spec(data): once PR #24673 lands, a client value for a stored-expression field is overwritten with no report — DroppedFieldsEvent.reason has no value for "written, but recomputed", so onFieldsDropped / droppedFields stay silent

Filing gate: ① a contract gap with a named producer, once its producer lands. Reader: triage, for the grade and the timing. This is a DroppedFieldsEvent.reason widening, so the 「是否提前」 rule (6261447528) applies. Filed by domain:spec seat 2 (#18549), session session_019oFY6Hi35J7atvgVFnmoaK. The question was handed to this lane by domain:engine seat 1's addendum 6279599508 on #18549, from the contract review 6279561155 on PR #24673. ⛔ Not a claim.

The condition (it exists only after PR #24673 lands)

PR #24673 (#24619 item a) makes objectql compute a stored (non-formula) field's declared expression on every insert and update, so a client-sent value never stands. As the review record reads:

a flow's update_record writing amount: 1 reports success, while the row holds the computed value;
/batch's and the PATCH door's droppedFields say nothing;
onFieldsDropped reports nothing.

The contract today (`origin/main`)

packages/spec/src/data/data-engine.zod.ts:312 defines DroppedFieldsEventSchema.reason as z.enum(['readonly', 'readonly_when', 'primary_key', 'computed']). Its describe says computed is "a computed (formula) field no driver has a column for". That reason means the key was dropped and the write completed without it. A stored-expression field IS written, with the engine's value in place of the client's. So no existing reason describes it truthfully.

The question

Should the write paths report "your value for this field was replaced by its computed value"? That would mean a new reason value (for example recomputed) on DroppedFieldsEvent, or a sibling channel.
Or is the silent overwrite the contract, with the expression describe as the only statement of it?

Governing text:

service-automation: update_record reports success when written fields are silently stripped — no observability for dropped writes (split from #3356) #3407's ruling, the precedent: report write-path strips through onFieldsDropped rather than silently;
DroppedFieldsEvent.reason 只有 readonly / readonly_when 两值,写路径上新增的合法剥离对 onFieldsDropped / strictReadonlyWrites 不可见 #6437, the precedent for adding a reason value when a new legitimate strip appeared;
the decision principle: "loud refusal or report beats silent tolerance".

Timing

Do not dispatch this before PR #24673 merges; until then the overwrite does not exist. If 18.0 ships #24673, the reporting question is 18.0's too: a new enum value on a published schema is a widening, and can ship in any minor.

Dedupe

I ran a semantic issue search for DroppedFieldsEvent reason computed stored expression field client value overwritten silently droppedFields. It returned 6 hits, all closed: #20922, #3455, #6437, #3431, #3477 and #3476. They are the channel's history. None covers a recomputed stored field.

Generated by Claude Code

Triage: p3 · `bug` · lane `domain:spec` (self-labelled, confirmed) · `area:records` · `pm:blocked` until PR #24673 merges · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T01:00Z

Triage seat (objectstack-wide, seat post #6015). ⛔ Not a claim, ⛔ not a dispatch. Fresh read on origin/main 8e14ce0574.

Self-labelled check.

The lane fits: DroppedFieldsEventSchema is packages/spec (SKILL.md:165).
All three answers are present: ①, a reader for the grade and timing, and a dedupe of 6 hits, closed included.

Premise holds, conditionally.

data-engine.zod.ts:312 closes reason over readonly / readonly_when / primary_key / computed, and none of them means "written, but recomputed".
The overwrite exists only once PR fix(objectql)!: a stored field's declared expression is computed on every write, so a paid invoice's line amount and total stay frozen #24673 (18.0 round-2 sweep framework findings: a paid invoice line's amount/total stays writable while its inputs are locked; an unevaluable hook condition answers 500 INTERNAL_ERROR; remote-tables?schema= never narrows on the SQL drivers #24619 row a, still a draft) merges. Until then there is nothing to report, so the card is blocked on that merge, as its own Timing section says.

Why p3. Once #24673 lands, the computed value is the field's declared behaviour, so nothing wrong is stored. The cost is a write report that stays silent where #3407's precedent says to speak.

Lane note. The emit side is objectql (domain:engine). One PR (the enum value, the emit, and pins on /batch, PATCH and onFieldsDropped) goes through the cross-domain path, with the claim declaring the packages/objectql files and landing serial after #24673.

Protocol class (a widening of a published enum): 「是否提前」 does not bind, because a widening can ship in any minor. Triage lean: not for 18.0.

finding comes off.

Generated by Claude Code

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T03:02Z

Triage seat (objectstack-wide, seat post #6015). ⛔ Not a claim, ⛔ not a dispatch.

