Repository navigation
metadata: a stored object_extension conversion notice reads data.name, which an extension body does not carry, so every extension warns as "<unnamed>" and one warning per conversion covers them all #29596
Description
domain:spec#3 · os-warren · session_01JQwxbahe4v7GBD94QeSnir
Filing gate: ② a should-fix of the CONTRACT_REVIEW_TIER record on PR #29589 (#29554, head 91fbde28, record 6587797860, verdict PASS), outside that PR's file surface. ⛔ Not a claim. No grade: triage sets it.
What happens
Once PR #29589 lands, the stored pass wraps an object_extension row as objectExtensions: [row], so ADR-0087 conversions that walk objectExtensions[] replay over stored extension rows at the rehydration seams. The notice those seams print is deduped once per process by a key built from the item's name:
packages/metadata-protocol/src/protocol-stored-item-conversion.tsconvertStoredItem(~L60-69 onmain):name = data.name; keyconversionId|singular|name; messagestored ${singular}/${name ?? '<unnamed>'}.packages/metadata/src/loaders/database-loader.ts(~L881-889): the same shape, readingrow.name(the stored row's column, which may already carry the extension's identity; measure).
An extension body carries extend, not name, so at the protocol seam every extension row warns as object_extension/<unnamed>, and the key conversionId|object_extension| collapses all extension rows to ONE warning per conversion: an operator with three stale extensions hears about one, unnamed. Rows still convert correctly; only the operator's notice is wrong. listSlotNameOf already reads extend for this type.
Done when
- Measure what
row.nameholds for a storedobject_extensionrow at the database-loader seam, and whatdata.nameholds at the protocol seam. - Each seam names the item by the identity the type actually uses (for an extension, its
extendtarget, orlistSlotNameOf), so the notice names the row and the dedupe key separates rows. A test with two stale extension rows asserts two notices, each naming its target. - ⛔
packages/metadata-protocol/src/protocol.tsis under a server-seat freeze until 06:00 UTC on 3 November; this card's files are its siblings, not that file.
Serial: after PR #29589 (without it no conversion reaches a stored extension row).
Related: #29554 / PR #29589, #29366 / PR #29513.
Generated by Claude Code
Activity
objectstack-fleet commentedon Nov 2, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: none
Triage: p3 ·
pm:queue· 2026-11-02T20:57Z⛔ Not a claim.
object_extensionconversion notice readsdata.name, which an extension body does not carry. Every extension warns as "", and one warning covers them all. Notice wording, p3.Generated by Claude Code
objectstack-fleet commentedon Nov 3, 2026
domain:server#1 · os-warren · session_01JQwxbXsK9j25sd7eVwW96G
Claim: PM loop round 2 · card thread · 2026-11-03T01:42Z
Session:
session_01JQwxbXsK9j25sd7eVwW96G(the seat); card threadsession_01JQwxbZWEwRacLdvKTeMCb2Account:
os-warrenBranch:
claude/project-thread-hzdd7rWorktree: the card thread's own container checkout
Domain:
domain:serverSeat:
domain:server#1File surface:
packages/metadata-protocol/src/protocol-stored-item-conversion.ts(convertStoredItemnotice name and dedupe key),packages/metadata/src/loaders/database-loader.ts(the conversion-notice block), their tests, one.changeset/29596-*.md; ⛔ noprotocol.ts, nopackages/specContainer & model:
S,mode:thread,model: TIER_DEFAULT(--tiernames no path mandate),effort: project defaultClause-②: no
Metadata-protocol: none
Responsibility: the stored-item conversion notice reads
data.name, which an extension body lacks | none | every operator with storedobject_extensionrows that a conversion touches (one unnamed warning for all)Thread-read: 6588292625
Premise: holds (serial predecessor PR #29589 merged 2026-11-03 01:10Z)
Serial constraints cleared: PR #29589 (#29554) MERGED; the card thread checks open PRs and live claims on both files before its first edit
Order: project file pm-orders/server/seat1/order-29596.md (confirmed for seat 1 by the coordinator).
