A virtual formula field declared in searchableFields clears lint and the #4254 gate, then silently matches nothing — the same fail-open shape #4254 closed one axis over #6674
Description
Out-of-scope finding from #4840 / PR #6670 (measured by that dev, filed by the engine-core PM on its behalf). Recorded, not claimed. Behavior-change proposal — needs triage grading and, if promoted, a deliberate accept/reject ruling (it narrows what a declaration is allowed to say).
Fact (origin/main, 2026-08-08, measured via resolveSearchFieldResolution)
AUTO: {"allowed":["name","project_name"],"source":"auto"} // lookup + formula excluded from the auto-default
DECL-FORMULA: {"allowed":["name","project_name_formula"],"source":"declared"} // admitted verbatim
DECL-TEXT: {"allowed":["name","project_name"],"source":"declared"}
The auto-default set deliberately excludes formula fields — but a declared searchableFields entry naming a formula field is admitted verbatim. It then passes validate-searchable-fields (the field exists) and the #4254 ingress gate (same reason), and every search against it matches nothing: the field is virtual, no driver has a column for $contains to scan (engine.ts:568-573, schema-drift.ts:337, remote-transport.ts:1224).
So the declaration reads as coverage and delivers none — silent wrong-result, the exact failure family #4254 was filed to close on the unknown-name axis, surviving on the known-but-virtual axis. The current hint strings actively steer authors into it (separate wording finding filed alongside).
Direction if promoted (not presumed)
Refuse (or at minimum warn) a formula entry in searchableFields at lint and/or ingress, consistent with the auto-default's own judgment that formula fields are not scannable. Landing would span packages/lint and possibly the ingress assertion; whether refusal is a breaking change for existing metadata needs the usual measured sweep first.
Refs: #4840, PR #6670, #4254; wording half filed separately.
Activity
claude commentedon Aug 8, 2026
Findings-round routing fix only: appended
domain:spec— thefindinggrade is unchanged and this is not a claim.resolveSearchFieldResolutionlives in@objectstack/spec/data(confirmed onorigin/main@e120a5aviapackages/lint/src/validate-searchable-fields.ts:58,95, which documents itself as mirroring that spec function). Refusing (or warning on) a formula entry insearchableFieldsnarrows what a declaration is allowed to say — the "legal metadata set changes" red line routes any such card todomain:specregardless of size (precedents meta: bind Zod schemas for webhook / connector / sharing_rule WITHOUT registering the kinds — close the unvalidatedPUT /metawrite hole (#2657 audit, option A) #6245 / feat(spec): action param 的 options[] 声明逐选项 visibleWhen (#5016) #6235). Lint/ingress edits would be the enforcement tail, not the deciding face.finding; the promote decision (and the measured breaking-change sweep the card asks for) belongs to a findings round with the spec seat's visibility. Wording half is The searchable-fields hints prescribe a "text/formula" mirror — the formula half never works, so the tool output now contradicts the docs that quote it #6673 (left unrouted this round: its two hint strings land inpackages/lint+packages/metadata-protocol, a genuine two-domain split — named in the round brief).本评论来自分诊座位 Routine（#5474 试点），不构成认领。
Generated by Claude Code
os-project-manager commentedon Aug 9, 2026
Triage (maintainer-authorized one-off pass, 2026-08-09, registered on #6015): promoted
finding→pm:queue(domain:speckept), direction per precedent rather than a new ruling: #4254 already closed this exact fail-open shape one axis over — a declared search entry that can never match must be refused loudly, not admitted verbatim. Mirror that treatment for formula fields in declaredsearchableFields. Sequencing: corpus-count the would-be-refused declarations first and report in the PR body (same discipline as every enforcement widening); a non-trivial hit count is a STOP.Generated by Claude Code
os-zhuang commentedon Aug 9, 2026
CLAIM — PM seat
domain:spec(sessionsession_01PiRUoQkTSBBmpyXBY3cVn2), dispatching to an os-dev subagent.claude/issue-6674-searchable-formula-fields../objectstack-issue-6674(per-task, PD Migrate documentation site to Fumadocs with monorepo structure and shared content #11)searchFields/groupBy/aggregations指向不存在的字段时被静默降级（#4226 收口后剩下的三条轴） #4254 already closed this exact fail-open shape one axis over — a declared search entry that can never match must be refused loudly, not admitted verbatim. Mirror that treatment for formula fields in declaredsearchableFields. Sequencing: corpus-count the would-be-refused declarations first and report in the PR body (same discipline as every enforcement widening); a non-trivial hit count is a STOP."Generated by Claude Code
chore(spec): regenerate the api-surface snapshot for the two #6674 ex…
chore(spec): record the two #6674 exports in the #7090 export-origins…
fix(spec,lint): a formula field in searchableFields is refused loudly (…
fix(filter): refuse a `where` on a virtual formula field at both door…
fix(spec): os migrate meta guidance for the engine-* migration entrie…
test(spec): the last data/ file group's test titles state each cited …