plugin-kanban: cardTitle falls back to titleField with || in the card list and ?? in the detail drawer — an authored empty string resolves two different ways on one board

Found while implementing objectui#8201 (declaring the scalar object-kanban keys), measured on origin/main 998d846aa.

The two sites

ObjectKanban.tsx resolves the same one authoring choice — "which record field is the card title", spelled cardTitle with titleField as its legacy alias — in two places, with two different operators:

ObjectKanban.tsx:403 (the card list, inside the effectiveData memo):

const explicitTitleField: string | undefined =
  schema.cardTitle || schema.titleField;

ObjectKanban.tsx:1077 (the record-detail drawer heading, inside the navigation overlay branch):
```
const titleField = schema.cardTitle ?? schema.titleField;
```

Why the difference is observable

|| and ?? differ on exactly the falsy-but-present values, and cardTitle is a string key, so the reachable divergence is the empty string:

authored	`:403` (`\|\|`) resolves to	`:1077` (`??`) resolves to
`{ cardTitle: '', titleField: 'name' }`	`'name'`	`''`

So a board authored with cardTitle: '' and a titleField fallback renders its cards titled from name, and opens a detail drawer whose heading falls through to the Record #<id> / object-label floor instead — one board, one authored document, two answers to one question. '' is not an exotic value on this path: an authoring UI that clears a text input, or a saved view that round-trips an unset field as '' rather than dropping the key, produces it.

Neither operator is obviously the intended one; what is defective is that the same choice is read twice with two precedence rules. That is the "one authored key read twice with two incompatible meanings" class objectui#7711 closed on object-calendar's filter and objectui#4034 closed on the map.

Scope note

Not fixed in objectui#8201's PR: that card declares the two spellings as authoring surface and does not change how either is resolved. The declaration makes the divergence more reachable, though, because the html tier now tells authors both keys exist — which is why it is filed rather than left as a comment.

Whichever operator wins, the fix should read the pair once (an exported resolver, the way resolveKanbanCardFields already handles cardFields) so a third read site cannot invent a third precedence.

Generated by a Claude Code os-dev seat, session session_01YBWFb5YgMU5dw8p2VKj16S.

Triage routing

Labels bug plugin plugin: kanban finding domain:ui pm:queue priority:p3 · Type Bug

Lane — domain:ui. Fix lands in packages/plugin-kanban, a published library. Symptom and fix are the same package.

⛔ Anchors have drifted since filing. The card cites origin/main 998d846aa; on 83d4fda16 the two sites are:

card says	actual today
`ObjectKanban.tsx:403`	`:404` — `schema.cardTitle \|\| schema.titleField;`
`ObjectKanban.tsx:1077`	`:1110` — `const titleField = schema.cardTitle ?? schema.titleField;`

Read the lines, not the numbers. The divergence itself is confirmed live: both operators are still there, still on the same pair.

Admission — class (a), reproducible defect. One authored key, two precedence rules, one board. The card's table is the reproduction and it is correct: { cardTitle: '', titleField: 'name' } titles the cards from name and drops the drawer heading to the Record #<id> floor.

Grade p3. The divergence is real and the class is right, but the reachable input is a single value — cardTitle: '' — and no producer in the repo writes it. Re-grade to p2 on either of these, and both are cheap to check before starting: an authoring surface that round-trips a cleared text input as '' rather than dropping the key, or any saved-view/metadata read that can emit it. If one exists, this stops being a latent divergence and becomes an observed one.

On the remedy. The card's own closing instruction is the load-bearing part and I am keeping it as a fence: read the pair once, through an exported resolver, the way resolveKanbanCardFields already handles cardFields. ⛔ Do not fix this by making :1110 say ||. That repairs today's two sites and leaves the property that produced them — the pair is readable ad hoc — fully intact, so a third read site invents a third precedence.

Which operator wins is genuinely open and is not a decision-box question: both branches only change how an already-broken input resolves, so neither widens the accepted set. Pick one in the claim comment, state it, and pin it. Note index.tsx:525-526 already publishes the intended precedence in prose — "cardTitle wins when both are authored" — which argues for ??'s intent but does not settle the empty-string case; the resolver's docblock should say what '' means, because today nothing does.

Scope fence. ⛔ Not a defect in objectui#8201 and not a reason to hold it — #8201 declares the two spellings and does not touch their resolution. It does make this more reachable, which is why the card exists.

Triage seat only — no claim, no dispatch, no code.

Generated by Claude Code

Claim: session session_01611D6ZaRaMmwTNQmSbk8MH · branch claude/issue-8308-kanban-card-title-precedence

PM dispatch（domain:ui 席）。assignee 与本评论由 PM 代设，dev 继承二者、⛔ 不写 assignee 字段、⛔ 不再发第二条 claim。

零、⭐ 运算符：裁 `||` 语义（分诊席把这一步交给认领席，此处照办并说明理由）

分诊评论 5583020403 写着："Which operator wins is genuinely open and is not a decision-box question … Pick one in the claim comment, state it, and pin it." ⇒ 本席裁：

cardTitle 只在「非空」时胜出；'' 视作未设置，落到 titleField。 也就是 || 的语义，但 ⛔ 不是通过写一个 || 来实现（见第一节）。

理由，可证伪：

⭐ cardTitle 是一个字段名，而 '' 不可能命名任何记录字段。 ⇒ 空串在这个键上不存在有意义的解读，它只可能是空输入的产物。⛔ 若你能在树里找到一个反例（某处把 '' 当作一个有意义的字段名解析），停下来报告，本裁决作废。
卡片自己给的两个来源都印证「未设置」这一解读：清空了文本框的授权界面、把未设字段回写成 '' 而不是丢键的 saved view。两者都不是「作者刻意选择了空字段名」。
已发布散文不与之冲突。 index.tsx 的 "cardTitle wins when both are authored" 在 || 下依然成立（非空的 cardTitle 仍然胜出）；那句话本来就没有覆盖空串这一格 —— 分诊席也是这么说的。

⚠️ ?? 的 intent 论据（分诊席点到的）本席已读并驳回：?? 表达的是「只要 present 就胜出」，而本键的 present 与 meaningful 不是一回事 —— 理由 1 就是它们分开的地方。

一、⛔⛔ 围栏：结构修法不可协商

分诊席把卡片自己的收尾指令立为围栏，本席原样保留：

读这一对键一次，通过一个导出的解析器，就像 resolveKanbanCardFields 已经处理 cardFields 那样。
⛔ 不要靠把 :1110 改成 || 来修。 那只修好今天这两处，而产生它们的那个性质 —— 这对键可以被随手就地读 —— 原封不动，于是第三个读点会发明第三种优先级。

⇒ 交付物是那个解析器，两处读点都改为调用它。⭐ 它的 docblock 必须写明 '' 的含义 —— 分诊席指出今天没有任何东西说这件事，这正是缺陷得以存在的原因。

二、⛔ 锚点已漂移，⛔ 不要照抄任何行号

分诊席实测（在 83d4fda16 上）：卡片说的 :403 实际是 :404，:1077 实际是 :1110。⚠️ 而那也是它那时的读数 —— 对你的树同样可能已经错。⇒ 读行，不要读数字；按锚点（schema.cardTitle || schema.titleField / schema.cardTitle ?? schema.titleField）定位。

三、⚠️ 开工前先做的一次便宜检查（可能改定级）

分诊席给了 p3，判据是可达输入只有 cardTitle: '' 这一个值，而仓库里没有任何生产者写它。⇒ 动手前先查两件事（都很便宜）：

有没有授权界面把清空的文本框回写成 '' 而不是丢键？
有没有任何 saved-view / 元数据读路径能吐出 ''？

⭐ 只要有一个存在，本卡就从「潜在分歧」变成「已观察到的分歧」⇒ 报告并建议升 p2（⛔ 你不自行改标，报告即可）。若都不存在，也在报告里明说 —— 这决定 PR 正文里这条缺陷该被描述得多重。

四、判别力

⭐ 主体：{ cardTitle: '', titleField: 'name' } 在卡片列表与详情抽屉标题上必须解析出同一个答案。修复前两处不同（'name' vs 落到 Record #<id> 地板），修复后一致。修复前这条断言必须红。
⭐ 控制（必须两个世界都绿）：{ cardTitle: 'subject', titleField: 'name' } 两处都解析为 'subject'；{ titleField: 'name' }（无 cardTitle）两处都解析为 'name'。⚠️ 若控制跟着主体一起动，它就不是控制。
⚠️ 另加一条结构性的 pin：两处读点都走那个解析器 —— 这样第三个读点若绕过它，将来有东西会红。

五、⛔ 边界

⛔ 不是 objectui#8201 的缺陷，也不构成拖住它的理由（#8201 只声明这两种拼写，不触碰它们的解析）。⛔ 不改 resolveKanbanCardFields 已有的行为 · ⛔ 不顺手统一 kanban 里其它的 ||/?? · ⛔ 不扩范围 · ⛔ 不编辑共享主检出（用专属 worktree）· ⛔ 绝不 git stash · ⛔ 绝不按名杀进程 · ⛔ 不碰 content/docs/releases/ · ⛔ 不碰 governed surface。记得 changeset。

六、⚠️ Clause-② —— 由你判 + 载体预授权

机械下限：新导出符号、或已发布载荷上的新键，恒 yes；拿不准 ⇒ yes。⚠️ 本卡大概率触发下限 —— 交付物就是一个新的导出解析器。⛔ 但仍由你按实际 diff 判并给依据（它是否真的经包 barrel 导出？以 dist/*.d.ts 为准）。
⭐ 载体预授权：固定拼写行必须落在这条 claim 评论里；本席 MCP 面上无编辑评论的工具，dev 席的 REST 有 PATCH /issues/comments/{id} ⇒ 预先授权你把测得的 Clause-②: 行追加到本评论末尾（保留上方原文，注明 dev 测得、PM 授权）。⚠️ 送零页脚正文并回读；若补出两个，至多修剪一次，回不去就停手、⛔ 不进循环；冒号必须是 ASCII 0x3a。判 yes 则 needs:contract-review 双载体同笔挂。

七、⭐ 位置轴与工具链

位置轴判据是 stored vs computed：只有被存下来、且被当作键或做相等比较的字面量才会过期。已确证载体 scripts/check-doc-example-types.mjs 的 UNGATED_EXAMPLES（键含行号），⚠️ cross-file-line-citation-census.mjs 看不见它 —— 两类都查，判 (A)/(B) 靠找锚点⛔ 不是比行号；强形式是在你的分支对 main 的一次真实 git merge --no-ff 结果树上再跑一次。
⚠️ 内容轴往两个方向枚举：不只「谁调用」，还要「谁监视」—— 读你改动文件的源文本 / 路径 / 形状的量具可以住在任何一个包里（本班 PR #8907 就是被 packages/sdui-parser/ 里的一条普查打红的）。

⚠️ 工具链：pnpm --filter PKG exec vitest run FILE 被 objectui#3378 守卫拒绝会静默跑别的文件报绿；pnpm --filter PKG test 在无 test 脚本的包上匹配零个脚本 exit 0。⛔ 都不是读数。从仓根：pnpm exec vitest run packages/PKG/src/FILE。
⚠️ 门 exit 2 = PREREQUISITE NOT MET ⇒ NOT MEASURED，⛔ 不是红也不是绿。验收引运行器自己的 VERDICT command-exit 行。
⚠️ REST /search/* 被出口代理按设计拒绝（403）—— 用 search_issues 去重前先用已知存在的正样本验证量具。

八、Clause-② —— dev 席测得，PM 预授权追写（第六节）

Clause-②: yes

依据（按实际 diff 实测，以 dist/*.d.ts 为准）：本卡交付物是一个新的导出解析器 resolveKanbanTitleField，建后它确实落在 packages/plugin-kanban/dist/ObjectKanban.d.ts （export declare function resolveKanbanTitleField(schema: KanbanTitleFieldSource | null | undefined): string | undefined;）—— 机械下限命中。

⚙ 一个限定，一并报上而不拿它降级判定：它并未经包 barrel 导出（dist/index.d.ts 零命中），且包的 exports 只路由 . 与 ./style.css，所以消费者今天触不到它 —— 与 resolveKanbanCardFields 同形。但本卡另有一项真正的契约动作：'' 在 cardTitle / titleField 上被裁为未设置，这是已发布渲染器对一个已声明授权键的解析语义变更 —— 单凭这一点也应走契约复审。拿不准 ⇒ yes。

PR：#8917 （draft）。needs:contract-review 已双载体挂上（本卡 + PR）。

上方原文为 PM 派发单，未动；本节由 dev 席按第六节预授权追写。

Generated by Claude Code

os-dev-report

{
  "issue": 8308,
  "status": "done",
  "branch": "claude/issue-8308-kanban-card-title-precedence",
  "pr": "https://github.com/objectstack-ai/objectui/pull/8917",
  "premise_still_valid": true,
  "summary": "Both read sites of the cardTitle/titleField pair now go through one new exported resolver, resolveKanbanTitleField in packages/plugin-kanban/src/ObjectKanban.tsx — the card list's effectiveData memo (was `schema.cardTitle || schema.titleField`) and the record-detail drawer heading (was `schema.cardTitle ?? schema.titleField`). Anchors had drifted AGAIN since triage: the sites were at :552 and :1303 on 348725a7c, not the card's :403/:1077 nor triage's :404/:1110 — located by anchor text, never by number. The PM's ruling is carried and, per the fence, written down for the first time: the resolver's docblock states that '' means UNSET on either key because cardTitle names a record field and '' can name none, so cardTitle wins only when NON-EMPTY. No counterexample exists in the tree — swept every cardTitle/titleField read in packages/ and apps/; collectors filter `typeof f === 'string' && f`, both view adapters floor an empty titleField with `|| 'name'`, and the nearest near-miss (ObjectGallery's `titleField != null`) still cannot resolve '' to a field because item[''] is undefined. The assignee was already set by PM and the newest Claim: comment names this branch and this session; no assignee write and no second claim were made.",
  "tests": "All runs serialised on the container's shared heavy-verify lock (slot objectui-issue-8308); each verdict quoted is the runner's own VERDICT command-exit line, never a bare $?. [1] `pnpm exec turbo run build --filter=...@object-ui/plugin-kanban --concurrency=2` -> 40/40 tasks successful, BUILD_EXIT=0. [2] `pnpm --filter @object-ui/plugin-kanban run type-check` -> echoed `tsc --noEmit && tsc -p tsconfig.test.json`, TYPECHECK_EXIT=0. [3] `pnpm exec vitest run packages/plugin-kanban/` (repo-root form, never the package-filtered form the objectui#3378 guard silently redirects) -> 'Test Files 43 passed (43) / Tests 271 passed (271)', re-run at the FINAL commit 241b39f1f. [4] cross-package watchers that read ObjectKanban.tsx off disk (object-kanban-record-source-7780, object-kanban-group-by-limit-7322, kanban-calendar-filter-sort-8174, base-bind-declared, kanban-arm-batch70-7742, kanban-plugin-dialect-authoritative-7664) -> 'Test Files 6 passed (6) / Tests 95 passed (95)'. [5] gates derived from the diff, all exit 0: check:control-bytes ('scanned 7088 tracked text file(s)'), check:handler-key-reads, check:element-data-source-declaration, check:sdui-registration-pins, check:entry-guard, check:unreferenced-sources, check:vi-mock-specifiers, check:comment-mask-corpus (1 pre-existing disagreement in apps/console/src/pages/DocsIndex.tsx, within the objectui#7882 ceiling; ObjectKanban.tsx is NOT among them, which independently vouches for the masker my census uses), check:changeset-presence ('1 source file(s) of 1 released package(s) changed, and this change declares 1 changeset(s)'). `pnpm --filter @object-ui/plugin-kanban run lint` exit 0 (181 pre-existing warnings, 0 errors). NOT MEASURED, not a finding: check:readme-exports exits 1 on exactly 8 rows, all in @object-ui/cli and @object-ui/plugin-ai — two packages outside this diff and outside the closure I built — each printing 'its type entry ./dist/index.d.ts is not on disk -- run `pnpm build` first'. @object-ui/plugin-kanban IS among the 35 judged packages and is clean. Repo-wide scans (`pnpm lint`, the full suite, the whole-tree-build doc gates) are declared to CI. ABLATION — two legs, both mutating ObjectKanban.tsx on disk with a trap on EXIT INT TERM whose handler is the restore, each proving the mutation landed by grep -c on the injected and deleted text PLUS a blob-hash comparison against HEAD, and each restored with `git checkout HEAD -- PATH` and proven restored by blob hash (10cf8041...) AND an empty `git diff HEAD`. No build is involved and the reason is stated rather than assumed: the spec imports the component RELATIVELY (`../ObjectKanban`) and the census reads the file off disk, so nothing resolves through dist/. LEG 1 (both operators put back; resolver call count in the component 2 -> 0, on-disk proof '|| schema.titleField' -> 1, '?? schema.titleField' -> 1, 'resolveKanbanTitleField(schema)' -> 0): 'Tests 3 failed | 8 passed (11)'. Named rows — RED: the SUBJECT render row (drawer accessible name read 'Contacts Detail', the object-label floor, while the cards read 'Ada Lovelace'), plus BOTH structural pins (census listed 4 offenders by name and offset; the call-site count read undefined against an expected 2). GREEN in the SAME run: both controls — `{ cardTitle: 'subject', titleField: 'name' }` and `{ titleField: 'name' }` — so they moved with neither world and are real controls. LEG 2 (a third ad-hoc read added: `const ablationThirdReadSite = schema.cardTitle ?? schema.titleField;`, on-disk proof -> 1): 'Tests 1 failed | 10 passed (11)' — ONLY the census reddened, naming the new read. That is the structural pin proving it catches the third read site a two-operator fix would have left reachable. Direction predicted before running and observed as predicted: turn RED, both legs.",
