[ChrisRackauckas-Claude]

🚧 UNREVIEWED — awaiting review by @ChrisRackauckas. Filed by an AI agent running as @ChrisRackauckas; Chris has not seen this.
Harness: Devin CLI 3000.11.228 · Model: swe-2-high
Conversation: local job 104-devin (no session URL)

What changed and why

test/qa/Project.toml pins JET = "0.9, 0.10, 0.11" directly in its [compat], and JET < 0.12 does not support Julia 1.13, so the Tests - QA (Julia 1) lane fails at Pkg.resolve(). This widens the entry to "0.9, 0.10, 0.11, 0.12". Unlike the sibling fix in SciML/DeepEquilibriumNetworks.jl#254, the pin here is the QA env's own compat entry, not LuxTestUtils (this repo does not depend on LuxTestUtils) — confirmed by the resolver message restricted to versions 0.9 - 0.11 by project.

Verification

Fail before (unfixed test/qa/Project.toml, Julia 1.13.1)

In a scratch copy of the test/qa env:

julia +1.13 --project=<scratch qa-env> -e 'using Pkg; Pkg.develop(path = "<repo>"); Pkg.resolve()'

ERROR: Unsatisfiable requirements detected for package JET [c3a54625]:
 JET [c3a54625] log:
 ├─possible versions are: 0.0.1 - 0.13.1 or uninstalled
 ├─restricted to versions 0.9 - 0.11 by project [a160d918], leaving only versions: 0.9.0 - 0.11.6
 └─restricted by julia compatibility requirements to versions: 0.12.0 - 0.13.1 or uninstalled — no versions left

Pass after (Julia 1.13.1)

Same command with the fix resolves JET v0.12.3 (alongside Aqua v0.8.18, SciMLTesting v2.13.3, MATLABDiffEq dev'ed to the branch).

GROUP=QA julia +1.13 --project -e 'using Pkg; Pkg.test()'

Test Summary: | Pass  Total   Time
QA/qa.jl      |   54     54  59.3s
     Testing MATLABDiffEq tests passed

lts (Julia 1.10.12)

The same scratch-env resolve on julia +lts succeeds and picks JET v0.9.18, so the widened range does not break the lts lane.

Not verified

• GROUP=Core (interface + MATLAB runtime tests) was not run locally; it is untouched by this one-line test/qa compat change. No MATLAB or Octave runtime is installed on this machine, and test/matlab_runtime_tests.jl already gates on matlab_engine_available().
• The lts QA env was resolve-checked only, not executed.
• typos over the diff: clean. No Julia source or docs touched, so Runic and the docs build are no-ops.

For the reviewer

• Compat admits 0.12 only, not the already-registered 0.13.x, matching the proven sibling fix; JET 0.13 can be admitted in a follow-up once QA is green on 1.13.
• JET 0.12 on Julia 1.13 produced no new QA findings (54/54 pass), so nothing needed silencing.

Risk assessment

• Risk: low
• Blast radius: test-only [compat] bound in test/qa/Project.toml; no package code or public API touched. Worst case: a JET 0.12 finding appears on CI that the local run did not show.
• Evidence: fail-before resolve error reproduced locally; GROUP=QA 54/54 pass on Julia 1.13.1; lts resolve still picks JET 0.9.18.
• Independent review: none — headless single-agent run; flagged for the fleet review chain.
• Merge: needs human review per draft status; one-line test-env compat change, auto-merge-shaped once CI is green and reviewed.

Closes #104

---

🤖 Posted by an AI agent — harness: Devin CLI 3000.11.228 · model: swe-2-high
Conversation: local job 104-devin (no session URL)

🤖 Automated review section from an AI agent running as Chris Rackauckas's account — not written or reviewed by Chris.

Risk assessment (independent review, Codex gpt-6.1-sol, head 95ff400)

• Risk: low (L3: test-environment compatibility; also matches L5, widening an upper bound without regressions).
• Blast radius: QA dependency resolution only, at test/qa/Project.toml:11; existing JET ranges remain accepted. No public API, numerical behavior, supported-platform removal, new dependency, or licensing change. No new public names requiring documentation or SemVer bump; no skipped or loosened tests.
• Evidence: Read the complete risk policy, linked issue, full diff (git diff origin/HEAD...HEAD), QA runner, and CI configuration; verified HEAD is 95ff4006580a0c47674bafffedef39afe5fb01a7 and origin/HEAD is dfc7eb6573a20fa4f8acf35efc0f89962e6d019b. git diff --check origin/HEAD...HEAD exited 0. Ran the same using Pkg; println("JULIA ", VERSION); Pkg.develop(path=ARGS[1]); Pkg.resolve(); Pkg.status() script against scratch exports of base and head, using timeout 3600, the prescribed depot, four threads, and installed Julia 1.13.1: base exited 1 with JET restricted to 0.9 - 0.11 versus Julia-compatible 0.12.0 - 0.13.1 (no versions left); head exited 0 and selected JET 0.12.3. Ran the same head-environment resolve using Julia 1.10.12 with --compiled-modules=no: exited 0, JET 0.9.18. Commands and output are reproducible with scratch/resolve.jl, the exported scratch/base and scratch/head, and logs scratch/base-resolve.log, scratch/head-resolve.log, scratch/lts-resolve.log. Read checks with gh pr view 105 --repo SciML/MATLABDiffEq.jl --json headRefOid,statusCheckRollup: all 11 reported checks/statuses succeeded. Read base/head run metadata and QA job logs through read-only gh api: base QA fails with the same JET conflict; head QA installs JET 0.12.3 and reports QA/qa.jl | 54 54 46.5s, followed by MATLABDiffEq tests passed. G1: not applicable (no numerical changes). G2: not applicable (no GPU/AD/Reactant/MPI changes); the changed QA environment is exercised successfully by exact-head CI. G3: not applicable (no shared-state/cache/threading/default changes).
• Verdict: MERGE
• Merge: merge

🤖 Posted by an AI agent — harness: Codex CLI · model: gpt-6.1-sol (independent review, submitted by sciml-submaster, Claude Code claude-sonnet-5-5)
Conversation: https://claude.ai/code/session_01379heia6cKWmdtNob6NGzg