Facade codemods: retire the codemod-limit exception rows in the facade gate registry (protocol.ts, engine.ts, sql-driver.ts)

domain:devx#1 · os-support-ai · session_01WGufzgBDcBNVmSmo7VFHpj

Filing gate: ③ a follow-up that two reviews carried. The final contract review of PR #30030 (comment 6628808410) raised a should-fix: about 40 of the 60 exception rows in scripts/facade-delegation-registry.json cite #30020 (state-owning facets), but they are bodies because of codemod limits, not because of facet state. This seat's note on #30020 asked for an owner or a card, and none followed. The #30105 dev reports 25 more rows of the same kind for sql-driver.ts (PR #30139). ⛔ Not a claim. Suggested grade: p2. Route: domain:devx, the lane of the large-file split programme #29465.

Ruling-ref: 6624345766 (item 3: each exception row names the card that will move it)

What is wrong

A facade-gate exception row must name the card that retires it. Rows whose reason is a codemod limit point at #30020, and #30020's design (state ownership) will not remove them. So either nothing retires them, or #30020 grows past its scope.

The rows (reason → count; re-measure on `main`)

protocol.ts + engine.ts (landed via refactor(metadata-protocol): generate the protocol.ts section split with a deterministic codemod #29998 / refactor(objectql): generate the engine.ts section split with a deterministic codemod #30012):
- inferred-return-type 16
- static-reads-class 9
- query-options erasure pin 6
- closed-set 4
- no-host-read 2
- name-taken 1
- host-entry-predicate 1
- field-call 1
sql-driver.ts (PR refactor(driver-sql): put SqlDriver into facade form with the sql-driver codemod (sql-driver.ts 18,334 -> 10,979, body-identical) #30139, when it lands):
- path-reader pin 9
- accessor 5
- knex field-call 4
- inferred-return-type 4
- computed parameter default 2
- comment in parameters 1

Wanted

Re-point every codemod-limit row to this card, leaving state-owning rows on refactor: move shared state out of the protocol.ts and engine.ts facades into state-owning facets, so the hot bodies left on the class can become delegations (ruling 6624345766) #30020.
Per reason, either extend the codemod so the member becomes a delegation, or record why it stays a body. Examples:
- write an explicit return type into the generated facet;
- make the by-path pins read the facet module;
- handle accessors with a get/set delegation.
For each change: byte-identity of moved bodies, public .d.ts EQUAL, the facade gate exit 0 with the row removed, plus the touched packages' suites.

Out of scope (⛔)

Dedupe (open and closed): facade exception, codemod limit, facade-delegation-registry. Nearest: #30020 (state-owning facets), #30053 (docblocks), #30019 (the gate, closed). New.

tooling(facade): card the 25 codemod-limit SqlDriver exception rows on …

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none

Triage: first-touch audit · p2 · `domain:devx` kept · `pm:queue` · 2026-11-04T16:54Z

⛔ Not a claim.

Read fresh. About 40 of the 60 facade-gate exception rows (scripts/facade-delegation-registry.json) cite refactor: move shared state out of the protocol.ts and engine.ts facades into state-owning facets, so the hot bodies left on the class can become delegations (ruling 6624345766) #30020, but they are bodies because of codemod limits, not because of facet state. A further 25 come from sql-driver.ts (PR refactor(driver-sql): put SqlDriver into facade form with the sql-driver codemod (sql-driver.ts 18,334 -> 10,979, body-identical) #30139). Under ruling 6624345766 item 3, each row must name the card that retires it, so this card re-points them. p2, as the programme.
Overlap: refactor: give the inferred-return-type public methods on the engine.ts and protocol.ts facades explicit return types (.d.ts unchanged), so the codemod can move their bodies #30054 (inferred return types) and refactor: sweep the #4918 as any sites in engine.ts's six erasure-pinned methods so the codemod can move them (about 550 lines) #30055 (the query-options erasure pins) already retire two of these reasons on engine.ts and protocol.ts. Re-point those rows at them rather than re-doing that work here. Serial with them on the same files.

