test(plugin-sharing): size the three scrypt cases near vitest's 5000 ms default to their measured work by objectstack-fleet[bot] · Pull Request #22426 · objectstack-ai/objectstack

objectstack-fleet · 2026-10-09T05:54:50Z

Fixes #22418
Clause-②: no

What changes

Three cases in packages/plugins/plugin-sharing get an explicit per-case timeout, sized to the work each one measures. Every other scrypt-reaching case in the package keeps vitest's 5000 ms default; the enumeration below gives each one's reading and the reason.

Case	Work (scrypt derivations)	Slowest `Test Core` reading	Budget
`share-link-password-webcontainer.test.ts:95`, "a password whose NFKC form differs from its input is one password on both paths" (the card's case, `:72` on `ca135dcc`)	4 pure-JS + 4 native	timed out at 5030 and 5039 ms; 5515 ms projected	`17_000`
`share-link-password-webcontainer.test.ts:74`, "a hash minted on WebContainer verifies natively, and the reverse"	3 pure-JS + 3 native	4136 ms	`13_000`
`share-link-password.test.ts:481`, "no log line carries the presented password, on any outcome"	9 native	3805 ms	`12_000`

Spelled the way this repo spells a per-case budget, as the third argument of it: two named constants in the webcontainer file and a literal 12_000 in the other. A comment beside each records the rule, the derivation count and the readings, so the next author can re-size from a fresh reading.

Not touched: the scrypt parameters (N, r, p), any production file, the vitest config, testTimeout anywhere, and the case bodies. No skip, no quarantine, no retry. No changeset: only two *.test.ts files change, and the package's files[] is dist, README.md and CHANGELOG.md.

The rule and the margin

Which cases get a budget: a case whose slowest CI reading is over half of the 5000 ms default. The default leaves every other case that has a CI reading at least 2x headroom over its slowest reading, and 2x is above the largest spread one case showed between the two CI runs: 1.88x (/resolve still accepts the ?password= query parameter, 588 ms in one run and 1104 ms in the other).
How big: 3x the slowest CI reading, rounded up to the next second. 3x covers that measured 1.88x spread with room for a heavier shard. The suspected trigger for these reds, per the card, is a shard re-pack (040184752c, chore(ci): refresh the Test Core shard-timings dataset #22368), which can change a shard's load again without touching this package.
The NFKC case's reading is projected, because CI cut it off at the timeout. Its derivations are 4/3 of the cross-verify case's (4 + 4 against 3 + 3), and the cross-verify case read 4136 ms in the same run, so the projection is 5515 ms. The 4/3 ratio checks out locally: the two cases read 926 and 675 ms idle (1.37x), and 5557 and 4682 ms under contention (1.19x). Under contention the projection overestimates, so it errs toward a larger budget.

Readings

CI (per-case durations from vitest's default reporter in the two red Test Core (5/6) job logs on the card, read through the job-log API): job 113665992253 (push, e75dceddd7) and job 113668129564 (PR #22215 merge of e75dceddd7). The log API returns only the last 5000 lines of a 19k to 48k line log. In both red jobs, plugin-sharing's last 24 files fall in that window, and the two share-link-eligibility.test.ts cases do not. Three hourly scheduled Test Core (5/6) jobs (113678118904 on ca135dcc, 113663806311, 113649948997) and the two green jobs on the card had no plugin-sharing lines in their windows. The direct REST log download is refused by this container's egress proxy, so no other log was reachable.

Local (a 4-vCPU container; every timing run was taken under the shared verify lock, before the gate battery started, with no other run alongside; the load average was 0.2 when the work began). The census and idle readings use --maxWorkers=2. The contention runs pin the four scrypt-reaching files to one core beside 5 busy loops on that core, with --maxWorkers=1. That load was picked because the NFKC case's CI-to-local ratio is at least 5.4x (at least 5030 ms against 926 ms). Case by case, it reads 0.95x to 1.48x the slowest CI reading, across the 22 cases that read over 1 s in CI and passed under it. So it is a load as heavy as the red shard, or heavier.

