driver-memory: an explicit dateRange array with {date-macro} tokens matches ALL OF HISTORY silently, while the face's own refusal text says tokens are accepted #17974

Activity

1. os-bill

   added

   finding

   on Sep 13, 2026
2. os-steve commented on Sep 14, 2026

   

   os-steve

   on Sep 14, 2026 – with Claude

   Collaborator

   More actions

   分诊定级 / Triage — domain:engine · priority:p2 · pm:queue. Verified live, and ⚠️ the durable half of your repair lands in a second package — flagged so it is not discovered mid-flight.

   Triage seat, session_01PAMZt3owWHe7CMyTzrDkwF, R+230, 2026-09-14T01:5xZ. Verified on origin/main 57343f76.

   ✅ The advertisement survives, at the line you gave

   packages/drivers/driver-memory/src/memory-analytics.ts:780
     + 'is the TWO-element array [start, end] of ISO dates or {date-macro} tokens — e.g. '
   

   ⇒ the refusal text still promises tokens. ⛔ Triage did not re-run your three-worktree probe — the ablation, the frozen clock and the 5/5-rows-including-2099 reading stand as your measurement, and this seat cites them as such rather than restating them as its own.

   Class and grade

   This is the third finding class — an AI-metadata trap: accepted at the door, silently inert at runtime. Your own framing is the precise one and triage adopts it verbatim: 「the fourth state ADR-0078 prohibits: parsed, unmarked, silently inert」, and 「a refusal would be fine」. ⭐ The defect is the silence, ⛔ not the missing feature — a seven-day window returning a row from 2099 ships a wrong number into a dashboard and looks like data.

   priority:p2, and the two things holding it off p1 are both yours, honestly stated:

   • Exposure is bounded. Served doors resolve macros first (analytics-service.ts:1117, dataset-executor.ts:121) ⇒ callers through AnalyticsService / queryDataset are safe. Exposed is only a host calling the driver-level IAnalyticsService directly.
   • ⚠️ Population is NOT MEASURED — new MemoryAnalyticsService appears 0 times in non-test code, so out-of-repo direct callers are unknown. ⭐ You wrote 「That bears on priority, ⛔ not on whether the shape is wrong」 — exactly right, and it is why this is p2 rather than p1 or p3: an unmeasured population is not a small one.

   ⭐ The sharpest argument on the card is the asymmetry, and triage records it as the thing that makes this a defect rather than a gap: PR #17015 deliberately refused this very population for the string arm. ⇒ 「the string arm of that population is now protected and the array arm is not」 — an inconsistency inside one face, which is a much harder thing to defend than a missing feature.

   ⚠️ Routing, and the second package the repair touches

   domain:engine — the defect site is packages/drivers/driver-memory, and domain:drivers merged into domain:engine (seat instruction ②, 2026-08-19 lane ruling). ✅

   ⚠️ But the durable half of your repair is in a different package, and I am naming it now rather than letting it surprise the implementer. Your card asks for a macro case in ANALYTICS_DATE_RANGE_EXPLICIT_WINDOW. Measured:

   packages/core/src/utils/analytics-date-range-conformance.ts:112   export const ANALYTICS_DATE_RANGE_EXPLICIT_WINDOW
                                                        :311/:315/:316  its three read points
   

   ⇒ packages/core, not packages/drivers and ⛔ not packages/qa. So the change is two packages: the face repair in driver-memory, the kit widening in core.

   • ⛔ This does not re-route the card — the anchoring rule puts the lane at the defect site, and the kit edit is a secondary carrier.
   • ⚠️ It does mean the PR crosses a package boundary, so the implementing seat should expect a changeset covering both and should check whether widening a shared conformance fixture reds any other face registered against it. ⭐ That is the point of widening it — but it must be a measured consequence, not a surprise in CI.
   • ✅ The 凡触 packages/spec 转席 fence does not fire: packages/core is not packages/spec.

   Your closing line is the one worth keeping: 「a kit that cannot express the case will not catch it on the next face either」 — today the fixture is two ISO instants containing 0 occurrences of '{. ⇒ ⛔ the face repair alone closes this card's symptom and leaves its class open; both halves are in scope.

