Skip to content

disposition-carrier plan: total-migration scope (_contracts coproducts + DeclarationRef→NodeRef grounding) - #5754

Merged
briansrls merged 11 commits into
mainfrom
session/quiet-lynx-746
Jun 24, 2026
Merged

briansrls merged 11 commits into
mainfrom
session/quiet-lynx-746

Conversation

@gunbai-bot

@gunbai-bot gunbai-bot Bot commented Jun 24, 2026 •

Copy link
Copy Markdown
Contributor

Works the total migration into the disposition-carrier plan tracker (operator directive: "i need a total migration/cleaned up"). The plan already covered the data:String fleet + the NodeRef-bind intent; two gaps are now captured. Authority is dsl/gunbc/plans/disposition_carrier.dag; docs/plans/disposition-carrier.md is its regenerated projection (drift gate green).

§6 — total migration surface. All five disposition regions in one table, ordered by what the substrate can enumerate now. Names the 20 _contracts CoproductWireContracts (0/20 tagged, across docker/github×6/llm×2/systemd) as the recommended next ratchet region — buildable today via enumerate_coproduct_decls(), no gunbc#5364 dependency. Answers the operator's "is tagging coproducts in _contracts required?" → not enforced yet; that is the #1 SubstrateMandatoryTag end-state, reached region-by-region, not by flag-day.

§7 — symbolic binds. Records the carrier drift: landed bind: DeclarationRef (stringly {module_path, decl_name, field}) vs §1's bind: NodeRef intent. Why it shipped stringly (field-granular resolution gated on gunbc#5364). The enforcement gap the operator flagged ("a lens to enforce that?") — no lens resolves a DeclarationRef today. Two-step grounding: (1) ungated resolution lens now over enumerated coproduct decls (RED on a dangling decl_name; the discriminating consumer that makes the _contracts migration non-inert); (2) DeclarationRef → resolved NodeRef, unresolvable-is-unwritable, once #5364 lands.

§1 carrier note + §5 sequencing updated to point at both.

Coordination: the disposition-carrier plan + std/disposition.dag carrier are the active slice-1 of neat-dove-397 / fierce-crane-13 (DESIGN-load-bearing, checkpoint to bright-stag-194). This PR edits only the tracker doc (not the carrier), so the owning lane reviews the added scope rather than getting a silent conflict. Notified.

Verification: gunbc compile --source-root dsl = 0 diagnostics; run_generated_artifact_drift_gate = ExitSuccess.

🤖 Generated with Claude Code

@gunbai-bot gunbai-bot Bot changed the title clean up comments as strings disposition-carrier plan: total-migration scope (_contracts coproducts + DeclarationRef→NodeRef grounding) Jun 24, 2026
@gunbai-bot
gunbai-bot Bot marked this pull request as ready for review June 24, 2026 20:22
briansrls and others added 6 commits June 24, 2026 20:28
…hority

#5745 hand-added the `→ [host-effect orchestration plan]` pointer to the
generated ROADMAP.md without modeling it in roadmap_authority.dag, so the
generated_artifact_drift_gate failed fleet-wide (emit(.dag) lacked the line
the committed .md carried) — main has been RED since that merge. An earlier
auto-committed regen on this branch "resolved" the drift by deleting the line,
which would in turn orphan docs/plans/host-effect-orchestration.md.

Fix the authority, not the projection: add the pointer as a section_prose
element immediately before the Host-operation band group (its text reads
"G1–G3 below are its consumers", which now directly follow it), then regen
ROADMAP.md from the .dag. emit(.dag) == committed .md → drift gate green and
the host-effect plan doc keeps a modeled inbound link.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…merge

The merge with main brought #5762's structure (Host-operation band label as a
section_prose, host-effect pointer below it, empty group label). Conflict was
resolved to main's side; regenerate ROADMAP.md from the .dag so emit == committed
and the generated_artifact_drift_gate stays green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@briansrls
briansrls merged commit 6ef1961 into main Jun 24, 2026
2 checks passed
@briansrls
briansrls deleted the session/quiet-lynx-746 branch June 24, 2026 23:30
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