feat: migrate core/behaviors/ Ports (1/5) — trivial base-only reparent - #2125

PR Triage — #2125

Overall: IMPROVE (2 in-scope fixes; no FAILs; CI failures pre-existing on main)

CI Summary

All failing checks are pre-existing on main:

No regressions introduced by this PR. All linters (black, flake8, mypy, pyright), CodeQL, fv demo, and fcv-reject demo pass.

Findings

IMPROVE-1 — CheckCaseUpdateOwnerNode missed in AC-1 sweep

File: vultron/core/behaviors/case/nodes/update.py:40

CheckCaseUpdateOwnerNode has no setup() override and no register_key() calls — it qualifies as Type-A but was left on DataLayerCondition. AC-1 requires every Type-A node under case/ to subclass DataLayerConditionWithPorts.

Fix: change base class + add a NoDataAvailable isolation test and a BTTestScenario happy-path test.

IMPROVE-2 — notes/py-trees-ports-adoption.md stale state section

File: notes/py-trees-ports-adoption.md:38

The "Current state (verified 2026-07-29)" bullet still reads: "A repo-wide search finds zero references to … BehaviourWithPorts in vultron/". This was accurate before the pilot (#1808), false since then, and now further stale after this PR migrates three more domains. Future agents reading this note get an inaccurate picture.

Fix: update the section to reflect: pilot + report/ done (PR #1808), case/status/note done (this PR), report/ and embargo/ Type-A remaining (PR #1809 chain continues).

Code Review

No logic errors, import issues, or behavioral regressions found. Both removed setup() overrides were verified pure no-ops. All 6 intentionally-skipped nodes confirmed to have substantive register_key() calls. 21 new tests pass locally.

Stacking note

PRs #2126 and #2127 fix the pre-existing failures this PR is blocked on. They can merge independently; stacking is not required for this PR's green gate since the failures are not regressions from this diff.

pr-execute — PR #2125

Outcome: SUCCESS — both fix-now findings from triage are fixed and pushed. Zero regressions introduced.

Fixes applied — commit 59439e1d

1. CheckCaseUpdateOwnerNode was a missed Type-A node (AC-1 gap)

vultron/core/behaviors/case/nodes/update.py:41 still subclassed DataLayerCondition
despite having no setup() override and no register_key() calls — its only blackboard
access is the inherited datalayer/actor_id, which makes it a textbook Type-A
base-only reparent. Changed to DataLayerConditionWithPorts and added 4 tests to
test/core/behaviors/case/nodes/test_typed_ports.py:

• NoDataAvailable isolated-port check (BTND-03-011)
• BTTestScenario coverage for case-not-found, actor-owns-case, and actor-is-not-owner

The other three nodes in that module carry substantive register_key() calls and are
correctly left for the later Type-B parts of the #1809 chain.

2. notes/py-trees-ports-adoption.md "Current state" was actively misleading

The section still claimed "A repo-wide search finds zero references to
input_ports, output_ports, BehaviourWithPorts, or PortInformation in
vultron/" — false since the pilot (#1808) landed the *WithPorts base classes.
Rewrote it with verified counts:

• 44 node classes migrated to the *WithPorts bases — case 29, status 7, note 4, report 4
• 168 still on the legacy bases — case 67, sync 32, report 27, embargo 26, status 7, sender 3, inbox 1, plus 5 generic helpers in helpers.py
• 275 register_key() call sites remaining

Also recorded the concrete five-part #1809 split (#1883 → #1887) in the issue sequence,
so the next agent in the chain reads the real decomposition rather than a single size:L Task.

Verification

test/core, test/bt, test/metadata, test/adapters, test/wire, test/architecture,
test/ci and the root-level test modules all pass locally. Run in chunks because the full
suite exceeds the local 120s pytest-timeout ceiling.

CI on head 59439e1d is authoritative: 7526 passed, 362 skipped, 3 xfailed, 1 xpassed,
5552 subtests passed, 2 failed (the pre-existing pair below). That is +25 passing vs
main's 7501 — exactly the new tests.

black, flake8, mypy and pyright (0 errors) clean; markdownlint and the
notes-frontmatter validator clean. xfail ratchet clean — #1991, #1992, #1994 all open.

CI failures — all 6 proven pre-existing

Each classification carries both required proofs:

• Clean-base: all 4 most recent main runs of both python-app.yml and
  demo-integration.yml conclude failure. main run 31261146314 fails the
  identical two pytest node IDs with identical 422 no routable recipients errors.
• Causality: the 25-file diff touches no path matching demo, config, pcr,
  ledger, or chain. It is confined to vultron/core/behaviors/{case,status,note}/,
  test/core/behaviors/, one note, and one plan-history file. The demo failures are the
  exact CLP-08-005 ReconstructChainTail signature from #2120, which no line of this
  diff can reach.

I also reproduced the #2086 pair locally with pytest test/demo -m "".

Issue bookkeeping: PR #2125 added to #2086's Blocked PRs list; handoff comments with
full evidence posted on #2120 and #2121.

⚠️ Two things needing your attention

1. Rebase needed before merge. The branch is 13 commits behind main. Two tests
in test/demo/test_integration_script_scenarios.py fail on the branch in isolation but
pass in CI, because CI tests the pull_request merge commit. main refactored that test
to read .github/demo-scenarios.json, while the branch still carries the version that
regex-greps demo: entries out of demo-integration.yml — and main moved the matrix
out of that workflow. Not a regression; a merge or rebase resolves it.

2. Local git worktree metadata was lost mid-session. This checkout's
.git/worktrees/clyde directory disappeared from the shared .git (only blinky
remains), so every git command in /workspaces/vultron_clyde now fails with
fatal: not a git repository. No work was lost — this happened after the fix
commit was pushed; remote head is 59439e1d, verified via the GitHub API, and all
working-tree files are intact. Per AGENTS.md I did not touch the shared .git
and ran no prune/gc — repair is left for you. This is also why I could not perform
the rebase in item 1 from here.

No new issues, no deferred asks

No NEW-ISSUE findings existed in triage, and PR #2125 has zero review threads.

Artifact: .claude/pr-2125-execute.json

PR Verify: #2125 — feat: migrate core/behaviors/ Ports (1/5) — trivial base-only reparent

Overall verdict: ❌ GAPS-FOUND — not because of this PR's code
CI status: ❌ failing (6 checks, all proven pre-existing on main)
Integrity check: ✅ all 2 findings accounted for

Finding Verdicts

Both code fixes were spot-verified at branch HEAD, not just in the commit diff:

• vultron/core/behaviors/case/nodes/update.py:41 reads
  class CheckCaseUpdateOwnerNode(DataLayerConditionWithPorts):, with the import
  present at line 34. The three Type-B siblings in that module are correctly untouched.
• test/core/behaviors/case/nodes/test_typed_ports.py:146 has
  TestCheckCaseUpdateOwnerNodePorts with all 4 new tests present.
• notes/py-trees-ports-adoption.md no longer contains the finds **zero** claim;
  the new counts and the five-part #1809 split are both present.

All three referenced issues confirmed OPEN: #2086, #2120, #2121.

Why the verdict is GAPS-FOUND

The pr-verify rule is mechanical: CI must be green. It is not, so the verdict cannot
be READY-TO-MERGE. But no failure is attributable to this PR — each has clean-base
and causality evidence recorded in the execute comment, and the diff introduces
+25 passing tests with zero regressions.

🟢 Actionable: stacking is now viable for the pytest failure

This changed since triage — PR #2126 (fix(test): stop demo config cache leaking across tests, Closes #2086) now passes all 10 of its own checks and is MERGEABLE /
CLEAN against main. At triage time it was still failing. Merging it should clear
the Tests (pytest) failure here.

Remaining after that:

So the path back to green main is: merge #2126 → land a working fix for #2120
(#2127 needs more work) → open a fix for #2121.

