Oxiland 0.12.0 release checklist¶
Status: published (v0.12.0 on crates.io and PyPI)
Tag target: v0.12.0
Release scope¶
0.12.0 closes the performance-optimization milestone: frozen competitive-parity
gate (ADR-028),
production-compile evidence on Linux/macOS/Windows release-default, RSS
budgets, and retained 0.11 parity on the candidate ancestry. See the
0.12 report.
Preflight¶
- Tip CI and qualification green on
mainbefore tagging (CI 30848513524, 0.11 Qualification 30848513591, 0.12 Qualification 30848514245 on57f3d3ddaffd74b022767d928b52e24988a53a5a) - Package and lockfile versions match
v0.12.0(python scripts/check-release-version.py v0.12.0) -
python scripts/check-0.11-release.py(parity retention; 6 profiles, 102 results, 1532 obligations) -
python scripts/check-0.12-release.py(3 profiles, 30 cases, 6 resource checks on revision34c0976fdca813aff7e701a1c2e617ed0832b99f) - Rust format, clippy, workspace tests, MSRV, C ABI allowlist, and docs (covered by tip CI)
- Rust API semver comparison against 0.9.0 as a minor release (covered by tip CI)
-
cargo publish -p oxiland --locked --dry-run - Release workflow enforces
scripts/check-0.12-release.pyforv0.12.*tags - Tip CI diagnostic performance ratios recorded in 0.12 report
The checked boxes record the tip preflight on 2026-08-03. Confirm CI is green on the commit you tag; the tag-triggered release workflow re-runs full CI and the 0.12 checker before publishing.
Publish¶
After the exact release commit is green on main:
The tag-triggered release workflow verifies the tag/version match, enforces
scripts/check-0.12-release.py, reuses the full CI gate, publishes oxiland
and oxiland-cli to crates.io, publishes the install-smoked wheels to PyPI,
and creates checksummed GitHub release assets.
After publish¶
- Confirm crates.io, PyPI, docs.rs, and the GitHub release show 0.12.0.
- Verify
SHA256SUMSagainst the uploaded wheels, crates, and source archive when auditing a specific release artifact set. - Keep tip banners on 0.12.0 until the next development line opens.
Notes¶
The optimized tip beats genuine Redland in all ten corrected strict cases on
macOS/arm64, with paired confidence intervals beyond parity. Tip CI Linux and
macOS diagnostic samples after library-path isolation show the same pattern
(large mutation/parse/serialize wins; ASK latency well under Redland). Cite
those results with host and run scope. Suite-wide faster-than-Redland is
authorized by the ADR-029 nine-cell bundle (0.13 report), not by
this 0.12 release checklist alone. The committed 0.12 release-gate bundle
remains the historical competitive-parity qualification record bound by
check-0.12-release.py.