Case	Contention, before (`ca135dcc`)	Contention, after (`616794ce1`)	Idle, before	Idle, after
NFKC (webcontainer `:95`)	`Test timed out in 5000ms` (5024 ms)	passed, 5557 ms	958 ms	833 ms
cross-verify (webcontainer `:74`)	passed, 3935 ms	passed, 4682 ms	675 ms	681 ms
no-log (`share-link-password.test.ts:481`)	`Test timed out in 5000ms` (5067 ms)	passed, 5016 ms	789 ms	789 ms
whole contention run, 4 files	2 failed, 94 passed (96)	96 passed (96)
idle run, the two named files			48 passed (48)	48 passed (48)

The before contention run reproduces the card's failure shape: the same case, as a timeout and not an assertion. It also shows the no-log case is the next one to go.

The enumeration pin: every case in the package that runs a scrypt-class derivation

The only scrypt-class key derivation in the package is share-link-password.ts: deriveKeyNode (node:crypto scrypt) and deriveKeyPureJs (@noble/hashes scrypt). A test reaches them through hashShareLinkPassword / verifyShareLinkPassword, through ShareLinkService.createLink with a password, through resolveToken against a stored scrypt$ hash, or through the legacy upgrade.

The list is measured, not grepped: a temporary vitest setup file, never committed, wrapped node:crypto's scrypt and scryptSync, synced them into the builtin ESM namespace, and counted the calls per case. It ran over the whole package at ca135dcc: 40 files, 1002 cases, 1002 census rows. 52 cases in 4 files reach a derivation:

share-link-password-webcontainer.test.ts: 4 cases;
share-link-password.test.ts: 42 of its 44 cases. The two others are the injected-hasher case and the authenticated list-route case;
share-link-internal-columns.test.ts: 4 cases;
share-link-eligibility.test.ts: 2 cases.

Pure-JS derivations cannot be counted that way, because @noble/hashes is a native ESM namespace. They are counted by reading the four webcontainer cases, which are the only cases in the package that set a WebContainer signal (a grep for STACKBLITZ, /bin/jsh and webcontainer across the package's tests finds no other).

"Local, idle" is the census run. "CI" is the push / PR red run.

File:line at HEAD	Case	scrypt work	Local, idle (ms)	CI (ms), push / PR	Budget
`share-link-password-webcontainer.test.ts:95`	a password whose NFKC form differs from its input is one password on both paths	4 pure-JS + 4 native	926	timed out at 5030 / 5039	17_000 ms = 3 × 5515 (projected), rounded up
`share-link-password-webcontainer.test.ts:74`	a hash minted on WebContainer verifies natively, and the reverse	3 pure-JS + 3 native	675	4136 / 3688	13_000 ms = 3 × 4136, rounded up
`share-link-password-webcontainer.test.ts:121`	the pure-JS key is byte-identical to node:crypto for the same password and salt	1 pure-JS + 1 native	216	1435 / 1137	default: 29% of 5000 ms
`share-link-password-webcontainer.test.ts:64`	the WebContainer path does not touch node:crypto.scrypt; the native path does	1 pure-JS + 1 native	270	1305 / 1393	default: 28% of 5000 ms
`share-link-password.test.ts:481`	no log line carries the presented password, on any outcome	9 native	793	3805 / 3417	12_000 ms = 3 × 3805, rounded up
`share-link-password.test.ts:352`	legacy plaintext row › still verifies, refuses a wrong password without touching the row, and is upgraded by the right one	5 native	584	2163 / 1342	default: 43% of 5000 ms
`share-link-password.test.ts:352`	legacy salted SHA-256 row › still verifies, refuses a wrong password without touching the row, and is upgraded by the right one	5 native	731	2132 / 1815	default: 43% of 5000 ms
`share-link-password.test.ts:617`	/resolve sends no-store + Vary on success and on every refusal	3 native	280	1573 / 938	default: 31% of 5000 ms
`share-link-password.test.ts:338`	the verifier accepts the right password and refuses a wrong one, an empty one and an unknown form	4 native	332	1407 / 1356	default: 28% of 5000 ms
`share-link-password.test.ts:284`	the list and the redemption result, through the service and through the route, carry no hash	3 native	291	1387 / 951	default: 28% of 5000 ms
`share-link-password.test.ts:617`	/messages sends no-store + Vary on success and on every refusal	3 native	289	1337 / 942	default: 27% of 5000 ms
`share-link-password.test.ts:582`	/resolve never falls back to a raw compare: an undecodable value that IS the raw password is still refused	2 native	212	1244 / 992	default: 25% of 5000 ms
`share-link-password.test.ts:391`	a refused upgrade write does not block the read, leaves the legacy form verifying, and is reported once without the password	3 native	260	900 / 1138	default: 23% of 5000 ms
`share-link-password.test.ts:552`	/messages still serves a raw with a %-escape password sent raw, with no encoding header	2 native	174	1110 / 896	default: 22% of 5000 ms
`share-link-password.test.ts:445`	/resolve still accepts the ?password= query parameter	2 native	197	588 / 1104	default: 22% of 5000 ms
`share-link-password.test.ts:535`	/messages does not guess: the same encoded value with no encoding header is compared raw	2 native	195	1092 / 955	default: 22% of 5000 ms
`share-link-password.test.ts:600`	/messages: the ?password= form still wins, and the header pair is then not read	2 native	212	1075 / 886	default: 22% of 5000 ms
`share-link-password.test.ts:552`	/messages still serves a Latin-1 password sent raw, with no encoding header	2 native	187	1072 / 822	default: 21% of 5000 ms
`share-link-password.test.ts:523`	/messages serves a emoji password sent percent-encoded under utf-8	2 native	196	994 / 1046	default: 21% of 5000 ms
`share-link-password.test.ts:600`	/resolve: the ?password= form still wins, and the header pair is then not read	2 native	188	1044 / 828	default: 21% of 5000 ms
`share-link-password.test.ts:552`	/messages still serves a raw with a stray % password sent raw, with no encoding header	2 native	174	1038 / 951	default: 21% of 5000 ms
`share-link-password.test.ts:552`	/resolve still serves a raw with a %-escape password sent raw, with no encoding header	2 native	183	938 / 1033	default: 21% of 5000 ms
`share-link-password.test.ts:535`	/resolve does not guess: the same encoded value with no encoding header is compared raw	2 native	187	988 / 1000	default: 20% of 5000 ms
`share-link-password.test.ts:326`	a new link stores scrypt, with no plaintext, and two hashes of one password differ	2 native	172	710 / 951	default: 19% of 5000 ms
`share-link-password.test.ts:523`	/messages serves a CJK password sent percent-encoded under utf-8	2 native	170	878 / 950	default: 19% of 5000 ms
`share-link-password.test.ts:433`	/resolve accepts the x-share-password header	2 native	179	712 / 945	default: 19% of 5000 ms
`share-link-password.test.ts:445`	/messages still accepts the ?password= query parameter	2 native	223	943 / 700	default: 19% of 5000 ms
`share-link-password.test.ts:552`	/resolve still serves a raw with a stray % password sent raw, with no encoding header	2 native	202	937 / 738	default: 19% of 5000 ms
`share-link-password.test.ts:460`	/resolve refuses a wrong password sent as a header	2 native	196	841 / 913	default: 18% of 5000 ms
`share-link-password.test.ts:552`	/resolve still serves a Latin-1 password sent raw, with no encoding header	2 native	210	912 / 865	default: 18% of 5000 ms
`share-link-password.test.ts:267`	the mint response, through the service and through the route, carries no hash	2 native	225	799 / 900	default: 18% of 5000 ms
`share-link-password.test.ts:433`	/messages accepts the x-share-password header	2 native	206	824 / 897	default: 18% of 5000 ms
`share-link-password.test.ts:305`	the list and the redemption result carry no hash even from an engine that does not strip it	2 native	191	884 / 895	default: 18% of 5000 ms
`share-link-password.test.ts:460`	/resolve refuses a wrong password sent as a query	2 native	217	894 / 874	default: 18% of 5000 ms
`share-link-password.test.ts:523`	/resolve serves a emoji password sent percent-encoded under utf-8	2 native	178	843 / 819	default: 17% of 5000 ms
`share-link-password.test.ts:582`	/messages never falls back to a raw compare: an undecodable value that IS the raw password is still refused	2 native	221	795 / 834	default: 17% of 5000 ms
`share-link-password.test.ts:460`	/messages refuses a wrong password sent as a header	2 native	173	773 / 770	default: 15% of 5000 ms
`share-link-password.test.ts:460`	/messages refuses a wrong password sent as a query	2 native	208	652 / 741	default: 15% of 5000 ms
`share-link-password.test.ts:523`	/resolve serves a CJK password sent percent-encoded under utf-8	2 native	200	616 / 691	default: 14% of 5000 ms
`share-link-password.test.ts:568`	/messages refuses an octet that is not UTF-8 with 400 VALIDATION_FAILED, before the token is looked up	1 native	87	574 / 513	default: 11% of 5000 ms
`share-link-password.test.ts:568`	/resolve refuses an encoding the server does not read with 400 VALIDATION_FAILED, before the token is looked up	1 native	88	550 / 365	default: 11% of 5000 ms
`share-link-password.test.ts:568`	/resolve refuses an octet that is not UTF-8 with 400 VALIDATION_FAILED, before the token is looked up	1 native	110	538 / 484	default: 11% of 5000 ms
`share-link-password.test.ts:568`	/messages refuses an encoding the server does not read with 400 VALIDATION_FAILED, before the token is looked up	1 native	104	537 / 450	default: 11% of 5000 ms
`share-link-password.test.ts:568`	/resolve refuses a truncated UTF-8 sequence with 400 VALIDATION_FAILED, before the token is looked up	1 native	86	446 / 497	default: 10% of 5000 ms
`share-link-password.test.ts:568`	/messages refuses a truncated UTF-8 sequence with 400 VALIDATION_FAILED, before the token is looked up	1 native	92	496 / 455	default: 10% of 5000 ms
`share-link-password.test.ts:375`	is not attempted for a switched-off object, whose redemption is refused	1 native	96	303 / 439	default: 9% of 5000 ms
`share-link-eligibility.test.ts:1191`	the `password_hash` shape — NEEDS_PASSWORD and WRONG_PASSWORD both become the unknown-token answer	5 native	662	NOT MEASURED (outside both logs' tail windows)	default (eligibility note below)
`share-link-eligibility.test.ts:1291`	fail-closed: an object whose schema the engine cannot answer for is refused, not probed	2 native	336	NOT MEASURED (outside both logs' tail windows)	default (eligibility note below)
`share-link-internal-columns.test.ts:143`	a password-protected link refuses without the password and with a wrong one, and opens with it	3 native	318	1361 / 1528	default: 31% of 5000 ms
`share-link-internal-columns.test.ts:155`	the redemption route answers password-required, refuses a wrong password, and opens with the right one	3 native	295	1489 / 1289	default: 30% of 5000 ms
`share-link-internal-columns.test.ts:130`	the generic read omits both columns; storage holds both	1 native	150	774 / 566	default: 15% of 5000 ms
`share-link-internal-columns.test.ts:182`	the creator's link list carries each link's token, and no hash	1 native	95	391 / 416	default: 8% of 5000 ms