⚠️ Also needs your attention

Rebase before merge. The branch is 13 commits behind main. Two tests in
test/demo/test_integration_script_scenarios.py fail on the branch in isolation but
pass in CI (CI tests the merge commit): main refactored that test to read
.github/demo-scenarios.json and moved the matrix out of demo-integration.yml,
while the branch still carries the old YAML-grepping parser. Not a regression.

Local git worktree metadata was lost mid-session. This checkout's
.git/worktrees/clyde directory vanished from the shared .git (only blinky
remains), so git commands in /workspaces/vultron_clyde now fail with
fatal: not a git repository. No work was lost — it happened after the push;
remote head 59439e1d verified via the API, working-tree files intact. Per AGENTS.md
I did not touch the shared .git and ran no prune/gc. This also blocked me from
doing the rebase above.

No deferred items

No NEW-ISSUE findings, no deferred-ask items, and zero review threads on this PR.

Artifacts preserved at .claude/pr-2125-{triage,execute}.json — verdict is GAPS-FOUND
solely due to pre-existing CI, so nothing here needs a re-run of /pr-execute.

PR Triage — #2125

Overall: ✅ No findings — clean migration

CI Status

All required checks pass. Five Demo Integration / Invariant Harness jobs fail, but all are pre-existing on origin/main and tracked:

origin/main currently has 15 failing demo jobs; this PR has 5 — it is a strict improvement. These are unrelated to the base-class migration in this PR.

Code Review

• All migrated nodes verified to have no register_key() calls ✓
• Intentionally-excluded nodes (CheckIsCaseManagerNode, EmitCloseCaseNode, BroadcastCaseUpdateNode, etc.) correctly retain legacy base classes ✓
• Legacy imports in partially-migrated files (update.py, vfd_role_guards.py, conditions.py) all have live users ✓
• suggest_actor/emit.py changes are docstring condensation + base-class swap only — no functional changes ✓
• ownership_transfer.py setup() removal is correct — those overrides only called super().setup(**kwargs) ✓
• Notes frontmatter valid; notes/py-trees-ports-adoption.md update is accurate ✓
• All 3 commits have Co-Authored-By trailers ✓
• PR body format correct: Closes #1883 at top, required sections, real test counts ✓

Spec Conformance

BTND-03-009: all migrated nodes subclass DataLayerConditionWithPorts / DataLayerActionWithPorts ✓
BTND-03-010: setup_ports() called via inherited setup() ✓
BTND-03-011: injected values read via get_input() in initialise() ✓
ADR-0044: followed correctly ✓

Test Coverage

21 new typed-ports tests across 3 new test files. Each file has NoDataAvailable isolation tests (BTND-03-011 coverage) and BTTestScenario happy-path tests. AC-4 satisfied ✓

Summary

Zero FAIL or IMPROVE findings. This is a clean homogeneous Type-A base-class migration as designed. Ready to merge pending human review.

PR Execute — #2125

Outcome: ✅ Nothing to fix — triage found 0 findings

Actions taken

• Phase 2 (fix-now): no findings to address
• Phase 3 (CI remediation): all CI failures are pre-existing (#2120, #2121); no branch-owned failures
• Phase 4 (test suite): 6478 passed, 0 failed (362 skipped, 2 xfailed → #1991, #1992)
• Phase 5 (NEW-ISSUE): no findings
• Phase 6 (review threads): 0 unresolved inline comments

Ready for pr-verify.

PR Verify — #2125

Overall verdict: ✅ READY-TO-MERGE

Finding Verification

Triage found 0 findings. Nothing to verify.

CI Summary

All 5 failing jobs are pre-existing bugs tracked on origin/main with open issues. origin/main has 15 failing demo jobs; this PR has 5 — it is strictly better than main.

Local Test Suite

6478 passed, 0 failed (362 skipped, 2 xfailed → #1991, #1992)

Conclusion

Clean migration PR. No code issues, all required CI green, tests pass. Ready to merge.