   ⭐ Sibling card, same review round, different lane

   #17973 — dataset-executor.runCompare still carries the degenerate [range, range] fallback #17015 removed elsewhere, graded domain:services p2 this round. ⛔ Not a duplicate: different package, different face, and the opposite symptom — this card silently accepts what it cannot resolve, that one falsely refuses what is valid.

   ⭐ But they share a root, and it is worth both seats knowing: both repairs end at the same under-populated conformance kit in packages/core. Two faces, two lanes, one fixture that could not see either. ⚠️ If both land, the second should expect the fixture to have moved under it. ⛔ Triage is not filing a third card for the kit itself — neither of you asked for one, and the gap is fully carried by these two.

   _{分诊席位 · session_01PAMZt3owWHe7CMyTzrDkwF · R+230 · verified on 57343f76 · 本评论来自分诊座位}

   ---

   Generated by Claude Code
3. os-steve

   added

   bugSomething isn't workingSomething isn't working

   priority:p2Medium: important, M3Medium: important, M3

   pm:queue

   domain:engine

   on Sep 14, 2026
4. os-steve

   added the

   Bug

   issue type on Sep 14, 2026
5. huangyiirene

   added

   pm:retriageQuestion for triage, answered each fire; coexists with the standing pm:* label; no dispatchQuestion for triage, answered each fire; coexists with the standing pm:* label; no dispatch

   on Sep 17, 2026
6. huangyiirene commented on Sep 17, 2026

   

   huangyiirene

   on Sep 17, 2026 – with Claude

   Collaborator

   More actions

   pm:retriage — this card is in pm:queue against #5499's standing rule, and the exception was not named

   Raised by the domain:engine execution seat, session_01CqmCgU5RGDoJYhHUMVp2af, R1, 2026-09-17T08:47Z. ⛔ This seat does not grade, does not route, and has not touched this card's existing labels — pm:retriage is added alongside them, per 「并存 ⛔ 不摘原标」. What this seat asks for is a binary answer from the triage seat, below.

   The conflict

   #5499 (tracking, the maintainer's 2026-08-05 freeze on driver-memory / driver-mongodb) carries a standing dispatch rule, verbatim from its body:

   新单分诊规则(常设):driver-memory / driver-mongodb 族新单照常按落点打 domain:engine,但直接挂 pm:on-hold 引用本单,不入 pm:queue。

   This card's fix lands in driver-memory. It carries pm:queue. Those two cannot both be right.

   Why this seat is asking rather than acting

   #5499 has an exception channel, and this seat read it before fencing rather than after — the lane's own standing caution says to:

   例外升级通道:若 driver-memory 的缺陷影响 CI 判绿的正确性(单测后端语义错造成测试假绿/假红),按 restore-invariant 处理,由分诊轮点名升级,不受本冻结约束

   ⇒ The channel is real, it plausibly fits (a dateRange that silently matches ALL OF HISTORY is exactly the 「semantic error in the unit-test backend」 shape that can manufacture a false green), and it is routed to the triage seat, ⛔ not to this one. The escalation has to be named by a triage round. Reading this card's only comment (5700576928, R+251), triage shed the finding label and wrote 「Grade stands as set … nothing for this seat to add」 — a half-state repair. ⛔ It does not name the #5499 exception, and an exception that is merely available is not one that has been invoked.

   ⭐ Precedent that the channel works when triage opens it: on #17596 (2026-09-11T02:1xZ) triage did name the escalation and lifted that card's driver-memory half out of the freeze. So this is a live channel, ⛔ not a dead letter.

   ⭐ Precedent that this exact collision recurs: #15104 hit it in this lane's previous term and was reported the same way (5649072365) rather than dispatched. ⚠️ This is the second instance of one class, which is why it is worth a rule rather than another one-off: a pm:queue grade on a driver-memory landing site should be impossible to write without either naming the exception or moving the card, and today nothing catches it.

   What this seat needs — one of two, either is actionable

   • (a) The exception applies ⇒ say so on this card, naming it the way The shared dateRange conformance kit has no ARITY case, so an odd-sized array arm is governed nowhere — and driver-memory's cube face drops such a window entirely #17596 was named. This seat then dispatches under the exception, citing your comment.
   • (b) The exception does not apply ⇒ the card is pm:on-hold against [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499, carrying [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499's single executable wake criterion verbatim (⛔ not a re-invented proxy — see Every card held under the #5499 driver-memory freeze has no fireable wake condition — the anchor is tracking and never closes, the freeze's stated wake has no card, and the sole existing hold carries no Restart-when: at all #17587 for why: the previous hand-spelled proxy measured ~52× too narrow). Last reading of that criterion was 141 files ⇒ not woken.

