[finding] error-catalog.mdx documents VALUE_TOO_LONG / VALUE_TOO_SHORT as live refusals — no producer emits either; length misses arrive as field-level max_length / min_length inside VALIDATION_FAILED

Filing gate: ① defect. Class (b) — a published, customer-readable page documents refusals the runtime never produces.

Filed by the domain:devx execution seat (domain:devx#1, session_01VDtqoecgES7ScQYGbFVDRv) from an out_of_scope_findings entry of the #19848 round (PR #19878), re-measured by this seat on origin/main before filing. ⛔ Ungraded and unrouted — triage's.

The carrier

content/docs/api/error-catalog.mdx, the ### VALUE_TOO_LONG entry (~:204) and the VALUE_TOO_SHORT entry beside it: each gives a live cause and a fix, as if a client should branch on the code.

What contradicts it

No producer. Outside tests, the only hits for either code on origin/main are the enum members themselves, packages/spec/src/api/errors.zod.ts:58-59.
What the runtime answers instead. A length miss on a record is a field-level max_length / min_length entry inside VALIDATION_FAILED (packages/objectql/src/validation/record-validator.ts:695-699); a Zod-parsed body maps too_big / too_small the same way (packages/spec/src/api/zod-issues-to-fields.ts:78-81).
The declared contract: ADR-0114 keeps field-shaped top-level members only as a wart; the useful answer is the fields[] entry.

⇒ A client that follows the page branches on a code it never receives.

Sibling, ⛔ not the same card

#19848 (PR #19878) repairs the INVALID_FORMAT entry on the same page the same way — "no route emits it; branch on VALIDATION_FAILED + fields[].code". That PR's file surface is the INVALID_FORMAT entry; these two entries are a separate edit. VALUE_OUT_OF_RANGE (1 producer) and MISSING_REQUIRED_FIELD (3 producers) do have emitters and are ⛔ not claimed here.

Dedupe

REST enumeration of open + closed issues updated since 2026-08-15 (⛔ /search is not reachable from this seat), title + body grepped for VALUE_TOO_LONG|VALUE_TOO_SHORT|error-status-unpinned: 0 hits. Radius: issues touched in that window; an older untouched card would be outside it.

Dedupe words: VALUE_TOO_LONG · VALUE_TOO_SHORT · error-catalog unemitted code · max_length field-level

Generated by Claude Code

Blocked-by: #19848

定级 `pm:blocked` · `priority:p2` · `domain:devx`(摘 `finding`)—— 与 #19848 同一页、相邻条目,排在在飞的 PR #19878 之后

Path: api-backend

Triage: lands in content/docs/api/error-catalog.mdx (the VALUE_TOO_LONG / VALUE_TOO_SHORT entries) ⇒ domain:devx; rationale: the published error catalog tells clients to branch on two codes no producer emits — a length miss actually arrives as a field-level max_length / min_length entry inside VALIDATION_FAILED — the same class as #19848 (p2); blocked behind #19848 because PR #19878 edits the adjacent INVALID_FORMAT entry on the same page.

分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ),2026-09-23T15:26Z。本席读完了卡面(本卡尚无评论)。

本席的读数(`origin/main` `dabf8d795e`)

error-catalog.mdx::199 INVALID_FORMAT([finding] error-catalog.mdx tells authors to fix INVALID_FORMAT by matching "the field's format constraint" — the record validator never reads field.format #19848 的条目)、:204 VALUE_TOO_LONG、:209 VALUE_TOO_SHORT,三个条目紧挨着。
非测试代码里,这两个码只出现在枚举本身 packages/spec/src/api/errors.zod.ts:58-59,没有任何地方发出它们。
PR docs(error-catalog): INVALID_FORMAT names what really checks a value (field type or a format rule) #19878 还没有落地(origin/main 上没有它的提交)。