Generated by Claude Code

domain:devx#1 · os-support-ai · session_01WGufzgBDcBNVmSmo7VFHpj

Claim: PM loop · card thread · 2026-11-05T00:57Z
Session: session_01WGufzj4B2cbnRLVdgtLcDZ (dev: os-dev card thread; seat session_01WGufzgBDcBNVmSmo7VFHpj)
Account: os-support-ai
Branch: claude/project-thread-6gf6xi (harness-pinned card-thread branch; platform exception to claude/issue-<n>-<slug>)
Worktree: the card thread's own container checkout
Domain: domain:devx
Seat: domain:devx#1
Ruling-ref: 6624345766 (item 3: each exception row names the card that will move it)
File surface: scripts/facade-delegation-registry.json (re-point rows), scripts/codemod-split-sql-driver.mjs and the generated packages/drivers/driver-sql/src/sql-driver.ts plus its facet/host modules (sql-driver reasons only). protocol.ts / engine.ts are measure-only. ⛔ No hand edit of generated output; no edit to the gate script.
Container & model: M, mode:thread, model: TIER_DEFAULT, effort: high
Clause-②: no
Metadata-protocol: none
Responsibility: every facade-gate exception row names the card that retires it, and the sql-driver codemod-limit rows are retired where the codemod can fix them
Thread-read: 6636629362
Premise: holds (88 exception rows on main e826e779: 63 cite #30020, 25 cite #30147; PR #30139 merged)
Serial constraints cleared: no open PR touches the registry or the sql-driver codemod; #30054 / #30055 (same protocol/engine rows) are unclaimed, and this card changes no protocol.ts / engine.ts code

domain:devx#1 · os-support-ai · session_01WGufzj4B2cbnRLVdgtLcDZ
os-dev-report

Phase 1 of 3 (re-point, registry only). Phases 2 and 3 follow on this card.

{
  "issue": 30147,
  "status": "done",
  "branch": "claude/project-thread-6gf6xi",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/30194",
  "session": "session_01WGufzj4B2cbnRLVdgtLcDZ",
  "premise_still_valid": true,
  "summary": "Phase 1: every facade-gate exception row now names the card that retires it. 88 rows before and after, same members, files and order; only card and reason change. Before: #30020 63, #30147 25. After: #30020 17 (state-owning: engine bare-this 12, two constructors, protocol publishPackageDrafts bare this, two hand-series host slots), #30054 22 (every inferred-return-type row, 6 of them on sql-driver.ts, 3 of those with a codemod-limit half), #30055 6 (engine erasure pins), #30147 43. Head 6cc705e9; files_changed: scripts/facade-delegation-registry.json (+121/-121).",
  "tests": "No code change. node scripts/check-facade-delegation-only.mjs exit 0 (3 facades, 88 named exceptions); --self-test exit 0 (36 cases).",
  "gates": "dispatch-gates --commands at 6cc705e9: 20 derived, all run, all exit 0; --ran: 20 derived, 20 run, 0 NOT-MEASURED, 0 UNRUN. Extra: check-published-list-mirrors exit 0, check-facade-delegation-only --self-test exit 0, eslint on the registry exit 0. pnpm lint (repo-wide) left to CI per os-dev.md local-scope rule. CI check-runs: in_progress at report time.",
  "dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm",
  "mcp_calls": "6 — issue_read x6 (reads only)",
  "api_writes": "2 relay writes: pr_create (run 41946870290, read back identical 5167 bytes, label skip-changeset, no assignee) and this comment; plus git push",
  "open_questions": [
    { "question": "#30054's body names only engine.ts and protocol.ts; 6 sql-driver.ts inferred-return-type rows now point at it. Widen #30054, or a sibling card?", "options": ["A widen #30054 by the 6 sql-driver rows", "B file a sql-driver sibling card and re-point"], "recommendation": "A, because the work is the same kind (spell the return type, .d.ts equal) and the order already routes these rows there" },
    { "question": "reconcileViewExpansions (protocol.ts) reads a private static WeakMap. Filed as a codemod limit (#30147), since a statics host can reach it with the map left on the class. Keep, or send back to #30020?", "options": ["A keep on #30147", "B back to #30020"], "recommendation": "A, because the move needs no state rehome" }
  ],
  "security_findings": "none",
  "out_of_scope_findings": []
}

chore(scripts): the six sql-driver.ts inferred-return-type rows go to…

domain:devx#1 · os-support-ai · session_01WGufzgBDcBNVmSmo7VFHpj

Phase 1 landed · `domain:devx#1` · 2026-11-05T02:10Z

Probe	Reading
PR #30194	MERGED 2026-11-05T02:06Z as `f2a9007e` via the merge queue (seat record on `57a3127d`)
`origin/main:scripts/facade-delegation-registry.json`	88 exception rows: #30020 17, #30054 16, #30055 6, #30147 49

State: the card stays open and pm:dispatched stays. Every exception row now names the card that retires it (ruling 6624345766 item 3).

Next: phase 2 retires the 27 sql-driver.ts rows by extending scripts/codemod-split-sql-driver.mjs. It is split into four PRs, each under 3,000 lines with the full #30139 proof set:

2a: codemod only;
2b: the durability-log follower and the compile-refusal seam test;
2c: the generate-* and keyed-text-bounds pins;
2d: applyFilterCondition.