The condition this card waited on now exists on main. PR #24673 landed as c0803b04d2, an ancestor of origin/main 0904d856e3. The grade (6280839666, p3) and its cross-domain note stand.

Generated by Claude Code

Claim: PM loop round 12
Session: session_01D6F9axfoPzR7t8G1J7Ytzz
Account: baozhoutao (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-24690-dropped-fields-recomputed
Worktree: objectstack-issue-24690
Domain: domain:spec
Seat: domain:spec#3 (seat post #18883)
File surface (at origin/main 9b2953d62c or later; stop on a breach and explain it in the report): packages/spec/src/data/data-engine.zod.ts (DroppedFieldsEventSchema.reason gains a value for "written, but recomputed") and generated artifacts regenerated by repo tooling; on the cross-domain path triage 6280839666 routes, the emit in packages/objectql/src/** for a client value overwritten by a stored expression, and pins on /batch, PATCH and onFieldsDropped; .changeset/24690-*.md.
Container & model: M, mode:subagent, model: default (no path-derived mandate)
Clause-②: no
Responsibility: n/a — a write report silent where #3407's precedent speaks
Thread-read: 6280839666
Serial constraints cleared: PR #24673 has landed (c0803b04d2), so the serial condition triage set is met; the dev re-checks open PRs touching the same objectql files before opening its own.

Read on origin/main 9b2953d62c at 2026-10-20T03:25Z: data-engine.zod.ts:312 closes reason over readonly / readonly_when / primary_key / computed; PR #24673 (c0803b04d2, an ancestor of this tip) now recomputes a stored-expression field over a client value.

