Skip to content

release: v3.2.5 - #131

Merged
avrabe merged 1 commit into
mainfrom
release-v3.2.5
Aug 21, 2026
Merged

release: v3.2.5#131
avrabe merged 1 commit into
mainfrom
release-v3.2.5

Conversation

@avrabe

@avrabe avrabe commented Aug 21, 2026

Copy link
Copy Markdown
Contributor

Release prep for v3.2.5 — the patch carrying FEAT-074 (#125) and FEAT-075 (#126), both already merged to main.

What's here

  • Version bump to 3.2.5 across the workspace and every pinned inter-crate dependency — 11 sites in 4 Cargo.tomls. They must move together: cargo publish resolves each leaf from the crates.io index, so a stale version = "3.2.4" pin would publish a 3.2.5 core depending on the previous release.
  • CHANGELOG entry with the falsification statement.
  • README / claims.yaml status line.

The claim-check gate earned its keep

Bumping Cargo.toml turned STATUS-VERSION red immediately:

✗ STATUS-VERSION
    claim evidence absent: 0 < required min 1  [/version = "3\.2\.4"/]
    — the doc asserts it but the source no longer shows it

That is precisely the ~40-version drift from #97 that the gate was built to prevent, caught within one command of being reintroduced. README, the verbatim claim text, and the count-min pattern move together by design. Now 5/5.

Falsification statement

This release is wrong if a module that analysed successfully under 3.2.4 now returns an error or a different verdict set.

Stated with its own limitation: the before/after A/B tested exactly one module (identical results — 8530 advisories, 6490 trap checks, 28 proven-safe). A corpus-wide comparison would falsify it properly, and #126's reporter has offered to re-run theirs.

Known issue shipped with the release

#128 — fixing the panic unmasked a pre-existing operand-stack defect it was hiding. Listed in the CHANGELOG rather than omitted.

Release-machinery finding

While asserting the pre-tag invariants I found that main has no required_status_checks — the CI gate is advisory, not enforced. Filed as #130; it needs repo-admin rights so I haven't changed it unilaterally. Every merge this session was verified by hand instead.

105 core tests + 34 viz · clippy -D warnings clean · fmt clean · rivet validate PASS · claim-check 5/5 · rivet release status v3.2.5 ✓ Cuttable.

Version bump touches the workspace version AND every pinned inter-crate
dependency together (11 sites across 4 Cargo.tomls). They have to move as one:
`cargo publish` resolves each leaf from the crates.io index, so a stale
`version = "3.2.4"` pin on scry-sai-interval would publish a 3.2.5 core that
depends on the previous release.

The claim-check gate did its job here, unprompted. Bumping Cargo.toml made
STATUS-VERSION go red immediately — README still said "v3.2.4 shipped" while the
source no longer showed that version. That is exactly the ~40-version drift
(#97) the gate was built for, caught within one command of being introduced.
README, claims.yaml's verbatim text, and its count-min pattern all move together
by design.

CHANGELOG carries the release's falsification statement, per methodology: this
release is wrong if a module that analysed successfully under 3.2.4 now returns
an error or a different verdict set. Stated with its own limitation — the A/B
tested exactly ONE module, so a corpus-wide before/after would falsify it
properly, and #126's reporter has offered to re-run theirs.

Known issue #128 is listed rather than omitted: fixing the panic unmasked a
pre-existing operand-stack defect the crash was hiding.

105 core tests + 34 viz · clippy -D warnings clean · fmt clean ·
rivet validate PASS · claim-check 5/5.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KkNzkNYzPh7366DkNijeNc
@avrabe
avrabe merged commit 5cd2a89 into main Aug 21, 2026
10 checks passed
@avrabe
avrabe deleted the release-v3.2.5 branch August 21, 2026 12:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant