Skip to content

chore(OMN-17292): advance omnimarket contract pin to d2b0d9986ed7 - #4677

Closed
onexbot-occ-writer[bot] wants to merge 5 commits into
devfrom
bot/omnimarket-contract-pin-20261007184111
Closed

onexbot-occ-writer[bot] wants to merge 5 commits into
devfrom
bot/omnimarket-contract-pin-20261007184111

Conversation

@onexbot-occ-writer

@onexbot-occ-writer onexbot-occ-writer Bot commented Oct 7, 2026 •

Copy link
Copy Markdown
Contributor

Automated omnimarket contract-pin advance (OMN-17292).

Advances .github/omnimarket-contract-pin.yaml from f859c7af5e1d2a4df043bbb5d5e9d873a6c1c495 to d2b0d9986ed70f3ddab8b1b7342c935f406abb26, also moves the omnimarket entry of .github/sibling-pins.yaml and the inline workflow ref: shas that read it (OMN-20331), and regenerates the derived application-database TABLE grants and rendered catalogs in the same commit.

This PR is the surface upstream contract drift is supposed to land on. Before OMN-17292 this same drift appeared as a red required check on every open omnibase_infra PR simultaneously, with no action available to their authors (OMN-17141, OMN-16930, OMN-17290).

The advance was proven forward-only against real omnimarket history (scripts/ci/check_omnimarket_contract_pin_advance.py), so it cannot silently revert grants the way the stale pin in OMN-15701 did, and the regenerated output already passes --check --prove locally in the bot run.

Evidence-Ticket: OMN-17292
Evidence-Source: OCC#13210

@onexbot-occ-writer

Copy link
Copy Markdown
Contributor Author

OCC autobind did not mint a companion for this PR: no changed-file candidate could be proven RED against the merge base, and emitting a PR-existence probe instead would be non-falsifiable evidence (OMN-15247). Hand-authored evidence is required.

Considered 13 changed file(s):

  • .github/omnimarket-contract-pin.yaml: contract-pin file with no net-new omnimarket_contract_ref sha absent from the merge base, or the diff changes a path that is not the pin or an output derived from it
  • .github/sibling-pins.yaml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/call-occ-autobind.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/call-occ-companion-effect.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/ci.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/contract-topic-graph.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/contractor-integration-note.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/delegation-consumer-kwarg-parity.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/dispatcher-route-coverage.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/exposure-reader-coverage.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/r1-front-door-probe.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/runtime-rebuild-trigger.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)
  • .github/workflows/skill-node-mapping-sync.yml: no candidate grammar reads this file type (only Python declarations, uv.lock lines, release-artefact lines and contract-pin lines are proposed)

This verdict is occ-autobind's, derived from this head and its merge base alone. A second producer, the occ-companion-effect backstop, consumes its own command on the lab bus and may still mint a companion under the generic binding. Whether it does depends on that consumer running, not on this diff (OMN-18876).

Deleted LEGACY_MIGRATION_TABLE_DECLARATIONS entries for delegation_routing_feedback and delegation_budget_applied_events in src/omnibase_infra/topology/table_grant_derivation.py. Removed both keys from _INTERIM_ENTRIES in tests/ci/test_supplemental_declaration_expiry_omn18863.py. The pinned contract d2b0d9986ed7 now declares these relations via omnimarket#3416 and omnimarket#3436, resolving the test failure.

Regenerating against the pin wrote no instances and no catalogs; --check --prove reports PASS=90 FAIL=0 RESIDUAL=0 for prod, stability-test and test.
…es stay retired

A new integration test file mirrors the existing board probe results bridge test to verify that delegation_routing_feedback and delegation_budget_applied_events no longer have LEGACY_MIGRATION_TABLE_DECLARATIONS entries. The test confirms the vendored migration file remains in the tree while local, onex-dev, and onex-prod topology instances still declare the relation, satisfying the Integration Test Coverage gate for changes to src/omnibase_infra/topology/table_grant_derivation.py.
jonahgabriel pushed a commit to OmniNode-ai/onex_change_control that referenced this pull request Oct 7, 2026
…13210)

* evidence(OMN-17292): author OCC companion for OmniNode-ai/omnibase_infra#4677

OCC companion by node_pr_lifecycle_fix_effect (OMN-13317 F1 / OMN-13990 / OMN-14285). Product PR head 9e6af9926b9e027c0bb587166dedd653dbc5d620.

* evidence(OMN-17292): self-bind OCC#13210 + rebind contract_sha256

---------

Co-authored-by: omnimarket-bot <bot@omninode.ai>
@jonahgabriel

Copy link
Copy Markdown
Collaborator

Superseded by omnibase_infra#4760, the newer PR of the same bot series; closing so it is never landed (drain map 2026-10-09T07:32Z, class superseded-bot).


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant