Dissolution pre-plan backlog — §3 VAGUE+INVALID-GATE not rollable under P1–P10 until concretized [#3243]

[briansrls]

Auto-opened by session-dashboard for session eager-koi-94.
Pushing to session/eager-koi-94 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]

Addressed cursor/composer-2 Practice 9 finding: typescript.dag Status: was split across two // lines; it is now a single // Status: … line (commit 9ce101a8d). Scope / Owns / Consumes / Status remain exactly four one-line fields after the path line.

Re optional note on #3244 plan anchor blocks in json/yaml/toml Deferred sections: agreed those files are already long-form; the anchors are intentional pointers into DECISIONS.md / dissolution-inventory for §3 roll-up traceability, not a move toward minimal prose in those blocks.

[briansrls]

Item 1 (relay of prior thread): Re-verified on current HEAD 9ce101a8d: src/v4/extdeps/languages/typescript.dag lines 2–5 are exactly one path line plus four single-line Scope / Owns / Consumes / Status fields; Status is no longer split. No additional commit needed for this point — it is already addressed by 9ce101a8d.

Merge readiness (re-check): GitHub mergeStateStatus is CLEAN / mergeable MERGEABLE; CI rollup on latest push is green (fmt, affected, ci, v4 succeeded). dashboard-ops reviews for this PR still shows 0/2 distinct dashboard APPROVE verdicts (meets_two_approval_rule: false; the composer-2 run is dashboard_only / verdict: unknown in the summary, and it was recorded against the previous SHA ebb9adb210d2e14766f816b5696ed26ac06624ec before the Practice-9 fix). So per project policy this PR is not merge-ready on approvals yet — waiting on api-review signal, not code or CI.

Per standing instruction: not running gh pr merge here; operator merges manually.

— sent from eager-koi-94

[briansrls]

codex (JSON unique-name “live cite”): Verified on prior HEAD — src/v4/extdeps/formats/json.dag only repeated SL-3229-T4-FORMAT-T6T7 and SL-3229-LLVM-WIDTH after JsonValue, while DECISIONS.md SL-3229-JSON-UNIQUE-NAMES was labeled Gate (live cite, Practice 9).

Fix: caab4c75c adds the missing third gate one-liner (feature text matches the Part 6 row) so JSON matches the yaml.dag “third axis” pattern. docs/audit/dissolution-inventory.md §3.2/§3.3 updated (24 lines, three post-coproduct gates).

CI is re-running on that commit; not claiming full merge-readiness until checks + api-review quorum catch up.

[briansrls]

Dashboard relay: The quoted block is the author’s own follow-up on codex’s JSON unique-name finding, not a new review thread. No further code change: HEAD is still caab4c75c; json.dag carries three post-JsonValue // 🟡 gated lines including SL-3229-JSON-UNIQUE-NAMES, and docs/audit/dissolution-inventory.md §3.2/§3.3 match (24 lines / three gates). GitHub Actions on that commit are green (fmt, affected, ci, v4).

Merge readiness (from gh): mergeable=MERGEABLE, mergeStateStatus=CLEAN, CI SUCCESS. Standard gh pr view --json reviews shows no submitted GitHub reviews from this token’s view — please confirm ≥2 api-review APPROVE (and no REQUEST_CHANGES) in the dashboard / UI; I am not running gh pr merge per operator policy.

— sent from eager-koi-94

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 077d6f46 · Trigger: schedule
• Thinking: 236s wall

---

BLOCKING (1)

Root Cause

• docs/audit/dissolution-inventory.md new format-specific gate was added after the P1/P3 queue was already ranked → add SL-3229-JSON-UNIQUE-NAMES to the appropriate §1.1 row/counts or give it its own explicit owning row before calling it concretized

⚠️ One newly added gate is documented, but not wired into the authoritative burn-down plan.

[briansrls]

codex blocking (schedule @ 077d6f46): Verified — SL-3229-JSON-UNIQUE-NAMES existed in json.dag + §3.2/§3.3 but §1.1 P3 only named SL-3229-T4-FORMAT-T6T7, so the new gate was not on the authoritative ranked queue.

Fix: c09ff9009 expands §1.1 P3 “unblocks” to enumerate P3-axis Part 6 slugs (T6T7 + JSON-UNIQUE-NAMES + YAML canonical keys + TOML datetime / table-syntax), explicitly excluding SL-3229-LLVM-WIDTH (P1). §1.2 P3 row bumps the in-file bucket 7→8 and ties the landing event to T-4.6 format gates; §3.3 headline now splits P1 vs P3 format cites (no accidental “all P3” wording).

[briansrls]

Inline blocking (dissolution-inventory.md:152 / P5): Verified on eff68ef09. Line 152 is the §1.1 P3 row: the unblocks cell already lists SL-3229-JSON-UNIQUE-NAMES (json.dag) next to the other format-axis Part 6 slugs, with owner compiler / T-6, T-7 and 🟡-count 2 + ~8 in-file — so the gate is on the ranked burn-down queue (not orphan debt). The relay text matches pre-c09ff9009 HEAD.

Extra (this push eff68ef09): Added the missing Roll-up: pointer under SL-3229-JSON-UNIQUE-NAMES in DECISIONS.md, mirroring SL-3229-T4-FORMAT-T6T7 → docs/audit/dissolution-inventory.md §1.1 P3, so the DECISIONS ↔ inventory trace is explicit both ways.

Thread item (codex blocking / c09ff9009 note): Same resolution — no further work unless a new finding references a newer SHA.

— sent from eager-koi-94

[briansrls]

openai-pro / gpt-5.5-pro (review 14117, relayed log): The artifact body pasted in the dashboard stops mid–chatgpt-reviewer upload / composer typing (sha=9ce101a8, not current HEAD). There is no substantive BLOCKING finding text after that prefix in the relay — nothing to patch against eff68ef09 without the rest of stdout.log or a fresh run on current HEAD. If a later paragraph in the full log still blocks, please re-relay that section or re-trigger the reviewer on the current tip.

claude-opus-4-7 (14173, APPROVE): Taken as read — no code change from that verdict. On the exploratory note: SL-3229-VERILOG-D3200 already records the merge-base “first consumer” dissolution story as superseded and re-concretized under D2-reversal / #3244 + feature: T-4 Phase-3 (DECISIONS.md §SL-3229-VERILOG-D3200). Other “first consumer” wording elsewhere (e.g. OS-1 coproduct table intro) names the T-4.5 OS seam for filesystem/process carriers — different scope than the Verilog #3200 ledger, not a contradiction with the D3200 supersession block. A repo-wide prose disambiguation sweep is reasonable follow-up, not required to land this PR.

Merge readiness (mechanical gh): mergeable=MERGEABLE, mergeStateStatus=CLEAN, CI green on the workflow that includes eff68ef09. Api-review quorum still needs UI/dashboard confirmation (gh JSON reviews often empty here). Not running gh pr merge per operator policy.

— sent from eager-koi-94

[briansrls]

codex (review 14177, APPROVE on eff68ef0 worktree): Verified — local HEAD is eff68ef09 (same tip family). No further code change: the diff already keeps DECISIONS.md authoritative, adds the json.dag SL-3229-JSON-UNIQUE-NAMES Practice-9 one-liner, wires §1.1 P3 + Roll-up: for that slug, and leaves P10 / INTEGER-GROUP-COMPLETION honestly ⛔ until an owning T-# exists.

Note for dashboard: An earlier codex schedule relay on 077d6f46 was BLOCKING on missing §1.1 wiring; that is addressed in c09ff9009 + eff68ef09. Treat 14177 as the current codex verdict on the tip.

Merge readiness (gh): mergeable=MERGEABLE, mergeStateStatus=CLEAN, CI green on the workflow covering this tip. Standard gh pr view --json reviews still shows older COMMENTED rows on 077d6f46, not GitHub-native APPROVED — please confirm ≥2 distinct api-review APPROVE artifacts (e.g. this codex pass + openai-pro / other) in the dashboard. Not running gh pr merge per operator policy.

— sent from eager-koi-94

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: eff68ef0 · Trigger: schedule
• Thinking: 267s wall

---

✅ No blocking concerns in the changed docs or JSON gate line.

[briansrls]

codex schedule (sha:eff68ef0): Verified — repo HEAD is eff68ef09194fc8d70c1a187833e14ed1619b378 (eff68ef09…), matching the review prefix. No code change: the relay is a clean “No blocking concerns” pass on the current docs + json.dag gate line set.

Merge readiness (gh): mergeable=MERGEABLE, mergeStateStatus=CLEAN, CI green on the workflow for this tip. Treat this artifact as codex’s current api-review approval on eff68ef09 (alongside any other provider rows the dashboard already has, e.g. claude-opus APPROVE). gh pr view --json reviewDecision does not surface GitHub-native APPROVED for these api-review bodies — please confirm ≥2 distinct approved artifacts + no REQUEST_CHANGES in the dashboard before calling quorum satisfied (openai-pro still the weighted signal per team policy). Not running gh pr merge here.

— sent from eager-koi-94

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 69529d1a · Trigger: schedule
• Thinking: 201s wall

---

✅ This mixed docs/model PR keeps the JSON gate as a terse pointer while making DECISIONS.md Part 6 the authority; no blocking concerns found.

[briansrls]

codex schedule (sha:69529d1a): Verified — local/remote tip matches 69529d1a13323e0d8ea57031d7e3447be0d2c6c3 (merge of origin/main into this branch). json.dag still carries three terse post-JsonValue gates including SL-3229-JSON-UNIQUE-NAMES; DECISIONS.md still has the §1.1 P3 Roll-up: for that slug. No code change — codex’s “No blocking concerns” matches tree state.

Merge readiness (gh): mergeable=MERGEABLE, mergeStateStatus=CLEAN, CI green on the workflow run for this tip. Quorum / REQUEST_CHANGES / openai-pro weighting remains dashboard/UI; not running gh pr merge.

— sent from eager-koi-94

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 69529d1a · Trigger: manual
• Comparison: main @ e7b8a8d6 ... session/eager-koi-94 @ 69529d1a
• Conversation: View conversation

---

1. Story of the diff

This PR tries to turn the old dissolution “pre-plan backlog” into named gates: it rewrites docs/audit/dissolution-inventory.md so former §3 VAGUE / INVALID-GATE entries are described as concretized into P1/P3/P4/P10 roll-ups, adds new src/v4/DECISIONS.md Part 6 disposition text for integer group-completion, Verilog D3200, float nominal-width, and JSON unique-names, and adds one new JsonValue gate line in src/v4/extdeps/formats/json.dag:23. The mechanism is a prose authority chain: the audit inventory says DECISIONS.md Part 6 is authoritative, DECISIONS.md carries the detailed gate text, and the .dag file keeps a terse slug pointer back to that decision row. That is the key load-bearing choice, and it is also where the PR is now stale against the current modeling discipline.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

Finding — P2 single authority / substrate-planning surface. docs/audit/dissolution-inventory.md:23 makes src/v4/DECISIONS.md Part 6 authoritative for the concretized gates, and docs/audit/dissolution-inventory.md:31-35 says the single authority path is DECISIONS.md Part 6 plus in-file one-liners. Current docs/modeling-discipline.md:8-17 and docs/modeling-discipline.md:92 explicitly retire src/v4/DECISIONS.md and docs/audit/dissolution-inventory.md as maintained ledgers; the inline model mark plus PR review / allowed non-comment receipts are the authority. This is substrate-facing planning metadata, not implementation-only Rust, so keeping a retired ledger as the authoritative gate source reintroduces parallel authority. chatgpt-review-beda1c0a-7e85-4f…

1. INVARIANTS.md + modeling-discipline.md.

Finding — P2 Boundary Discipline + P5 Progress Is Dissolution. src/v4/extdeps/formats/json.dag:23 adds // 🟡 gated — feature: JSON unique-name object profile parse/emit validation — DECISIONS.md Part 6 · SL-3229-JSON-UNIQUE-NAMES., so the live carrier now points to a decision ledger that the current modeling discipline says must not be maintained as a parallel prose authority. The same authority says every 🟡 needs a concrete gate and bound plan, but that record belongs on the inline carrier / TASKS.md / INVARIANTS.md receipts / PR review, not in a retired DECISIONS.md ledger (docs/modeling-discipline.md:173-190, docs/modeling-discipline.md:560-563). chatgpt-review-21f9a8b4-6d40-41…

1. CODING.md.

N/A — no Rust implementation code, helper shape, method-vs-function choice, or Rust error/result surface changes in this diff.

1. TESTING.md.

N/A — the only .dag change is a comment-only gate line; there is no executable semantic change to exercise with unit/integration tests.

1. LOCKED DESIGN DECISIONS.

N/A — I did not find this PR altering a design item explicitly marked “LOCKED”; the issue is against the current ledger-retirement / single-authority discipline, covered above.

1. TRACKED vs UNTRACKED DEBT.

Finding — incomplete tracked bridge. src/v4/DECISIONS.md:954 adds a 🟡 gate while explicitly stating the owning TASKS.md T-# is not assigned and that no committed substrate PR / owning task exists. docs/modeling-discipline.md:180-185 says the bound dissolution plan must include the substrate PR or task that will land it, and a 🟡 with no bound dissolution PR is not a valid 🟡. Even before the retired-ledger problem, this row is not a merge-valid tracked bridge; it names the missing property rather than supplying it.

2.5. Top-down PM intent review

Finding. The highest-level intent is to reduce duplicate authority and keep cost-of-change at one: the thesis says edits to a compiler concept should stay at one .dag file, with no parallel hand-maintained authority, and the current modeling discipline narrows this specifically by deleting src/v4/DECISIONS.md and docs/audit/dissolution-inventory.md as maintained ledgers (docs/modeling-discipline.md:8-17, docs/modeling-discipline.md:92). This PR lands the opposite shape: docs/audit/dissolution-inventory.md:31-35 declares DECISIONS.md Part 6 plus in-file one-liners as the single authority path, while src/v4/extdeps/formats/json.dag:23 adds a live model comment that points back to that retired decision ledger. A worker following this faithfully would continue maintaining the old ledger path instead of moving the gate record to the current inline/TASKS/PR-review authority surfaces. chatgpt-review-beda1c0a-7e85-4f…

3. Verdict

REQUEST_CHANGES. The PR’s main mechanism depends on DECISIONS.md and docs/audit/dissolution-inventory.md as authoritative ledgers after the current modeling discipline retired those exact docs from that role. Please rebase the gate recording onto the inline carrier / allowed task or invariant receipt surface and assign a real owning T-# for the P10 bridge before merging.