chore: promote CHANGELOG for 1.18.0 - #141
Conversation
|
@codex review Release commit for 1.18.0. No code changes — the changelog promotion plus the The one judgement call worth checking is the version number. 1.18.0 was bumped on 2026-08-12 but never tagged, never released and never published, and the eigensolver fix from #140 then landed on top of it in Every step of |
|
Codex Review: Didn't find any major issues. 🚀 Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
@codex review Bumping this one. It is a documentation-only change (CHANGELOG promotion plus the CITATION.cff release date, no code), so if that is why it was skipped, a short confirmation is enough and I will tag from here. |
|
Codex Review: Didn't find any major issues. 🚀 Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
1 participant
Promotes the changelog for the 1.18.0 release.
1.18.0was bumped inpyproject.tomlback on 2026-08-12 but never tagged, never released on GitHub and never published to PyPI, and the solver fix from #140 then landed on top of it in[Unreleased]. Rather than leave a version that exists in the changelog but cannot be installed, this folds the two together: 1.18.0 now carries the issue work and the eigensolver fix, redated to today.No code changes.
pyproject.tomland the__init__fallback were already at 1.18.0, so only the changelog and theCITATION.cffrelease date move.Pre-tag verification
Every step of
docs/release_checklist.rstwas run locally:--strict)TwSSM2ShWARNTwo things worth recording rather than burying.
The one manifest WARN is
bmodes, an archive-only clone with no git HEAD to pin. It is optional and maintainer-local, it pre-dates this release, and every required clone hashes clean.Regenerating the reference decks moved two displayed ratios on
iea34_landby a last digit (56.9to56.8,64.0to64.1). The polynomial coefficient files are byte-identical and every verdict is unchanged, so this is a rounding boundary in a derived display figure rather than a numerical shift, and it gets no Changed entry. The regeneration was reverted so this commit stays limited to the two release files.