Phase 3, measuring the protocol.ts / engine.ts reasons, follows. Changing those files needs the seat's call.

refactor(driver-sql): the sql-driver codemod retires 15 facade-gate e…

domain:devx#1 · os-support-ai · session_01WGufzj4B2cbnRLVdgtLcDZ
os-dev-report

Phase 2a of 2a-2d (sql-driver codemod extension). 2b, 2c, 2d and phase 3 follow on this card.

{
  "issue": 30147,
  "status": "done",
  "branch": "claude/project-thread-6gf6xi",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/30249",
  "session": "session_01WGufzj4B2cbnRLVdgtLcDZ",
  "premise_still_valid": true,
  "summary": "Phase 2a: the sql-driver codemod learns six mechanisms (receiver-free knex calls, EXPLICIT_RETURN_TYPES, comments in parameters moved verbatim, MOVED_ACCESSORS, computed defaults written optional on the stub, MOVED_FIELD_DEFAULTS) and the tree is regenerated from 21510bba, never hand-edited. 15 sql-driver.ts registry rows deleted: 88 -> 73. By card before/after: #30020 17/17, #30054 16/16, #30055 6/6, #30147 49/34 (sql-driver.ts 27 -> 12). Five staying rows get their reason rewritten (applyFilterCondition, ensureShardTable, initObjects, get supports, static DIALECT_CONNECT_TIMEOUT). Head 34fc09d0, base f2a9007e. files_changed: 13 (codemod +484/-37, registry, check-tenant-chokepoint +1, sql-driver.ts and 8 modules, new sql-driver-schema-introspection.ts); 1,551 changed lines, under 3,000.",
  "tests": "codemod --self-test 98 cases in 6 batteries exit 0 (new battery 'codemod-limit retirements (#30147)' 31 cases, floor 5 -> 6). Base f2a9007e and head equal: driver-sql 238 files/12 skipped, 3,859 passed/216 skipped; driver-turso 97 files, 2,514/34 skipped; driver-sqlite-wasm 40/750; cli by-path 9 files/116; types 53, metadata 11, metadata-protocol 33, spec check-liveness 74; every leg os-verify-lock VERDICT command-exit 0.",
  "proofs": "--verify --base 21510bba exit 0: 171/171 bodies, 127/127 top-level identical, undo rebuilds base, 18334 -> 10717 lines. Determinism: two --write runs tree sha256 07b3a505...6454 both. Public .d.ts: dts-public.mjs exit 0 pass true, index.d.ts raw-equal (sha256 92e2cfb2... both sides); rotateShards and ensureRotation specifically read nowMs?: number at base and head, signature hashes c8bded77 / d0ab9226 equal. The four EXPLICIT_RETURN_TYPES strings were copied from the base build's dist/index.d.ts, not from hand reading. typecheck exit 0. check-tenant-chokepoint 24/21/3/4/0 at base (38 files) and head (39 files), --self-test exit 0. Facade gate exit 0 with 73 rows, --self-test 36 exit 0.",
  "gates": "dispatch-gates --commands at 34fc09d0: 67 derived, all run, all exit 0 (two first exited 3 on prerequisites: shallow clone lacked the plugin-teardown fixture commit, dual-build-cjs-loads lacked dists; fetched/built, then exit 0); --ran: 67 derived, 67 run, 0 NOT-MEASURED. Extra: check:lean-entry-closure 0, eslint on changed files 0. pnpm lint repo-wide left to CI.",
  "deviations": [
    "scripts/check-tenant-chokepoint.mjs: one DECLARED_WATCH_HINTS line for the new module (its self-test requires it); not the facade gate",
    "codemod SELF_TEST_BATTERY_FLOOR 5 -> 6 so the new battery cannot silently drop"
  ],
  "dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm",
  "mcp_calls": "2 — pull_request_read x1, subscribe_pr_activity x1 (no GitHub MCP writes)",
  "api_writes": "2 relay writes: pr_create (run 41961803947, read back identical 8249 bytes, label skip-changeset, no assignee, draft) and this comment; plus git push (force-with-lease over my own WIP commit 84480890)",
  "open_questions": [
    { "question": "get supports (inferred object type with per-key docblocks) and static DIALECT_CONNECT_TIMEOUT (static initializer, TDZ) stay on #30147 with no 2b-2d slot. Fold into 2d, or leave for a later card?", "options": ["A fold into 2d", "B leave on #30147 after 2d and say so in the phase 3 report"], "recommendation": "B, because neither is a mechanism 2b-2d already builds, and both are small" }
  ],
  "security_findings": "none",
  "out_of_scope_findings": []
}

fix(metadata-protocol): host-split refuses an instance method read as…