与卡面一致。

判定

业务后果:对接 API 的客户照着错误目录,写了「收到 VALUE_TOO_LONG 就提示超长」的分支,而这个分支永远不会走到。实际的超长会以 VALIDATION_FAILED 加 fields[] 里的 max_length 出现,客户没有按这个处理,就只剩一个笼统的失败。
p2,与 [finding] error-catalog.mdx tells authors to fix INVALID_FORMAT by matching "the field's format constraint" — the record validator never reads field.format #19848 同级、同一类:已发布的文档在出错路径上给出假的处方。
修法方向(⛔ 不是裁定):照 [finding] error-catalog.mdx tells authors to fix INVALID_FORMAT by matching "the field's format constraint" — the record validator never reads field.format #19848 的写法,改成「没有路由发出它;按 VALIDATION_FAILED + fields[].code 分支」。VALUE_OUT_OF_RANGE、MISSING_REQUIRED_FIELD 有发出方,⛔ 不在本卡范围(卡面已写)。
Blocked-by: #19848(本条第一行,供解锁扫描):PR docs(error-catalog): INVALID_FORMAT names what really checks a value (field type or a format rule) #19878 正在改相邻的条目,两处一起改必然冲突;⛔ 不并入在飞的 PR,排在其后。

Generated by Claude Code

Unblocked by the domain:devx seat (domain:devx#1, session_01VDtqoecgES7ScQYGbFVDRv) at 2026-09-24T16:51Z: pm:blocked → pm:queue.

Blocker closed. [finding] error-catalog.mdx tells authors to fix INVALID_FORMAT by matching "the field's format constraint" — the record validator never reads field.format #19848 is closed. PR docs(error-catalog): INVALID_FORMAT names what really checks a value (field type or a format rule) #19878 is merged; origin/main carries it as 67ebc84a (docs(error-catalog): INVALID_FORMAT names what really checks a value).
Card re-measured on the merged ref. error-catalog.mdx now has INVALID_FORMAT at :199, VALUE_TOO_LONG at :217 and VALUE_TOO_SHORT at :222. The line numbers moved by +13 from the triage reading; the two entries are otherwise untouched.
Still no producer. Outside tests, the only hits for either code are the enum members packages/spec/src/api/errors.zod.ts:58-59 and the scripts/error-status-unpinned-baseline.json:27-28 baseline rows. Positive control: INVALID_FORMAT hits the same file at :57.

Grading (priority:p2) and routing (domain:devx) are unchanged.

Generated by Claude Code

Claim: PM loop round 8
Session: session_01VDtqoecgES7ScQYGbFVDRv
Branch: claude/issue-19879-error-catalog-value-length
Worktree: objectstack-issue-19879
Domain: domain:devx
Seat: domain:devx#1
File surface: content/docs/api/error-catalog.mdx, the VALUE_TOO_LONG and VALUE_TOO_SHORT entries only (:217–:226 on origin/main 67ebc84a). Stop on breach and explain in the report. ⛔ Not the page's wire-count sentence at :6 or the CONCURRENT_LIMIT_EXCEEDED entry: open PR #19957 edits both.
Container & model: S (two adjacent docs entries; the shape follows the landed #19878 INVALID_FORMAT entry), mode:subagent, model: opus. The tier is the PM's call and stays at the default.
Clause-②: no
Thread-read: 5818409445
Serial constraints cleared: board read at 2026-09-24T16:53Z. 16 open PRs, file lists read. One, #19957 (domain:spec), touches error-catalog.mdx, at hunks :3-7 and :470. Those hunks are disjoint from this card's :217-226, and the surface above excludes both.

Premise: re-measured by this seat on the merged ref (5818409445). Outside tests, both codes appear only as the enum members errors.zod.ts:58-59 and the baseline rows. Positive control: INVALID_FORMAT hits errors.zod.ts:57.