  "mcp_calls": "0 — the whole run went through repo-scoped REST (probe: GET /repos/objectstack-ai/objectui/issues/8308 -> HTTP 200) and git; zero MCP GitHub calls",
  "open_questions": [
    {
      "question": "Attribution footer form for the PR body: this task file mandates the session-URL form (`_Generated by [Claude Code](https://claude.ai/code/session_ID)_` after a rule line) for a created PR body, while the session's injected attribution reminder mandates a different two-line block ('Generated with [Claude Code](https://claude.com/claude-code)' plus a bare session URL). Flagged rather than silently sided with, per the no-quiet-side-taking rule.",
      "options": [
        "A — use the task file's session-URL footer for the PR body (what I did), commit trailers from the injected reminder",
        "B — use the injected reminder's block for the PR body too",
        "C — emit both blocks"
      ],
      "recommendation": "A. The task file is the repo-specific contract and explains its own mechanism (GitHub PATCH downgrades the session-URL form and appends a second footer unconditionally), so it is the form whose byte-level behaviour is known here; the injected reminder's commit trailers are not in conflict and were used verbatim in the commit. C is ruled out — two footers is the failure mode the contract tells me to trim, not to create. No code depends on this; it is a reporting question only."
    }
  ],
  "out_of_scope_findings": [
    "REACHABILITY / possible re-grade to p2 (PM section 3, reported not relabelled): a stored-view metadata path CAN emit cardTitle: '' onto the node. BOTH view adapters strip `titleField` from their kanban passthrough but NOT `cardTitle`, so `...restKanban` forwards a stored kanban.cardTitle verbatim — packages/plugin-view/src/ObjectView.tsx destructures `{ columns, groupByField, groupField, titleField, conditionalFormatting, ...restKanban }` and then writes `titleField: kanbanCfg.titleField || 'name'` UNCONDITIONALLY, so a stored `kanban.cardTitle: ''` yields exactly `{ cardTitle: '', titleField: 'name' }` — the card's own reproduction — with no authored titleField needed at all; packages/plugin-list/src/ListView.tsx has the same shape (titleField stripped and re-emitted only when truthy, cardTitle not stripped). CAVEAT stated so PM can weigh it rather than inherit a headline: @objectstack/spec's KanbanConfigSchema is $strict and declares NEITHER cardTitle NOR titleField under a view's kanban block, so such a stored block is not spec-legal — the passthrough is simply unvalidated at runtime. The other half of section 3 is a clean NO: no authoring surface in this repo writes cardTitle at all (packages/plugin-designer has zero kanban view-config editor; the only non-test cardTitle sites tree-wide are plugin-kanban itself, the types faces, the registry declaration and docs).",
    "noted, not filed: ObjectKanbanComponentProps's docblock in the file I edited still cites ':350' / ':1024' for the titleField read. Its substantive claim (both union arms are load-bearing) is still true and only the addresses are stale — and they were ALREADY stale before this change, since the reads sat at :552/:1303. Class is a prose line-address citation, which objectui#8875 explicitly reserves the disposition of and which census:cross-file-line-citations reports without gating, so filing would pre-empt that card. Carrier: the next PR touching this prop docblock — this PR's own file.",
    "noted, not filed: the two view adapters above forward keys the spec's $strict KanbanConfig block does not declare (cardTitle through `...restKanban`). Not filed because the general property is already documented in place — ObjectView.tsx's own comment names 'author-written kanban.groupBy riding this repo's .passthrough() mirror' and the known residual is pinned as ObjectView.kanbanGroupByRetired-8213 — so a new card would most likely duplicate that tracked class. Carrier: whoever next works objectui#8213's residual.",
    "noted, not filed: packages/plugin-list/src/ObjectGallery.tsx gates on `titleField != null && item[titleField] != null` rather than on non-emptiness. Not a counterexample to the '' ruling and not a defect — item[''] is undefined so the second conjunct rejects it — but it is the one place in the tree that does not reject '' up front. Different component, different key, out of this card's fence. Carrier: any future sweep that unifies field-name emptiness handling."
  ]
}

Generated by Claude Code

Contract review pointer — director seat, summon #18 segment 6 (session_017Js5kTpTtxieBjPyScgxJ3, huangyiirene). PR objectui#8917 @ head 241b39f1f4: PASS, verdict comment on the PR: #8917 (comment) — tier-verified isolated review (69 stamps). PASS, no rework; F4–F6 non-blocking nits (regex-shaped census pin, three unpinned precedence cells, designer-face description lacks the '' clause). Overlap with #8865 measured clean in both orders. mergeable_state: behind is base drift only — queue-eligible.

Generated by Claude Code

needs:contract-review cleared on card objectui#8308 and PR objectui#8917 in one stroke — landing provenance on the PR: #8917 (comment) (verdict #8917 (comment)). Card state and assignee untouched until MERGED close-out.

Generated by Claude Code

落地注记 — PR #8917 已合并（`4fa0eb9c4`），按内容验收通过

⚠️ 本 PR 在实现席交报告之前就由维护者从合并队列直接合入，⇒ 派发席未及验收。本席裁的是这张卡的运算符语义（||：'' 视作未设置），因此补验它是否被正确实现 —— ⛔ 不按 sha，按内容。

围栏（读一次，走导出的解析器）—— ✅ 满足

export function resolveKanbanTitleField( 在 packages/plugin-kanban/src/ObjectKanban.tsx（⚠️ 行号是当刻读数，⛔ 后来者不要照抄）。
两处就地读已消失：全仓搜 schema.cardTitle || schema.titleField / ?? schema.titleField，可执行位置零命中；唯一残留在测试文件的 docblock 里，是描述旧状态的散文，⛔ 不是代码。

语义 —— ✅ 正是裁决那一条

if (schema?.cardTitle) return schema.cardTitle;
if (schema?.titleField) return schema.titleField;
return undefined;

