[matt-pharr]

Release note

• Audience: developers
• Numerical impact: none (harness @ 1883bee): all 27 cases unchanged; gal_resistive_pe is N/A by design
• Migration: run pre-commit run --all-files before pushing. The pre-commit check now runs on every pull request into develop and main. To bring an open branch in line, see the notes below.

The repository is formatted with the pinned pre-commit hooks, and the pre-commit workflow runs on every pull request, so formatting churn stops showing up in PR diffs. Closes #515.

What changed

Commit	Change
(Pharr's commits)	.github/workflows/format.yaml adds the pull_request trigger for develop and main.
9e57323f9	The output of pre-commit run --all-files on develop at 961484b92, and nothing else. JuliaFormatter 1.0.62 is the version .pre-commit-config.yaml and the workflow pin. 99 files change: Julia formatting, trailing whitespace, and end-of-file fixes (including seven coil-geometry .dat files and Project.toml). Every TOML convention check already passed.
ee76864c6	The workflow comment now says the check is enabled together with the sweep (it used to say it was expected to fail until the sweep). The check also runs on pushes to develop, because two pull requests that are each clean can merge into unformatted code.
f7fe63ce8	docs/src/set_up.md pins JuliaFormatter 1.0.62. The unpinned Pkg.add("JuliaFormatter") it used to suggest installs 2.x, which formats differently and would fail this check on a contributor's first PR. CLAUDE.md's formatting section now describes the pinned version and the CI check.
1883beeb0	.git-blame-ignore-revs lists the sweep commit, so GitHub's blame skips it. Locally, run git config blame.ignoreRevsFile .git-blame-ignore-revs.

How to review the sweep

The sweep commit is reproducible: regenerating it from 961484b92 in a fresh worktree reproduces it byte for byte. To check:

git checkout 961484b92        # develop the sweep was made from
julia -e 'using Pkg; Pkg.add(PackageSpec(name="JuliaFormatter", version="1.0.62"))'
pre-commit run --all-files    # run twice: the end-of-file and formatter hooks interact once
git diff 9e57323f9 -- . ':!.github'   # expect no output

• The coil .dat files: they only lose trailing spaces, and read_coil_dat splits on whitespace, so they read the same.
• This PR's own pre-commit job: it runs here with the new trigger and must pass.

Bringing an open branch in line

1. Merge develop as of just before the sweep: git merge 6cca8d77c.
2. Run pre-commit run --all-files and commit the result.
3. Run git merge -X ours <this PR's merge commit>. Where both sides reformatted the same lines, this keeps the branch's own formatted lines.
4. Run pre-commit run --all-files again; it must come back clean.
5. Check that git diff against develop shows only the branch's real changes.

Done this way, the branch's diff carries no formatting churn. PRs #625, #626 and #627 have already been brought in line this way.

Regression report

Harness on feynman: develop 961484b92 vs this branch at 1883beeb0 (the later commits touch only .github/ and docs)
tj_n1                          25 unchanged
diiid_slayer_n1_toroidal       5 unchanged
gal_resistive_diiid_galinner   11 unchanged
ggj_ray_q500i                  4 unchanged
solovev_kinetic_nuzero         14 unchanged
diiid_error_field              10 unchanged
diiid_slayer_n1                22 unchanged
diiid_truncation_n1            12 unchanged
gal_resistive_pe               8 missing/N/A
tj_kinetic_calculated          20 unchanged
diiid_slayer_n1_coupled        7 unchanged
gal_resistive_diiid            11 unchanged
solovev_n1                     22 unchanged
ggj_reference                  4 unchanged
tj_kinetic_ntv                 7 unchanged
solovev_multi_n                15 unchanged
diiid_n1                       53 unchanged
diiid_slayer_n1_chi_perp_e     22 unchanged
diiid_n1_riccati               17 unchanged
lar_ggj_tearing_n1             3 unchanged
diiid_multi_n                  19 unchanged
diiid_slayer_n1_winding        22 unchanged
efit_fixedbdy_separatrix       5 unchanged
tj_kinetic_multiion            6 unchanged
diiid_kinetic_multi_n          8 unchanged
diiid_n1_riccati_precompiled   17 unchanged
lar_reversed_shear_n1          22 unchanged

gal_resistive_pe reads N/A on both refs by design: its perturbed-equilibrium stage waits on #635.