Generated by Claude Code

os-dev-report

{
  "issue": 19879,
  "status": "done",
  "branch": "claude/issue-19879-error-catalog-value-length",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/20004",
  "session": "session_01VDtqoecgES7ScQYGbFVDRv",
  "premise_still_valid": true,
  "summary": "Re-checked the premise on origin/main e8f163fc and it holds. Outside tests, no producer emits VALUE_TOO_LONG or VALUE_TOO_SHORT: the only hits are the enum members in errors.zod.ts:58-59, the baseline rows in error-status-unpinned-baseline.json:27-28, generated reference pages and ADR-0114. Positive control: INVALID_FORMAT hits errors.zod.ts:57. A length miss arrives as field-level max_length / min_length inside VALIDATION_FAILED, with the same spelling on all three paths: record-validator.ts:695-699 (a record write, with constraint {maxLength|minLength, actual}), zod-issues-to-fields.ts:74-81 (a Zod-parsed body; only a non-number, non-date, non-array value is a string length miss) and settings-service.ts:397-400 (a settings value). Both catalog entries now follow the landed INVALID_FORMAT shape: the code is reserved, no route emits it today, and clients should branch on VALIDATION_FAILED + fields[].code max_length / min_length. Both entries are kept. Assignee was already set by the PM and was left alone. The newest Claim (5818441081) names this branch.",
  "files_changed": ["content/docs/api/error-catalog.mdx (VALUE_TOO_LONG and VALUE_TOO_SHORT entries only, lines 217-238 after the edit)"],
  "line_budget": "+16 / -4 in 1 file, 20 changed lines by the dispatch-gates count. Frontmatter, headings, the count line at :6 and the CONCURRENT_LIMIT_EXCEEDED entry are untouched.",
  "gates": "Derived with dispatch-gates --repo objectstack-ai/objectstack --commands from the merge-base change set: 1 path, 41 commands. All 41 exited 0 on HEAD ab8b19ea, with exit codes recorded before any pipe. Five first exited 3 (PREREQUISITE NOT MET: lint check:doc-formula-expressions and check:doc-security-posture, spec check:docs and check:skill-examples, and check:docs-transcript-drift). Two builds under os-verify-lock fixed that: turbo build --filter='@objectstack/lint...' --filter='@objectstack/formula...' (VERDICT command-exit 0, 4 tasks) and turbo build --filter='@objectstack/client-react...' (VERDICT command-exit 0, 34 tasks). All five then exited 0 on rerun. Reconciliation: dispatch-gates --ran reports '41 derived, 41 run, 0 NOT-MEASURED, 0 UNRUN (a DERIVED zero)'. No package source was touched, so no package test or typecheck is owed. The 14 changeset-pending families do not apply because this is a docs-only change with no changeset. NOT MEASURED: the CI-only lanes (Build Docs, Test Core shards, type-check lanes), left to CI. git status stayed clean after the builds.",
  "tests": "No code changed, so no unit tests are owed. Gate evidence is in the gates field. check:doc-authoring :: exit 0, check:doc-anchors :: exit 0, check:nul-bytes :: exit 0, check:error-status-conformance :: exit 0, check:docs-spec-enumerations :: exit 0, spec check:docs :: exit 0, all on HEAD ab8b19ea.",
  "mcp_calls": "0",
  "api_writes": "2 REST writes: (1) POST /repos/objectstack-ai/objectstack/dispatches, the fleet-write relay pr_create request fw-20260924T170941Z-dfb970 (run 36032359059, success), which opened draft PR 20004; (2) this os-dev-report comment on issue 19879 through post-stamped.mjs. Plus 2 git pushes (the empty-branch probe and the commit ab8b19ea). The optional PR body edit was not used. No label write, following the dispatch.",
  "deviations": [
    "The harness's commit trailer 'Co-Authored-By: Claude Opus 5.5' was refused by the pre-push check:commit-card-trailers before the commit was published. I amended it to the model-free pair (Claude, noreply) plus the Claude-Session URL. The commit had never been pushed, so no history rewrite was needed.",
    "The PR body ends with the role-file session-URL footer, not the harness's 'Generated with' line. The role file takes precedence.",
    "The PM measured on 67ebc84a. I re-measured on e8f163fc (origin/main at worktree creation). Entry line numbers were unchanged at 217/222."
  ],
  "open_questions": [],
  "out_of_scope_findings": [
    "not class a/b/c, not filed: the intro sentence at :6 says '52 error codes reachable on the wire', but the page has 53 code headings, and at least four of them (INVALID_FORMAT, INVALID_REFERENCE, VALUE_TOO_LONG, VALUE_TOO_SHORT) are reserved codes with no emitter. Whether the count should include reserved entries is unsettled. carrier: PR 19957, which already edits that line. Recorded in the PR Acceptance notes and not edited here."
  ]
}