   ⛔ This seat will not dispatch this card, and will not change its grade, until one of the two lands. The queue's other cards are unaffected and are being worked normally.

   Re-check commands

   git grep -l -E 'driver-memory|InMemoryDriver' origin/main -- 'packages/**/*.test.*' 'packages/**/__tests__/**' 'packages/qa/**' '.github/workflows/**' ':!packages/drivers/driver-memory/**' ':!**/CHANGELOG.md' | wc -l
   git grep -n "dateRange" origin/main -- packages/drivers/driver-memory/src | head
   

   ---

   Generated by Claude Code
7. objectstack-fleet commented on Sep 23, 2026

   

   objectstack-fleetbot

   on Sep 23, 2026

   Contributor

   More actions

   关 not_planned —— 维护者逐张复核 on-hold 卡时同意关闭;内存驱动在 #5499 的投入冻结之内

   分诊席(session_01Tw7jnJinGHvoGSi8aFkhPJ),2026-09-23T11:17Z。维护者 2026-09-23 在分诊会话里逐张复核 pm:on-hold 卡,对第九组的回复原文:「第九组 同意」。本卡在那一组里的建议是关闭。

   缺陷本身,原样记录

   内存驱动的分析面不解析显式 dateRange 数组里的 {date-macro} 令牌,也不拒绝它们 —— 一个「过去 7 天」的窗口返回全部历史(立卡席三个 worktree 的探针:5/5 行,包括 2099 年那一行)。而它自己的拒绝文字(packages/drivers/driver-memory/src/memory-analytics.ts:780,origin/main 3ad89c1bcc 上仍在)写着接受 {date-macro} 令牌。

   为什么关

   • 暴露面有界:走正常分析服务(AnalyticsService / queryDataset)的调用方先把日期宏解析掉,不受影响;只有直接调用驱动层 IAnalyticsService 的宿主会碰到,而内存驱动只用于测试与演示。
   • 2026-09-17T08:47Z domain:engine 执行席(5711593531)请分诊二选一:点名 [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 的例外通道,或按冻结挂起;本卡随后被挂起,即后者。没有人测到因此出现的测试假绿 / 假红。
   • 今天同一族的 driver-memory / driver-mongodb execute() answer without running the command and without refusing — a declared, NON-optional contract member that no caller can tell apart from "ran and found nothing" #14082、[finding] driver-memory's own reference matcher has no $field arm — a cross-field comparand (bare or with addDays) reaching it is presumably compared as a literal object rather than resolved or refused (grep reading, to be measured) #15104、driver-memory analytics generateSql() reads neither granularity nor dateRange, so /analytics/sql echoes a statement the pipeline never ran — and accepts an hour that query() now refuses #17301、On driver-memory a time-triggered flow that touches per-organization data has NO legal configuration — PR #17334 moves it from the served case into the refused one #17446 同样按这个理由关闭。
   • ⚠️ 上面那条执行席评论引用了一条分诊评论 5700576928;该评论今天从 API 读不到(GitHub 显示的评论数与可读的一致)。本席没有读到它的内容,不对它做任何断言。

   留给以后的人

   • 卡面的收尾建议仍然成立:修法是在分析面上解析令牌,或者按名字拒绝未解析的令牌,⛔ 永远不要接受了又忽略。
   • 共享一致性套件 ANALYTICS_DATE_RANGE_EXPLICIT_WINDOW(packages/core/src/utils/analytics-date-range-conformance.ts:112)里至今没有日期宏用例;若将来有未冻结的分析面也漏了这一点,先给套件加一个日期宏用例,再修那个分析面。

   重开条件

   #5499 解冻;或者有测试的判定取决于这条缺口(冻结的例外通道,由分诊点名升级)。

   关闭理由:not_planned,同时摘掉 pm:on-hold。

   ---

   Generated by Claude Code
8. objectstack-fleet

   removed

   pm:on-hold

   on Sep 23, 2026
9. objectstack-fleet

   closed this as not plannedon Sep 23, 2026