'' 落到 titleField，与裁决一致。docblock 也把理由独立复述了：这是唯一一种不把空串默默当成「任何记录都不可能有的字段名」的读法。

⭐ 两处比派发单要求的更讲究

它刻意选了真值测试而不是 typeof === 'string' 收窄，并写明了理由：在一个字符串键上两者等价，但收窄会顺带开始丢弃两个旧运算符本来都放行的非字符串脏值 —— 而那是 objectui#8308 没有裁过的变更。⇒ 它把改动严格限制在被裁定的那一条上，并把「为什么不顺手多做一点」说了出来。⭐ 这正是本班反复强调的纪律，用在了一个我没有预见到的分支上。
它把结构性 pin 做成了源码普查：__tests__/ObjectKanban.titleFieldPrecedence-8308.test.tsx 在这个函数之外对任一键做属性读时会红。⇒ 比「两处都调用了解析器」强得多 —— 它让第三个读点在出现的那一刻就红，而「第三个读点会发明第三种优先级」正是卡片和分诊席共同立下的那条围栏所防的东西。

仍然留给后续的

⚠️ 派发单要求开工前做的那次便宜检查（有没有授权界面把清空的文本框回写成 ''、有没有 saved-view / 元数据读路径能吐出 ''）—— 若答案是「有」，本卡应从 p3 升为 p2。本席未在合并前拿到该读数；若实现席已量过，请把结论补在这条下面，⛔ 不必重开 PR。

Generated by Claude Code

Navigation Menu

Repository navigation

plugin-kanban: `cardTitle` falls back to `titleField` with `||` in the card list and `??` in the detail drawer — an authored empty string resolves two different ways on one board #8308

The two sites

Why the difference is observable

Scope note

Triage routing

零、⭐ 运算符：裁 `||` 语义（分诊席把这一步交给认领席，此处照办并说明理由）

一、⛔⛔ 围栏：结构修法不可协商

二、⛔ 锚点已漂移，⛔ 不要照抄任何行号

三、⚠️ 开工前先做的一次便宜检查（可能改定级）

四、判别力

五、⛔ 边界

六、⚠️ Clause-② —— 由你判 + 载体预授权

七、⭐ 位置轴与工具链

八、Clause-② —— dev 席测得，PM 预授权追写（第六节）

落地注记 — PR #8917 已合并（`4fa0eb9c4`），按内容验收通过

围栏（读一次，走导出的解析器）—— ✅ 满足

语义 —— ✅ 正是裁决那一条

⭐ 两处比派发单要求的更讲究

仍然留给后续的

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

plugin-kanban: cardTitle falls back to titleField with || in the card list and ?? in the detail drawer — an authored empty string resolves two different ways on one board #8308

Description

The two sites

Why the difference is observable

Scope note

Activity

os-zhuang commented on Sep 8, 2026

Triage routing

os-zhuang commented on Sep 9, 2026

零、⭐ 运算符：裁 || 语义（分诊席把这一步交给认领席，此处照办并说明理由）

一、⛔⛔ 围栏：结构修法不可协商

二、⛔ 锚点已漂移，⛔ 不要照抄任何行号

三、⚠️ 开工前先做的一次便宜检查（可能改定级）

四、判别力

五、⛔ 边界

六、⚠️ Clause-② —— 由你判 + 载体预授权

七、⭐ 位置轴与工具链

八、Clause-② —— dev 席测得，PM 预授权追写（第六节）

claude commented on Sep 10, 2026

huangyiirene commented on Sep 10, 2026

huangyiirene commented on Sep 10, 2026

os-zhuang commented on Sep 10, 2026

落地注记 — PR #8917 已合并（4fa0eb9c4），按内容验收通过

围栏（读一次，走导出的解析器）—— ✅ 满足

语义 —— ✅ 正是裁决那一条

⭐ 两处比派发单要求的更讲究

仍然留给后续的

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

plugin-kanban: `cardTitle` falls back to `titleField` with `||` in the card list and `??` in the detail drawer — an authored empty string resolves two different ways on one board #8308

零、⭐ 运算符：裁 `||` 语义（分诊席把这一步交给认领席，此处照办并说明理由）

落地注记 — PR #8917 已合并（`4fa0eb9c4`），按内容验收通过