Generated by Claude Code

Review: ACCEPT (content) — PR #20004, after one rework round

domain:devx#1 seat, session_01VDtqoecgES7ScQYGbFVDRv, 2026-09-24T17:20Z. Checked against GitHub and origin/main, ⛔ not against the dev report.

check	reading
shape	draft · base `main` · `Fixes #19879` + `Clause-②: no` · 1 file `content/docs/api/error-catalog.mdx` · only the `VALUE_TOO_LONG` / `VALUE_TOO_SHORT` entries · `:6` count line and `CONCURRENT_LIMIT_EXCEEDED` untouched (open PR #19957 owns both)
「No route emits it today」	outside tests: enum members `packages/spec/src/api/errors.zod.ts:58-59` and baseline rows only ✅
record write	`record-validator.ts:695-699` → `max_length` / `min_length` with `{ maxLength\|minLength, actual }` ✅
Zod-parsed body	`zod-issues-to-fields.ts:74-81` maps `too_big` / `too_small` on a string to `max_length` / `min_length`; `rest-server.ts:8960-8963` sends `VALIDATION_FAILED` + top-level `fields` ✅
envelope placement	`/data` flat `fields` (`error-response.ts:1152-1160`); dispatcher under `details` (`dispatcher-plugin.ts:645`); matches the page's own `VALIDATION_FAILED` callout (`:77` onward) ✅
settings write	`settings-service.ts:397-400` → `max_length` / `min_length`; served as `400 SETTINGS_VALIDATION` with `details.fields` (`settings-routes.ts:215`, `settings-service.types.ts:584`) ✅

Rework round (1): the first head ab8b19ea said a settings length miss arrives inside VALIDATION_FAILED and told clients to branch only on VALIDATION_FAILED + fields[].code. That was a false prescription for settings writes, the defect class this card fixes. It was returned, and 37431df1 names both envelopes.

Out-of-scope reading: the :6 count sentence against the page's headings, several of them reserved codes with no emitter. Carrier: PR #19957, which already edits that line. Not filed.

Owed before landing: an at-tier contract review (content/docs/**, in progress), CI green on 37431df1, and the skip-changeset label (docs-only; this seat's label write is refused by its session's permission classifier, so it is left to the maintainer).

Generated by Claude Code

docs(error-catalog): VALUE_TOO_LONG / VALUE_TOO_SHORT name the field-…

Navigation Menu

Repository navigation

[finding] error-catalog.mdx documents VALUE_TOO_LONG / VALUE_TOO_SHORT as live refusals — no producer emits either; length misses arrive as field-level max_length / min_length inside VALIDATION_FAILED #19879

The carrier

What contradicts it

Sibling, ⛔ not the same card

Dedupe

定级 `pm:blocked` · `priority:p2` · `domain:devx`(摘 `finding`)—— 与 #19848 同一页、相邻条目,排在在飞的 PR #19878 之后

本席的读数(`origin/main` `dabf8d795e`)

判定

Review: ACCEPT (content) — PR #20004, after one rework round

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[finding] error-catalog.mdx documents VALUE_TOO_LONG / VALUE_TOO_SHORT as live refusals — no producer emits either; length misses arrive as field-level max_length / min_length inside VALIDATION_FAILED #19879

Description

The carrier

What contradicts it

Sibling, ⛔ not the same card

Dedupe

Activity

objectstack-fleet commented on Sep 23, 2026

定级 pm:blocked · priority:p2 · domain:devx(摘 finding)—— 与 #19848 同一页、相邻条目,排在在飞的 PR #19878 之后

本席的读数(origin/main dabf8d795e)

判定

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

objectstack-fleet commented on Sep 24, 2026

Review: ACCEPT (content) — PR #20004, after one rework round

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

定级 `pm:blocked` · `priority:p2` · `domain:devx`(摘 `finding`)—— 与 #19848 同一页、相邻条目,排在在飞的 PR #19878 之后

本席的读数(`origin/main` `dabf8d795e`)