The eligibility note. Neither share-link-eligibility.test.ts case has a CI reading, because both finished before either red log's window opened. They keep the default, on two measured comparisons:

The same work as the legacy rows. The password_hash-shape case does 5 native derivations, exactly as many as each legacy row, and its idle reading is between theirs: 662 ms against 731 and 584 ms. The legacy rows' slowest CI reading is 2163 ms, 43% of the default.
Under the contention load it passed, at 2832 ms (before) and 2902 ms (after), beside the legacy rows at 2714 / 2638 ms and 2800 / 2988 ms in the same runs. That load reads up to 1.48x the CI shard, and the case was still at 58% of the default.

Its CI time is NOT MEASURED, and is stated as such rather than inferred into the table. The fail-closed case does 2 derivations and read 336 ms idle and 1113 ms under contention.

Mechanism assumptions, checked

"Two test files run scrypt-class derivations." Four do. share-link-internal-columns.test.ts (4 cases) and share-link-eligibility.test.ts (2 cases) reach the derivation through ShareLinkService. Neither is anywhere near the budget, so neither is edited, and the file surface stays the two claimed files.
"5030 / 5039 ms for the NFKC case and 1.1 to 4.1 s for its siblings." Confirmed from the same two job logs: siblings 1305 / 1393, 4136 / 3688 and 1435 / 1137 ms. The logs also show a third case near the budget in the other file, the no-log case at 3805 / 3417 ms. The contention run reproduces it as the next timeout.
"Follow test(rest): meta-state-route-engine-outage.test.ts's multi-kernel wiring case times out at vitest's 5000 ms default under the hourly full run, and has turned main's hourly run red three times in two days #21920's timeout spelling and margin comment." The precedent has no per-case timeout. test(rest): meta-state-route-engine-outage.test.ts's multi-kernel wiring case times out at vitest's 5000 ms default under the hourly full run, and has turned main's hourly run red three times in two days #21920 landed as PR test(rest): pay the state route's objectql load at collection, so the multi-kernel case stops timing out on the hourly run #21925 (7b6c65290), which states "No global testTimeout change, no per-case timeout". It moved a module load out of the clocked window into collection, because that cost was loading, not the case's own work. The shape does not transfer here. The cost here is the case's own work, the triage ruled that work stays, and the derivation's only module load (@noble/hashes, on the first pure-JS call) is small: the two one-derivation-per-path cases read 1305 / 1393 and 1435 / 1137 ms in CI. So the spelling follows the repo's own per-case convention instead: the third argument of it, with a comment stating why and how much. packages/drivers/driver-sql's "Why the it() below carries an explicit 60_000 budget" blocks are an example.

Tests and gates

All at 616794ce1 (git rev-parse --short HEAD), the PR head.

pnpm --filter '@objectstack/plugin-sharing^...' build (the dependency closure), under the verify lock: VERDICT command-exit 0.
pnpm --filter @objectstack/plugin-sharing test, under the lock: Test Files 40 passed (40), Tests 1002 passed (1002), VERDICT command-exit 0.
pnpm --filter @objectstack/plugin-sharing typecheck, under the lock: VERDICT command-exit 0, with check:test-typecheck: OK, 3 pre-existing ledgered errors held. tsc -p tsconfig.test.json --listFiles counts 40 src/**/*.test.ts in the program, both edited files among them.
The two named files, pnpm --filter @objectstack/plugin-sharing exec vitest run --maxWorkers=2 FILES: 48 passed (48), before and after. Per-case readings are in the table above.
The contention reproduction, before and after: shown in the readings table above.
Gates: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 57 commands for this diff, and all 57 exit 0. Three needed a second run, which is declared here:
- check:i18n and check:dual-build-cjs-loads first answered PREREQUISITE NOT MET (exit 3, no dist/). Each was run again after building its prerequisite (the i18n closure the gate names, then the workspace build), and both exit 0.
- check:type-check-debt first answered exit 3 because its own build collided with my concurrent closure build: plugin-webhooks' DTS step read a half-written service-messaging dist/. Run again alone, it exits 0.
- --ran reconciliation: "57 derived famil(ies) accounted for — 57 run, 0 NOT-MEASURED (a DERIVED zero — all 57 recorded an exit code and none of them is 3)".
Lint, narrowed to the two changed files: eslint --no-inline-config --format json on both files gives 0 errors and 0 warnings. The narrowing loses nothing, for three reasons:
- both files are inside eslint.config.mjs's population, since an ignored file would come back with an ignore warning;
- the JSON output counts 2 files;
- the config never enables type-aware linting (its own comment says so: no parserOptions.project, no typed rules) and loads no import-resolving plugin, so this diff cannot move the verdict on any file it does not touch.
The full pnpm lint is CI's run.
No ablation leg is needed beyond the before/after pair. The "before" leg is the committed ca135dcc tree, which is exactly this change removed, and it was run under the same load as the "after" leg.

