerelease

The make-room change, as its Hosts section says (kept word for word):

  • A match on HostAction, Reask, HintStatus or VerdictError gains a wildcard arm that keeps and surfaces what the host does not name (a re-ask by Reask::label, an error as the fault it is), never an arm that maps to HintStatus::Unknown or drops the value.
  • A Verdict built by literal becomes Verdict::new(status) with its other fields assigned after it.
  • A host that stores hints stores the serde form of HintStatus ({"status":"mined","height":900}) and reads it back by it, never by a hand map, and surfaces a status the reader refuses.

The pin itself asks nothing of a host of this crate.

The proof, at the tip fa607cc

CARGO_BUILD_JOBS=2, a private CARGO_TARGET_DIR, cargo 1.98.1 on aarch64-apple-darwin.

## cargo test --locked
test result: ok. 2 passed; 0 failed   (unit)
test result: ok. 9 passed; 0 failed   (arcade_bodies)
test result: ok. 14 passed; 0 failed  (chain_events)
test result: ok. 4 passed; 0 failed   (compact)
test result: ok. 11 passed; 0 failed  (compact_golden)
test result: ok. 5 passed; 0 failed   (labels)
test result: ok. 4 passed; 0 failed   (properties)
test result: ok. 7 passed; 0 failed   (reorg_by_root)
test result: ok. 7 passed; 0 failed   (scenarios)
test result: ok. 4 passed; 0 failed   (splits)
test result: ok. 10 passed; 0 failed  (transitions)
test result: ok. 2 passed; 0 failed   (verdicts)
test result: ok. 8 passed; 0 failed   (doc-tests)
TOTAL passed 87 failed 0
list: 87 base = 87 tip, identical
## RUSTDOCFLAGS=-Dwarnings cargo doc --locked --no-deps
   Generated .../doc/bsv_tracker/index.html
## cargo fmt --all -- --check
fmt exit 0
## cargo clippy --locked --all-targets -- -D warnings
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.86s   (exit 0)
## vectors: chain_events, the eight pinned_* by --exact
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 6 filtered out
## vectors: verdicts arc_p0_2_twenty_bodies_are_only_hints
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out
## cargo build --locked --target wasm32-unknown-unknown --features wasm
    Finished `dev` profile [unoptimized + debuginfo] target(s)
## msrv: rust_version 1.88; cargo +1.88 check --locked --all-targets
    Finished `dev` profile [unoptimized + debuginfo] target(s)
## cargo deny --locked check   (cargo-deny 0.20.2)
warning[unused-wrapper]: wrapper for banned crate was not encountered   (deny.toml:34, secp256k1, absent, as its comment says)
advisories ok, bans ok, licenses ok, sources ok
## cargo semver-checks --baseline-version 0.9.0-beta.8   (0.51.0)
Checking bsv-tracker v0.9.0-beta.8 -> v0.9.0-beta.9 (major change)
Checked 0 checks: 0 pass, 260 skip   (a prerelease step is read as major)
## cargo semver-checks --baseline-version 0.9.0-beta.8 --release-type minor
Checked 202 checks: 200 pass, 2 fail, 0 warn, 58 skip
enum_marked_non_exhaustive: HostAction, HintStatus, VerdictError, Reask
struct_marked_non_exhaustive: Verdict
Summary semver requires new major version: 2 major and 0 minor checks failed
## git diff origin/beta --stat
 CHANGELOG.md | 10 ++++++++++
 Cargo.lock   |  6 +++---
 Cargo.toml   |  4 ++--
 3 files changed, 15 insertions(+), 5 deletions(-)

The semver break is PR 49's five marks, as its changelog text states; nothing comes from the pin. The machete, secrets and actionlint jobs are CI's alone (their tools are not installed here).

Landing

The landing into beta and the tag v0.9.0-beta.9 are the captain's; this PR merges nothing and tags nothing.

🤖 Generated with Claude Code

Calgooon and others added 2 commits November 17, 2026 17:24
…-jcm5-4m3m's fix

The requirement moves from =0.9.0-beta.8 to =0.9.0-beta.9 so the toolbox's
and the CLI's beta lines can follow onto it: they pin bsv-rs exactly and a
build holds one bsv-rs, so they reach the patched prerelease only with a
tracker that pins it.

Cargo.lock moves by `cargo update -p bsv-rs --precise 0.9.0-beta.9` alone:
one row, bsv-rs's version and checksum; its dependency list is unchanged
and no other row moves.

No source line changes. This crate names bsv_rs::transaction::MerklePath
and MerklePathLeaf alone, and bsv-rs's src/transaction/merkle_path.rs is
byte-identical between the two packages; the diff of src is beef.rs,
beef/link.rs, the new beef/link_limits.rs and a CLAUDE.md. The 87 tests
keep their names and their answers (the `--list` of the base and the tip
compared, identical), and tests/compact_golden.rs passes unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ade before 0.9.0

The version moves to 0.9.0-beta.9 in Cargo.toml and in the crate's own
row of Cargo.lock, so the toolbox's and the CLI's beta lines can follow
onto bsv-rs's security prerelease through this crate.

CHANGELOG.md: the text under ## [Unreleased] (the make-room change of
PR 49) moves word for word under ## [0.9.0-beta.9] - 2026-11-17, and
## [Unreleased] stays above it, empty. Beside the moved text, in
0.9.0-beta.8's shape: "Changed: the pin to bsv-rs 0.9.0-beta.9" (no
source line changes, no answer moves, as the runs and the read of
bsv-rs's merkle_path.rs show) and "Security" (GHSA-33fm-jcm5-4m3m, its
door is BEEF linking, which this crate does not reach). The make-room
text's own "Hosts" is kept as it is, last.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Calgooon Calgooon added this to the 0.9.0 milestone Nov 17, 2026
@Calgooon Calgooon added the enhancement New feature or request label Nov 17, 2026
…e restored)

The Security entry gave the advisory summary as its own words and left out
"past the 128 MB Worker isolate" without marking the cut. Restored, so the
quotation is exact.

Co-Authored-By: Claude <noreply@anthropic.com>
@Calgooon
Calgooon merged commit cbc179c into beta Nov 17, 2026
12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant