[ChrisRackauckas]

Summary

• Update SciMLBase compat bounds to include v3 (in main and test Project.toml)
• Bump package version 1.0.0 → 1.1.0 (minor)
• No code changes needed - package doesn't use deprecated SciMLBase v2 patterns

Supersedes #92

🤖 Generated with Claude Code

[ChrisRackauckas]

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

Independent review (Devin CLI 3000.11.3, model fusion-claude-opus-5-5-high-sidekick-swe-2-medium): REJECT, risk low.

Full review

VERDICT: REJECT
RISK: low

Blocking findings

1. The PR is superseded and has nothing left to merge. Close it. (Project.toml:3, Project.toml:34, test/Project.toml:19)
   • Confirmed by running: gh pr view 92 -R SciML/ParallelParticleSwarms.jl --json state,mergedAt → MERGED, 2026-04-09T19:12:01Z. This PR was opened at 2026-04-09T12:05:33Z and says it "Supersedes Bump the all-julia-packages group across 2 directories with 1 update #92". That is backwards. Dependabot's Bump the all-julia-packages group across 2 directories with 1 update #92 (commit 508e6b0) made the same compat change (SciMLBase = "2.79, 3.0" in both Project.tomls) and merged seven hours after this PR was opened.
   • Confirmed by running: git log --oneline origin/main -G'^version = "1.1.0"' -- Project.toml → dc7440c "Bump version from 1.0.0 to 1.1.0" is already on main. Tags v1.1.0 through v1.6.4 exist. git show origin/main:Project.toml → version = "1.6.4", SciMLBase = "3.36". Main has gone past both of this PR's edits.
   • Confirmed by running: gh pr view 93 --json mergeable,mergeStateStatus → CONFLICTING / DIRTY. If someone resolved the conflicts by taking this PR's side, version would go back from 1.6.4 to 1.1.0 and the SciMLBase floor would drop from 3.36 to 2.79. The 2.x majors were removed from main on purpose in Drop older SciML major versions from compat #120 and Drop stale [compat] majors older than one year #129.
   • Required change: close as superseded by Bump the all-julia-packages group across 2 directories with 1 update #92 and the later commits on main. No rebase is worth doing, because the rebased diff would be empty.

Non-blocking findings

• The PR body claims "No code changes needed" with no command or output behind it. I tested this myself (confirmed by running, see below): at the PR head, SciMLBase 3.57.0 resolves and test/regression.jl passes. So at the time it was written the claim was correct, but it was unbacked.
• The PR's "CUDA Tests (Julia 1)" check failed. The log has expired (gh run view --log → HTTP 410), so I can't tell whether the PR caused it. GPU Tests on main also fail for v1.6.0 onward. This is moot because the PR should be closed.
• The repo has no AGENTS.md or CLAUDE.md at its root, so no repo-specific rules apply. Per the global rules, the PR body does not have the required agent-attribution banner and footer ("🤖 Generated with Claude Code" only, no model or session).

What I ran

• gh pr view 93, gh pr view 92 --json state,mergedAt,mergeCommit, gh pr view 93 --json mergeable,mergeStateStatus,closingIssuesReferences. No linked issue: closingIssuesReferences is empty.
• git diff origin/main...HEAD, which is 3 lines: the version bump plus SciMLBase compat in both Project.tomls. Also git show origin/main:Project.toml and git show 508e6b0.
• gh pr checks 93 and gh run list --workflow "GPU Tests".
• In a scratch copy of the PR head (git archive HEAD, under scratch/pkg), with Julia 1.11.9: Pkg.resolve() → SciMLBase v3.57.0, RecursiveArrayTools v4.5.3, Optimization v5.9.1, OptimizationBase v5.8.0, DiffEqGPU v3.21.4, SimpleNonlinearSolve v2.14.6.
• In a scratch test env built from the PR's test/Project.toml plus Pkg.develop of the PR head: @safetestset on test/regression.jl and test/reinit.jl → Regression 49/49 pass, Reinit 0 tests (the file defines no tests). Log: scratch/regress.log.

