R3 gate #4: tier3 effect carrier mirror dissolved

[briansrls]

Auto-opened by session-dashboard for session royal-heron-214.
Pushing to session/royal-heron-214 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: f0cbc5f9 · Trigger: manual
• Comparison: main @ 77678c04 ... session/royal-heron-214 @ f0cbc5f9
• Conversation: View conversation

---

1. Story of the diff

This PR consolidates the Stage 2b workflow-idempotency Rust projection into src/v3/compiler/src/dag/effects.rs, making the effects carrier module the native Dag witness surface for the std.effects algebra rather than keeping a separate effect-carrier mirror. The load-bearing move is that linear workflow idempotency now composes directly over existing OperationEffect / EffectShape carriers via compose_operation_effects, while non-linear workflow forms fail closed into WorkflowIdempotencyReport::IdempotencyUnsupported instead of pretending to have a verdict. The new analyze_workflow entry point reads the existing native Dag projection with lane2_workflow_effect_at and delegates to the same projection logic used by lane2_workflow_idempotency_report, so the API surface is one carrier model with two entry shapes: direct WorkflowEffect and native Dag root.

2. Invariant categories

1. LAYER MODEL — Compliant. This touches a substrate-adjacent carrier module, but it does not introduce a new Dag substrate type, field, or variant; the new code consumes existing carriers (OperationEffect, WorkflowEffect, EffectShape) and existing native Dag lookup (d.lane2_workflow_effect_at(&workflow_root)) at src/v3/compiler/src/dag/effects.rs:220, src/v3/compiler/src/dag/effects.rs:259, and src/v3/compiler/src/dag/effects.rs:296. That matches Boundary Discipline / single-authority use of declared substrate facts rather than a fresh parallel storage shape. chatgpt-review-0cda0389-3d9b-42…
2. INVARIANTS.md + modeling-discipline.md — Compliant. Fail-closed and typed-carrier discipline are handled: missing workflow facts return WorkflowIdempotencyReport::IdempotencyUnsupported at src/v3/compiler/src/dag/effects.rs:297, and unsupported non-linear workflow variants return the same typed unsupported carrier at src/v3/compiler/src/dag/effects.rs:263, src/v3/compiler/src/dag/effects.rs:271, and src/v3/compiler/src/dag/effects.rs:279. Facts flow forward through the existing partition: compose_operation_effects reads EffectShape::IsBreaking(_) directly at src/v3/compiler/src/dag/effects.rs:222 and returns either BrokenBy { first_breaker } or IdempotentComposition at src/v3/compiler/src/dag/effects.rs:229 and src/v3/compiler/src/dag/effects.rs:232.
3. CODING.md — Compliant. The added behavior is data + free functions with explicit dependencies and structured outputs: compose_operation_effects(effects: &[OperationEffect]) -> CompositionVerdict at src/v3/compiler/src/dag/effects.rs:220, project_workflow_idempotency_report(workflow: &WorkflowEffect) -> WorkflowIdempotencyReport at src/v3/compiler/src/dag/effects.rs:256, and analyze_workflow(d: &Dag, workflow_root: NodeId) -> WorkflowIdempotencyReport at src/v3/compiler/src/dag/effects.rs:295. That aligns with the coding rule that lenses/analyses are free functions over data, not methods hidden on Dag. chatgpt-review-56c84d01-3f6e-4a…
4. TESTING.md — Compliant. The diff itself adds no tests, but the changed surface is explicitly described as an emitted-output export exercised by the existing rustc round-trip in m2_lens_idempotency_migration_test at src/v3/compiler/src/dag/effects.rs:237–src/v3/compiler/src/dag/effects.rs:238. Since this PR is relocating/dissolving a mirror rather than adding a new workflow-idempotency semantic, I do not see a diff-citable missing-test finding. The shape also fits the current testing guidance that lens behavior should be pinned through minimal carrier/API surfaces where possible. chatgpt-review-e05961b6-c509-45…
5. LOCKED DESIGN DECISIONS — Compliant. The diff preserves the Pure Bootstrap / self-hosting direction rather than weakening it: the module banner names this as a typed Rust witness for currently unparsed std arrow bodies and says the standalone workflow_idempotency.rs mirror is retired at src/v3/compiler/src/dag/effects.rs:5–src/v3/compiler/src/dag/effects.rs:8. That is consistent with the higher-level target that hand-authored Rust remains scaffold pending dissolution, not a permanent parallel authority. chatgpt-review-1500ddde-ea5d-40…
6. TRACKED vs UNTRACKED DEBT — Compliant. The only bridge-like language is tracked and bounded in the module contract: this is the “typed Rust witness for unparsed std arrow bodies in bootstrap” at src/v3/compiler/src/dag/effects.rs:5, with the dissolution trigger named as the retired standalone mirror / R3 gate at src/v3/compiler/src/dag/effects.rs:6–src/v3/compiler/src/dag/effects.rs:8. I do not see a new TODO, temporary enum, second record shape, or unbounded scaffold introduced in the diff.

2.5. Top-down PM intent review

Compliant. The highest-level intent is to reduce hand-maintained parallel compiler authority and move toward .dag/generated/self-hosted structure; the thesis and Pure Bootstrap docs make “one .dag authority, shrinking Rust scaffold” the direction of travel. chatgpt-review-1500ddde-ea5d-40…

This diff’s concrete mechanism supports that: the module is declared as the “sole compiler-local implementation surface for this algebra” at src/v3/compiler/src/dag/effects.rs:6, and it folds the standalone mirror into the carrier module at src/v3/compiler/src/dag/effects.rs:7–src/v3/compiler/src/dag/effects.rs:8. I do not see a diff-cited semantic dilution where a must-have dissolution target becomes optional or permanent.

3. Verdict

APPROVE. The PR is small and appears to dissolve a mirror into the existing effects carrier authority without adding new substrate variants, duplicate record shapes, or fail-open behavior. The unsupported cases are explicit typed reports, and the native Dag entry point reads the existing workflow projection rather than inventing a second source of truth.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: f0cbc5f9 · Trigger: schedule
• Thinking: 135s wall

---

BLOCKING (1)

Root Cause

• src/v3/compiler/src/dag/effects.rs The projection was copied into the carrier module before the old workflow_idempotency.rs authority was atomically retired → move the implementation and update exports/census in the same diff, or keep one implementation surface.

⚠️ The core logic matches the existing algebra, but the dissolution PR currently adds a second authority instead of completing the retirement.

[briansrls]

BLOCKING: Adding compose_operation_effects here copies the existing Stage 2b authority from src/v3/compiler/src/workflow_idempotency.rs without deleting or rerouting that module/exports, leaving parallel implementations in violation of INVARIANTS.md P2/P5 single-authority dissolution.

[briansrls]

Closing as superseded by PR #2679 (gate #4 tier3 effect carrier mirror dissolved, warm-ibex-579 re-spawn after royal-heron-214 archive). Re-spawn substance (+130/-147) supersedes original (+100/-9). Closed by Director cleanup.