Release v5.0.0

[1-Bort-1]

TL;DR

Sets Project.toml to 5.0.0 and names the shipping changelog section
## VortexStepMethod v5.0.0 2026-09-07. That section also gains #262's entry,
which sat under the already-released v4.1.0 heading and would otherwise have
been dropped from the release notes; it is prefixed BREAKING:.

Verification

• bin/release's own version guard and notes-extraction stanzas, run
  against the working tree: version guard yields v5.0.0, matching
  Project.toml; notes come out as ### Added (5 bullets) plus
  ### Changed (1 bullet, BREAKING:). On main at 0a8f046 the guard
  yields v4.3.1 — see the surprise below.
• git diff --word-diff over CHANGELOG.md for the latest commit shows
  exactly three tokens: -v4.4.0, +v5.0.0, +BREAKING:. The 13-line
  bullet is rewrapped to 80 columns, and the word diff proves no wording
  moved with it.
• The Continuous wrapping method #262 block this PR moved up is byte-identical to main's (diff,
  exit 0) apart from that rewrap.
• main green: run 34126474537 on 0a8f046, all 7 jobs including
  Documentation. Rebased on main, which has not moved.
• No tests run and no local CI mirror: git diff origin/main -- src test docs README.md is empty, so the tree under test is main's. Nothing in src/,
  test/ or docs/ reads Project.toml's version or CHANGELOG.md
  (grepped). deno fmt not run — deno is not on the box and no workflow
  formats markdown; the changed lines are ≤ 80 columns. REUSE lint n/a.

Reviewer's guide

• Two files, and Project.toml is the whole of one of them.
• In CHANGELOG.md, read the header and the first three words of the
  ### Changed bullet. git show --color-moved=zebra 74c72cf greys the moved
  block; git show --word-diff 5612d24 is the reclassification on its own.
• Risk: the header says 2026-09-07; merging on a later day wants that changed.

Decisions

• v5.0.0, not the v4.4.0 asked for in Release v4.4.0 #276. Bart: "The shrinkwrap being
  continuous makes this a breaking release. It's results are different than the
  gridded approach from before." This repo bumps the major for that — its only
  other BREAKING entries sit under v4.0.0 — so a breaking release off
  v4.3.1 is v5.0.0. Say the word before merging if you meant to keep the
  number and only flag the notes.
• The BREAKING: prefix goes on Continuous wrapping method #262's entry, which is where the break is: the
  wrap is a different shape (a V3 canopy's aft strip comes out 2 * clearance
  thick where the grid gave 3.5 * cell_size), and ShrinkWrap's cell_size
  field is renamed min_clearance — construction still takes cell_size=
  (shrink_wrap.jl:42-44), but sw.cell_size now errors. AGENTS.md §4 spells
  the prefix BREAKING:; the v4.0.0 entries use BREAKING - . Followed the
  rules file.
• Two PRs have merged since v4.3.1, not one. Release v4.4.0 #276 said Give a wing's mesh and airfoil settings a home in the settings file #271 alone and asked
  me to check. Continuous wrapping method #262 is in v4.3.1..main too, and pivot_contour /
  min_clearance exist on main but not at v4.3.1, so it has shipped in no
  release. 536a67f returned its entry to the v4.1.0 section because an earlier
  attempt at that rewrite lived there before 2030908 pulled it out pre-v4.1.0.
  Moving it is the same idea as the rest of this PR, and §4 Releasing step 1
  wants it before bin/release runs — so it is here, not in a second PR the
  release would also wait for.
• Take the panel normal from the quarter-chord step #273 is still open (checked again just now), so it is not in this release and
  its ### Fixed entry stays on its own branch.
• Surprise, filed as bin/release: the version check scans every ## header, so an ## Unreleased section makes it re-register the last release #280: bin/release's version guard scans every ##
  header rather than the top one, so with ## Unreleased on top it matched
  v4.3.1 and would have re-registered an existing tag under the unreleased
  notes. agent release must not run until this PR merges.
• Also filed: docs: the settings file has no reference page — mesh: and airfoil: are reachable only through docstrings #278 (no settings-file reference page, and the README never
  mentions settings files — predates this release, so it did not hold it up)
  and cleanup: CI.yml carries fail-fast: false, so a broken matrix runs every job to the end #279 (fail-fast: false in CI.yml).

Scope

+18 / -15 across 2 files; the bulk is #262's entry moved into the shipping
section. Branch is still named agent/276-release-v4-4-0; renaming it would
rewrite a pushed branch, so it stays.

🤖 Generated with Claude Code

https://claude.ai/code/session_016TgCWZVmVByLWrA5sStHry

Closes #276 · task VortexStepMethod.jl-276

[1-Bart-1]

The shrinkwrap being continuous makes this a breaking release. It's results are different than the gridded approach from before

[1-Bort-1]

Local full suite: PASS (7 min, Julia 1.12.7, one cell of the matrix)

[1-Bort-1]

Independent review (advisory)

Verdict: APPROVE · 0 inline, 0 off the diff

Good

• Diff is one idea and minimal: git confirms only CHANGELOG.md and Project.toml changed (+18/-15, 13 of the deletions the moved block), matching the card's Scope.
• The #262 entry move is byte-identical: extracted the 13-line block from main (lines 330-342) and HEAD (lines 29-41) and diffed them — identical, so the 'pure move' claim holds.
• The move is substantively correct: pivot_contour/min_clearance grep absent at the v4.3.1 tag and present at HEAD, so #262 shipped in no release and its notes were misattributed under the already-released v4.1.0 heading.
• Release tooling now passes: bin/release's guard compares Project.toml (4.4.0) against the first versioned ## header (v4.4.0 2026-09-07), and its awk extraction yields exactly the new section.
• Removal leaves the v4.1.0 ### Changed subsection intact (SolverSettings/AIC/is_show bullets remain at lines 319-355) — no dangling empty heading.
• Card is honest per rule 6: 'No tests run' justified by grep, deno fmt absence disclosed, CI mirror started-not-waited disclosed, and the stale-date risk self-flagged.
• Surprises routed per rubric §5: the bin/release version-guard bug was filed as #280 with an explicit 'do not run agent release' warning instead of being silently worked around.
• Scope call (rescuing #262's entry inside the release PR rather than a second PR) is argued in Decisions against §4 Releasing step 1, not hidden.

Not good

• Header date 2026-09-07 is only right if merged today; the author flags this himself as the one risk.
• Rubric §9 calls Project.toml bumps Bart's job; defensible here because the release was requested and merging makes the bump his, but it is the one rubric line this PR touches.
• The 4.4.0-vs-breaking call on the cell_size→min_clearance field rename is the author's judgment, transparently deferred to Bart — registration is irreversible.

_{opencode, rubric CLEAN_CODE.md. A different lab from the implementer
on purpose: a reviewer sharing its blind spots would not flag its mistakes.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!