Skip to content

chore: release 7.1.0 - #120

Merged
michael-denyer merged 1 commit into
masterfrom
release/7.1.0
Jul 26, 2026
Merged

chore: release 7.1.0#120
michael-denyer merged 1 commit into
masterfrom
release/7.1.0

Conversation

@michael-denyer

Copy link
Copy Markdown
Owner

Version bump, CHANGELOG section, and uv.lock for 7.1.0. Four PRs since 7.0.0 — #110, #111, #112, #118 — plus five dependabot Action bumps.

Why minor and not major

The one candidate for a breaking change is save_kinship=True without kinship_output_dir now raising. Treated as a fix rather than a contract change, and the CHANGELOG says so explicitly rather than leaving the reader to infer it:

  • The behaviour removed is a silent no-op_computed_eigen_pairs tested save_kinship and kinship_output_dir is not None and skipped the write when only half the pair was set. No caller could have depended on it for output.
  • kinship_output_dir's own docstring already documented it as required when save_kinship is set. Nothing enforced it; that's what changed.
  • LocoConfig first shipped one release ago in 7.0.0, so the window in which anyone could have written that call is a single version.

What makes it a minor rather than a patch is additive public API: LocoConfig.artifact_suffix, eigen_stem(), eigen_paths(), kinship_path().

CHANGELOG additions beyond a straight move

  • The save_kinship entry is reworded with the reasoning above.
  • A new ### Internal heading records the pyrefly burn-down. chore: burn 98 errors off the pyrefly baseline #118 merged without a CHANGELOG entry because it was test-only, but the release notes should account for it.
  • That entry also carries the tooling correction: pyrefly check --baseline prints 0 errors (28 suppressed) and that count is inline suppressions, appearing identically with and without the baseline. I misread it as the baseline size earlier in this work; the note exists so the next reader doesn't.

Verification

Check Result
pytest tests/ 2241 passed, 10 skipped, 3 xfailed
prek run --all-files 25 hooks pass, exit 0
uv lock jamma v7.0.0 -> v7.1.0, uv.lock staged with pyproject.toml
import jamma __version__ == "7.1.0"

After merge

gh release create v7.1.0 — which build-wheels.yml picks up on release: published to build wheels across platforms and publish to PyPI. No dist/* passed to the release, per the repo's publishing notes.

Four PRs since 7.0.0: the thermo-nuclear review follow-ups (#110), zizmor
hardening (#111), the refurb cleanup (#112) and the pyrefly baseline
burn-down (#118), plus five dependabot Action bumps.

Minor, not major. The one candidate for breaking is save_kinship=True
without kinship_output_dir now raising, and the behaviour it removes is a
silent no-op: _computed_eigen_pairs checked the pair and skipped the write
when only half was set, so nothing could have depended on it for output. The
field's docstring already called the directory required, and LocoConfig
shipped one release ago, so the window is a single version. The CHANGELOG
entry says all of this rather than leaving the reader to infer it.

New public API on LocoConfig — artifact_suffix, eigen_stem(), eigen_paths(),
kinship_path() — is additive, which is what makes this a minor rather than a
patch.

Also records the pyrefly burn-down under a new Internal heading, since #118
landed without a CHANGELOG entry, including the note that pyrefly's
"28 suppressed" line counts inline suppressions rather than baseline entries.
@michael-denyer
michael-denyer merged commit b9e338f into master Jul 26, 2026
12 checks passed
@michael-denyer
michael-denyer deleted the release/7.1.0 branch July 26, 2026 22:45
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