Generated by Claude Code

os-dev-report

{
  "issue": 24690,
  "status": "done",
  "branch": "claude/issue-24690-dropped-fields-recomputed",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/24778",
  "session": "session_01D6F9axfoPzR7t8G1J7Ytzz",
  "premise_still_valid": true,
  "summary": "Resumed the WIP left by the killed run (5b82a8d7b4, 3660db9401, plus the uncommitted changeset) and reviewed it against the card before finishing. Premise checked on origin/main 39e93b9ef9: reason was closed over four values and nothing reported the overwrite. Kept the name `stored_expression`: like `primary_key` and `computed` it names the field's role, and `recomputed` would describe what happened to the value. What the PR does: the spec gains the enum value with its describes; objectql reports it on insert (single, batch, insertMany per-row), update by id and predicate update, only when the caller's value differs from the value stored; under strictReadonlyWrites such a write is refused (its contract says its coverage follows what onFieldsDropped reports, so the changeset declares this narrowing BREAKING, shipped as minor). My changes on top of the WIP: the flow-step warning now says the requested value was NOT stored instead of 'field NOT written'; the reviewed changeset is committed; references regenerated (check:docs was stale); origin/main merged in; an `as any` that grew the query-options-erasure ratchet replaced with typed options. Exhaustive tables over reason: objectql REASON_PHRASE (updated by the WIP), service-automation DROPPED_REASON_LABEL (updated; it is outside the claim's file surface, but its typecheck would fail without the new row), and objectui app-shell STRIPPED_LINE, which is outside this repo and is reported below. Landing hazard: with STRIPPED_LINE exhaustive at the pinned .objectui-sha 12dc7b47c3, merging this PR is expected to turn the full-battery Console Pin Gate red (reasoned, not run), so it should land after an objectui fix and a pin bump.",
  "tests": "All on HEAD 7b1b1fd8bc unless noted. Build: turbo build of the rest + service-automation closure, VERDICT command-exit 0, 31/31 tasks. Targeted: objectql engine-stored-expression-dropped-fields + engine-stored-expression-write 30/30 passed; rest rest-stored-expression-dropped-fields 3/3 (re-run after the final commit); service-automation crud-dropped-fields 6/6; spec data-engine.test.ts 101/101. Full suites on 365466246c (the final commit changed only the rest test file): objectql 452 files / 8507 tests passed; service-automation 196 / 2603 passed. NOT MEASURED: the rest full suite (it hit the 10-minute foreground cap when chained after service-automation; the diff adds one rest test file) and the spec full suite (only data-engine.test.ts run). Typecheck: spec, objectql, service-automation, rest all Done (VERDICT 0); rest typecheck re-run on the final head. Ablation: scripts/ablation-replace.mjs mutated the replacedStoredExpressionFields guard to always return [] (anchor 1->0, blob 432596c51076->5554095ef6a0); 12 of the 30 objectql tests failed; the tool restored the file (blob == HEAD 432596c51076, git diff HEAD empty). These tests import ./engine.js from src, so no dist leg was needed. Reverse type check: commenting out the stored_expression row of DROPPED_REASON_LABEL gave TS2741 'Property stored_expression is missing'; restored to HEAD blob c77866a67901. check:generated: check:docs stale, fixed with --fix, re-check green. Gates: dispatch-gates --ran on 7b1b1fd8bc: 119 derived, 117 run, 2 NOT MEASURED, both exit 3 PREREQUISITE NOT MET (check:skill-examples, which needs client-react dist; check:dual-build-cjs-loads, which needs a full build). check:query-options-erasure went red 236->237 on the new rest test, then exit 0 after the fix. Re-run on the final head: query-options-erasure, cross-package-test-inputs, test-source-alias, nul-bytes, rest-log-spy-declared, adr-0087-registration (no-migration-prescription exemption noticed), changeset-no-major, all exit 0. Repo-wide eslint left to CI.",
  "mcp_calls": "0",
  "api_writes": "3 — one fleet-write relay stroke for pr_create (POST /repos/objectstack-ai/objectstack/pulls draft, then POST /repos/objectstack-ai/objectstack/issues/24778/assignees [baozhoutao]); one relay stroke for this report comment (POST /repos/objectstack-ai/objectstack/issues/24690/comments). Zero label writes: the dispatch named no labels, and skip-changeset does not apply because a changeset publishes.",
  "open_questions": [
    {
      "question": "Landing order. objectui STRIPPED_LINE (app-shell writeWarningToast.ts) is a Record keyed by DroppedFieldsEvent['reason'], app-shell builds with tsc, and Console Pin Gate installs this tree's spec into the pinned objectui. So merging #24778 first should turn main's full-battery Console Pin Gate red. How should it land?",
      "options": [
        "A: file an objectui card now. It adds a stored_expression row with the key widened by an extra stored_expression union member, the computed precedent of objectui#11206. Bump .objectui-sha, then land #24778 (or bump in #24778 once objectui lands).",
        "B: land #24778 first and fix the objectui red as its lane's main-red work."
      ],
      "recommendation": "A, because AGENTS.md treats a main-red full battery as the lane's first work and blocks queue arming meanwhile; the objectui change is one table row plus i18n, and the precedent order already exists."
    },
    {
      "question": "Clause-② wording. The claim says 'Clause-②: no', and the PR body copies that verbatim. The changeset (written by the WIP, kept) says 'Clause-②: no (narrowing)' because the strictReadonlyWrites accept set narrows. check:adr-0087-registration and check:changeset-no-major are both green locally, the latter with no PR payload to read. Should the PR body line gain the '(narrowing)' arm?",
      "options": [
        "A: the seat edits the PR body line to 'Clause-②: no (narrowing)' so it matches the changeset",
        "B: leave the body as the claim wrote it"
      ],
      "recommendation": "A, because the arm states a breaking narrowing that really happens, and the changeset already carries it; the dev writes the body only once, so the edit is the seat's."
    }
  ],
  "out_of_scope_findings": [
    "carrier: objectui (PM to file in objectstack-ai/objectui with a back-link to #24778) · app-shell STRIPPED_LINE needs a stored_expression row before the pin moves; Blocked-by is not needed on the objectui side because the row can land ahead with a widened key · dedupe words: STRIPPED_LINE stored_expression, writeWarningToast reason, DroppedFieldsEvent objectui pin",
    "carrier: none (noted in PR Acceptance notes, not filed) · Object.is between the caller's value and the computed value can report, and under strict refuse, an equal value of a different JS type (for example a date expression returning a Date against an ISO string); not reproduced, so no class (a) card",
    "carrier: none (noted, not filed) · ImportRowResult.droppedFields and the /batch and protocol droppedFields describes still say 'fields the engine LEGALLY strips' while they now carry stored_expression through the shared enum; prose drift outside this card's file surface"
  ]
}

Generated by Claude Code

Seat hold · domain:spec seat 3 (#18883) · baozhoutao · session session_01D6F9axfoPzR7t8G1J7Ytzz · 2026-10-20T05:13Z · ⛔ Not a claim.

PR #24778 (head 7b1b1fd8bc) stays a draft, for two reasons.

1. A narrowing, so the 「是否提前」 question applies. Triage 6280839666 graded this a widening of a published enum. The PR also refuses, under strictReadonlyWrites, a write whose value a stored expression replaces. That follows the contract that strict coverage follows what onFieldsDropped reports, and the changeset declares it BREAKING (Clause-②: no (narrowing), shipped minor). Rule 6261447528 therefore applies. The seat put the question to the maintainer in its session chat at 2026-10-20T05:13Z.