What I did not verify

• gpu.jl, constraints.jl, lbfgs.jl, jet.jl, and the CUDA group at the PR head. I didn't run them because the PR should be closed as superseded no matter what they show, and there is no GPU on this machine.
• Why the PR's CUDA CI run failed (log expired).
• The Downgrade workflow: it triggers on master, but the repo uses main.

---

🤖 Posted by an AI agent — harness: Devin CLI 3000.11.3 (review), Claude Code 2.1.285 (posting) · model: fusion-claude-opus-5-5-high-sidekick-swe-2-medium
Conversation: local Claude session 3c311569-dda6-4687-bc9f-65febb212c33 (fleet master)

[ChrisRackauckas]

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

What was wrong

• The branch was CONFLICTING/DIRTY against main: both Project.toml and test/Project.toml conflicted.
• The PR's content (bump version 1.0.0 → 1.1.0, widen SciMLBase compat to "2.79, 3") is fully superseded: main is already at version = "1.6.4" with SciMLBase = "3.36" (v3 support landed via Bump the all-julia-packages group across 2 directories with 1 update #92 and later releases; the 2.x majors were dropped on purpose).
• The CUDA Tests (Julia 1) check failed on the old head; its log has expired so the cause can't be read, but GPU Tests on main were also failing in the same era and are green on current main — treated as pre-existing/environmental, not introduced by this 3-line compat diff.

What I changed

• Merged origin/main into scimlbase-v3-compat (merge commit 7ca8cf9, parents c12fc88 + a73c0a0).
• Resolved both conflicts by taking main's side: version = "1.6.4", SciMLBase = "3.36" in Project.toml, SciMLBase = "2.79, 3.0" in test/Project.toml. Choosing the PR's side would have reverted the version and re-admitted SciMLBase 2.x.
• Result: git diff origin/main...HEAD is empty — the merged tree is identical to main, so the PR now introduces no changes of its own. This is the honest resolution for a superseded compat PR; whether to merge or close is a maintainer decision.

Validation (all run on Julia 1.12.7, JULIA_NUM_THREADS=4 JULIA_NUM_PRECOMPILE_TASKS=4)

Command	Result
git diff origin/main...HEAD	empty (0 lines)
PPS_TEST_GROUP=Core timeout 3600 julia +1.12 --project=. -e 'using Pkg; Pkg.test()'	PASS, exit 0 — Regression 49/49, Reinit 0 tests, SimpleLBFGS polish 25/25, Algorithm interface 3/3, CPU backend (gpu.jl + constraints.jl) 50/50
PPS_TEST_GROUP=QA timeout 3600 julia +1.12 --project=. -e 'using Pkg; Pkg.test()'	PASS, exit 0 — QA 48/48 (Aqua/JET/ExplicitImports via SciMLTesting)
julia +1.12 --project=@runic -m Runic --check src test	clean, exit 0 (no Julia files differ from main, which passes format-check)
typos (repo, over the diff scope)	clean, exit 0

Pre-existing failures left alone

• CUDA Tests (Julia 1) (GPU Tests workflow, self-hosted V100 runners): the PR-time failure log is expired; GPU Tests on main are green as of a73c0a0. Not runnable in this environment (no free GPU), and moot since the branch's tree is now identical to main's green head.
• No test changes were made; nothing was skipped, silenced, or loosened.

---

🤖 Posted by an AI agent — harness: Devin CLI 3000.11.92 · model: swe-2-max
Conversation: local Devin CLI session 9c68557988420b22a2ac5d527374ed01

[ChrisRackauckas]

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

Closing: this PR changes no files. The SciMLBase v3 compat bump it carried is already on the default branch (the redundant-PR check found an empty diff against it). Reopen if a change is still missing.

---

🤖 Posted by an AI agent — harness: Claude Code 2.1.309 · model: claude-sonnet-5-5
Conversation: https://claude.ai/code/session_011mxnhsFoU1k8SSzfESULEd