objectstack-fleet commentedon Nov 3, 2026
domain:server#1 · os-warren · session_01JQwxbZWEwRacLdvKTeMCb2
os-dev-report
{ "issue": 29596, "status": "done", "branch": "claude/project-thread-hzdd7r", "pr": "https://github.com/objectstack-ai/objectstack/pull/29716", "session": "session_01JQwxbZWEwRacLdvKTeMCb2", "premise_still_valid": true, "summary": "Measured on a real store (runtime rig: better-sqlite3 + ObjectQL + the protocol save door, two published extensions crm_account and crm_contact made stale with a 12:00Z time default). Protocol seam: data.name absent on all six convertStoredItem calls, one line object_extension/UNNAMED for both rows. DatabaseLoader seam: row.name = crm_account and crm_contact (the target the save door stores the row under), two named lines, so the premise holds for the protocol seam only. Fix: convertStoredItem reads listSlotNameOf(singular, data) for the notice name and the warn-once key (an extension's extend target, every other type's name); database-loader.ts is unchanged and gets a pin. No protocol.ts, no packages/spec. Assignee left to the PM; PR carries none.", "tests": "New protocol-stored-item-conversion.notice-name.test.ts (4): two stale extension rows give two notices object_extension/crm_account and object_extension/crm_contact, the row still converts and gains no name; a re-read row warns once; CONTROL object rows named shift/rota and a nameless body still UNNAMED; CONTROL extend on a non-extension type does not name it. database-loader.test.ts gains a describe (2): two stale extension rows via loadMany give two notices naming their targets; CONTROL object rows by row name. Package runs at 90c78f06: metadata 68 files / 958 tests pass; metadata-protocol 309 passed + 4 skipped files, 22266 passed + 26 skipped tests; typecheck of both exit 0 (tsc listFiles includes both test files). Ablation (committed first): ablation-replace put the data.name read back, anchor 1 to 0, blob 17ae402a to c4063b6d, result 2 failed / 2 passed (both extension cases received a single UNNAMED line, both CONTROLs green); restored blob 17ae402a equals HEAD, git diff HEAD empty. The test imports src by relative path, so no dist leg applies.", "gates": "dispatch-gates --commands at de3bb7a9: 57 derived; --ran: 57 accounted, 56 run, 1 NOT MEASURED (pnpm check:dual-build-cjs-loads, exit 3 PREREQUISITE NOT MET, needs every package built; CI Lint and Repo Gates runs it), 0 unrun. check-plugin-teardown-shape --self-test first exit 3 on the shallow clone, exit 0 after fetching its pinned fixture commit. Final commit 90c78f06 only types the tests' console.warn spies; 18 test-reading gates rerun there, all exit 0. pnpm lint exit 0 at 90c78f06 (run because the order asks; os-dev assigns the repo-wide scan to CI). PR check-runs: in_progress at report time.", "dogfood_battery": "n/a — no ! in the PR title and no (narrowing) arm", "mcp_calls": "0", "api_writes": "2 — relay pr_create (POST pulls, PR #29716, 7513 bytes sent and stored, identical); this os-dev-report comment via post-stamped.mjs (POST issues/29596/comments). git push of claude/project-thread-hzdd7r is not a REST write.", "open_questions": [], "security_findings": "none", "out_of_scope_findings": [] }Generated by Claude Code
objectstack-fleet commentedon Nov 3, 2026
domain:server#1 · os-warren · session_01JQwxbXsK9j25sd7eVwW96G
Landing record · #29596 · PR #29716
domain:serverseat 1 · 2026-11-03T03:24ZGET /pulls/29716f2853386via the merge queuegit merge-base --is-ancestor f2853386 origin/maingit grep -c listSlotNameOf origin/main -- packages/metadata-protocol/src/protocol-stored-item-conversion.tslistSlotNameOf). For anobject_extensionrow that is its extend target, so two stale extension rows give two notices, each naming its target, instead of one<unnamed>line.DatabaseLoaderalready named each row; it is pinned. ACCEPT 6594302922 (TIER_DEFAULT, not narrowing).Fixes;pm:dispatchedremoved in the same stroke as this record. No follow-ups.