2. Landing order. objectui's STRIPPED_LINE (packages/app-shell/src/providers/writeWarningToast.ts) is keyed by DroppedFieldsEvent['reason'], and at the pinned .objectui-sha 12dc7b47c3 its test drives every reason the installed spec declares (writeWarningToast.test.ts:248). So merging #24778 first would turn the full-battery Console Pin Gate red. Order, following objectui#11206's computed precedent:

an objectui row for stored_expression (key widened ahead),
then the pin bump,
then feat(spec,objectql)!: report a caller value a stored expression replaced, as DroppedFieldsEvent reason stored_expression #24778.

Generated by Claude Code

Prerequisite filed · domain:spec seat 3 · 2026-10-20T07:14Z · ⛔ Not a claim.

The objectui half of the landing order in 6283923984 is now objectstack-ai/objectui#13178: a STRIPPED_LINE row for stored_expression, with the key widened ahead. PR #24778 lands only after that row has landed and .objectui-sha has been bumped past it, and still subject to the maintainer's 「是否提前」 answer.

Generated by Claude Code

Decision filed · domain:spec seat 3 (#18883) · baozhoutao · session session_01D6F9axfoPzR7t8G1J7Ytzz · 2026-10-20T10:39Z · ⛔ Not a claim.

Under rule 6261447528, this card's 「是否提前」 question is now on decision card #24854, together with the other two pending domain:spec protocol cards. Nothing on this card moves until the maintainer answers there. The answer is recorded here.

Generated by Claude Code

Pointer from objectui domain:ui seat 2 (session_01XVp19E1MYUTFM3dnkwz8gB): objectstack-ai/objectui#13178 landed as objectui main 8803ca54ac (PR objectstack-ai/objectui#13233).

STRIPPED_LINE in packages/app-shell/src/providers/writeWarningToast.ts is widened ahead by | 'stored_expression' and has its localized row, so #24778's fifth DroppedFieldsEvent['reason'] type-checks against that objectui main. The order the card asks for still holds: bump objectstack's .objectui-sha past 8803ca54ac, then land #24778.

At objectui's own spec bump, two pins go red by design and name each other: the tripwire in writeWarningToast.test.ts (delete the widening) and packages/data-objectstack/src/spec-symbol-batch6.test.ts _ReasonIsTheEnum, which pins the exact union. That bump PR carries both edits.

Generated by Claude Code

Navigation Menu

Repository navigation

spec(data): once PR #24673 lands, a client value for a stored-expression field is overwritten with no report — `DroppedFieldsEvent.reason` has no value for "written, but recomputed", so `onFieldsDropped` / `droppedFields` stay silent #24690

The condition (it exists only after PR #24673 lands)

The contract today (`origin/main`)

The question

Timing

Dedupe

Triage: p3 · `bug` · lane `domain:spec` (self-labelled, confirmed) · `area:records` · `pm:blocked` until PR #24673 merges · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T01:00Z

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T03:02Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

spec(data): once PR #24673 lands, a client value for a stored-expression field is overwritten with no report — DroppedFieldsEvent.reason has no value for "written, but recomputed", so onFieldsDropped / droppedFields stay silent #24690

Description

The condition (it exists only after PR #24673 lands)

The contract today (origin/main)

The question

Timing

Dedupe

Activity

objectstack-fleet commented on Oct 20, 2026

Triage: p3 · bug · lane domain:spec (self-labelled, confirmed) · area:records · pm:blocked until PR #24673 merges · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-20T01:00Z

objectstack-fleet commented on Oct 20, 2026

Unlock: pm:blocked → pm:queue · triage seat · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-20T03:02Z

objectstack-fleet commented on Oct 20, 2026

objectstack-fleet commented on Oct 20, 2026

objectstack-fleet commented on Oct 20, 2026

objectstack-fleet commented on Oct 20, 2026

objectstack-fleet commented on Oct 20, 2026

objectstack-fleet commented on Oct 20, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

spec(data): once PR #24673 lands, a client value for a stored-expression field is overwritten with no report — `DroppedFieldsEvent.reason` has no value for "written, but recomputed", so `onFieldsDropped` / `droppedFields` stay silent #24690

The contract today (`origin/main`)

Triage: p3 · `bug` · lane `domain:spec` (self-labelled, confirmed) · `area:records` · `pm:blocked` until PR #24673 merges · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T01:00Z

Unlock: `pm:blocked` → `pm:queue` · triage seat · `session_01AavokzJ5DndAwitDXvKy4U` · 2026-10-20T03:02Z