Notes for reviewers

• Authorship: the trigger is Pharr's, and logan-nc added the sweep. That's why the reviewer should be someone other than either of them.
• Merge method: use a merge commit, not squash, so the hash in .git-blame-ignore-revs stays valid. If it's squashed, .git-blame-ignore-revs needs a follow-up with the new hash.
• Timing: merge promptly once approved. Any later merge into develop makes the sweep stale and means re-running it, which takes minutes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01VdRsGGg4YRG6E9fgLGbpdF

[logan-nc]

@matt-pharr assign someone or mark as draft

[logan-nc]

Measured what this gate would see today, since nobody had posted the size of the sweep it waits for. JuliaFormatter.format(".") at v1.0.62 (the rev pinned in .pre-commit-config.yaml, so the gate and a local run agree) on develop db90e82 rewrites 88 files, +1212/−1064, spread over every module; the hygiene hooks additionally flag trailing whitespace in seven text files (.github/workflows/test.yaml and six benchmarks/*.jl). So the gate as written would fail every PR until that sweep lands, as the comment in the diff says. Sequencing suggestion: the sweep collides with #569 (36 files), #574 (39), #581 (35), #551 (31) and #593 (17), so the cheap order is those five first, then the one-off sweep, then this PR the same day. The diff itself is fine; one nit: format.yaml has no concurrency block while #593 adds cancel-in-progress to test.yaml, so a rapid push sequence will queue formatter runs.

[matt-pharr]

@logan-nc I say this can go in after #602, which can end with a formatter commit.

[logan-nc]

closes #515

[logan-nc]

@matt-pharr Adding the one-off pre-commit run --all-files sweep to this PR, so the gate lands already green and closes #515 together with it. The sweep is a single mechanical commit plus a .git-blame-ignore-revs entry. Your trigger change stays as is. Say if you'd rather keep this PR to the trigger only, and I'll move the sweep out.

[logan-nc]

Review of the sweep and the CI change. No changes requested. Everything I found is already pushed to this branch.

The sweep (9e57323f9) is reproducible. Re-running pre-commit run --all-files (JuliaFormatter 1.0.62) on develop 961484b92 in a fresh worktree reproduces it byte for byte. The only difference is this PR's workflow file, which isn't on develop.

It's layout only.

• The only change inside string literals is five one-line docstrings turned into """ blocks, with the same text.
• No string macros or quoted code were touched, and no [a -b]-style matrix literal changed meaning.
• The seven coil-geometry .dat files only lose trailing spaces, and read_coil_dat splits on whitespace.
• Project.toml only loses two trailing blank lines.

CI: all green except the harness stamp. pre-commit, Tests, Documentation (so the reformatted docstrings render) and both Julia versions pass.

Fixed on this branch:

• docs/src/set_up.md told contributors to run Pkg.add("JuliaFormatter") without a version. That installs 2.x, which would fail this check on everyone's first PR. It now pins 1.0.62.
• CLAUDE.md said the formatter version "is not currently pinned"; it now describes the pinned version and the check.
• The workflow comment said the check was "expected to fail until the sweep lands"; corrected.
• Added a push trigger on develop, so two clean PRs that merge into unformatted code are still caught.

Pending: the harness run, develop vs this branch (feynman job 137273, ETA about 23:30 EST). It should be bit-identical, and I'll add the stamp to the release note when it finishes. The "Title and release note" check is failing only because that stamp is missing.

When merging: add it to the queue, which uses merge commits, so the .git-blame-ignore-revs hash stays valid. Merge it before #625, #626 and #627; they already contain this sweep.