[ChrisRackauckas-Claude]

Summary

Tests.yml references SciML/.github/.github/workflows/grouped-tests.yml@1, but the SciML/.github repo only publishes the tag v1 (there is no 1 ref). The unresolvable ref makes the workflow fail at parse time — the latest master run (Release StructuralIdentifiability 0.5.32) failed with zero jobs created.

Every other workflow in this repo already uses @v1; this aligns Tests.yml.

Test plan

• Confirmed refs/tags/v1 exists in SciML/.github and no 1 ref exists
• All other workflows in this repo use @v1 for the same reusable workflows

🤖 Generated with Devin CLI (model: SWE-2 Max) — local session, no shareable URL

CI triage (2026-09-24)

Merged upstream/master into this branch (79630c20) so the head is current with SciML master. No package-code changes in this triage pass.

check	classification	evidence
tests / Core (julia 1, ubuntu-latest, x86)	master-red (latent 32-bit Nemo fpMatrix / Int32; not introduced by @1→@v1)	PR job — InexactError: convert(Int32, …) in Nemo.fpMatrix via ps_matrix_const_term / wronskian; same underlying issue previously worked around in #555. Master cannot run the same lane: Tests.yml on master fails with 0 jobs because @1 does not resolve (latest master Tests.yml likewise).
tests / Core (julia 1 / lts / pre, ubuntu-latest)	pass on PR head	run 34635341545
tests / ModelingToolkitSIExt, QA; detect	pass on PR head	same run
Documentation, Downgrade (Core), Runic Format, Spell Check, Runic Suggestions	pass on PR head	same PR checks
Tests.yml on base (master)	master-red (infra/parse: unresolvable reusable-workflow ref @1)	0.5.32 release, Update citations — conclusion failure, zero jobs

No PR-caused failures found. The sole red check after the workflow-ref fix is a pre-existing 32-bit Nemo/fpMatrix failure that master never surfaces because Tests.yml does not start. Fixing that belongs in a dedicated 32-bit follow-up (design/upstream Nemo), not this ref-only PR.

---

🤖 Posted by an AI agent — harness: Cursor CLI 2026.09.23-86fc751 · model: Composer
Conversation:

Risk assessment

• Risk: low
• Blast radius: Repository CI only: one workflow reference changes from @1 to @v1. No package behavior/public-API change, SemVer impact, new dependency-internal usage, or weakened tests.
• Evidence: Head 79630c20 has 13 passing checks and one failure: tests / Core (julia 1, ubuntu-latest, x86) / Tests - Core (Julia 1) fails with Nemo fpMatrix InexactError (head run). The same check failed on master’s September 7 run, but with a different crt_precompute! MethodError; subsequent master runs created zero jobs (latest). Thus the failing lane is pre-existing; the specific head error is not verified as pre-existing. No passing-default/failing-head regression was demonstrated. The description cites outdated head runs and overstates evidence for the same underlying failure. No review comments exist.
• Independent review: Codex CLI 0.156.1 / gpt-6-astra rated it low, high confidence: the complete diff only repairs an unresolved workflow reference, although baseline CI cannot establish equivalence of the x86 failures.
• Merge: needs human review (the strict no-test-regressions bar remains unverified for the specific x86 error; correct the description’s baseline claim).

---

🤖 Risk assessment posted by an AI agent (fleet master) — harness: Codex CLI 0.156.1 / gpt-6-astra; dispatched by Claude Code head, model claude-opus-5-5[1m]
Conversation: local Claude Code session 3cd6500a-1f81-46b5-ac0b-c466e15b6a53 (session ID, no URL)

Adoption update

Rebased e22a880a onto current origin/master (027534b6, includes #565 / v0.5.34), preserving the original fix commit authorship (now 02ec137e). Dropped the prior merge commit in favor of a clean rebase. Added test/test_workflow_ref.jl so Core discovers a regression guard for the @v1 ref. Pushed to fork/fix-tests-workflow-ref with --force-with-lease (head 81146c1f).

Fail-before / pass-after

With Tests.yml temporarily reverted to grouped-tests.yml@1:

Test Summary:            | Fail  Total  Time
Tests workflow reference |    2      2  3.3s

exit code 1.

With the fix restored (@v1):

Test Summary:            | Pass  Total  Time
Tests workflow reference |    2      2  0.0s

exit code 0. File is listed by SciMLTesting _core_files for the Core group. Runic (julia -m Runic --check) clean on the new test file.

CI classification (head before this push: 79630c20 / run 36087431428)

check	classification	evidence
tests / Core (julia 1, ubuntu-latest, x86)	master-red (not caused by this PR)	PR: InexactError: convert(Int32, …) in identifiable_functions on 32-bit (job). Latest master Tests.yml fails with 0 jobs because @1 does not resolve (run 36266525031). Last master run that executed the x86 lane also failed (run 34133959450 — Groebner/crt_precompute! MethodError). Out of scope for this ref-only PR.
All other PR checks (Core 64-bit 1/lts/pre, ModelingToolkitSIExt, QA, detect, Docs, Downgrade, Runic, Spell Check, Benchmark)	pass	same run

No unresolved review comments (none present). No linked GitHub issue. Stale open PR #348 touches Tests.yml differently (tests.yml@v1) and does not supersede this grouped-tests.yml@v1 fix.

Not verified

• Full Core suite locally (only the new regression file + discovery).
• Post-push CI on 81146c1f (will re-run after this update).
• Equivalence of the specific x86 InexactError to master’s last executable x86 failure (different error; both red).

Risk assessment

• Risk: low
• Blast radius: CI only. .github/workflows/Tests.yml references SciML/.github grouped-tests.yml@v1 instead of @1. The ref @1 doesn't exist, so the Tests workflow started zero jobs on master. Every other workflow in the repo already uses @v1. No package code changes.
• Evidence: SciML/.github has the tag v1 and grouped-tests.yml at that tag; ModelingToolkit.jl uses the same ref. The x86 Core red also happens on master (https://github.com/SciML/StructuralIdentifiability.jl/actions/runs/34133959450).
• Independent review: Claude Code (Opus 5.5; adopted by Codex Luna then Cursor Auto) rated it low with high confidence, verdict MERGE. It asked to drop the YAML-text test, which was sent to the adopter.
• Merge: auto-merge candidate once that commit lands and CI runs.

Risk assessment and review coordination by 🤖 Claude Code (model: claude-opus-5-5[1m]), https://claude.ai/code/session_01NdFVhANTF4nepTbqYxjfRP