Acceptance notes

share-link-password.test.ts's header lists, as pinned, "the pure-JS scrypt the WebContainer path uses and node:crypto's produce interchangeable hashes". No case in that file pins it: the pin lives in share-link-password-webcontainer.test.ts. This is a stale doc bullet, an observation only, and not filed. Carrier: none.
The job-log API returns only a log's last 5000 lines, and this container's egress proxy refuses the REST log download. So per-case CI readings were reachable only for files that finish late in their shard. That is why the eligibility cases have no CI reading. It is an observation about the measuring instrument, not about the code.
Re-sizing: when a budgeted case's derivation count changes, its budget changes by the same ratio, from a fresh CI reading. Lowering N or r to make a case fit is ruled out by the triage on the card.

Generated by Claude Code

…ms default to their measured work The share-link password cases derive scrypt keys at the production parameters on purpose, so their wall time scales with the shard that runs them. Three of them read over half of the 5000 ms default on Test Core: the NFKC case (timed out at 5030 and 5039 ms), the cross-verify case (4136 ms) and the no-log case (3805 ms). Each now carries an explicit budget of 3x its slowest reading, rounded up to the next second. The other 49 scrypt-reaching cases in the package keep the default. No parameter, production file or vitest config changes. Co-authored-by: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WYYhVJ78u7PhwFViWo1EmQ

github-actions · 2026-10-09T05:55:41Z

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 27a8b33dec2eb98b73b3e5146e740067cdbd7806 → packageMentionDocs.

github-actions · 2026-10-09T06:45:48Z

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 37893672824 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

分类：failure —— 按下面的日志分诊。

失败的 job（日志抽取，best effort）：

Test Core (5/6) — 失败步骤: Check this shard's timing drift（日志不可读，点进 job 看）

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

⚠️ 断言这一侧有一类例外，判据是断言在测什么，不是它是不是 AssertionError。 断言的对象是产品行为（一个值、一个形状、一次拒收）⇒ 照上面读：真实的行为改变，去查，⛔ 不要重排掉；
断言的对象是这次实验自身的有效性前提（跑完的耗时、负载下的先后、任何只在时间预算内才成立的条件）⇒ 它跟超时是同一类，同样对负载敏感，重排一次是合法的判别手段。
识别是机械的：断言的消息或它比较的值本身点名了一段时长、一个时间戳、一个耗时计数。实测过的一对 —— AssertionError: SecurityPlugin.init() ran: expected false to be true 测的是产品行为（真回归）；
AssertionError: this run took over a second, so second-precision stamps could have differed too: expected 1006 to be less than 1000 测的是实验前提：它守护的那条不变式当时是绿的，同一个 head 原样重排一次即成功。
穿着 AssertionError 外衣的时间测量，仍然是时间测量。（⛔ 这只改「怎么读一次红」，不改「哪些测试可以重排」——后者由别处管。）

跨 PR 相同签名（24h，按失败测试文件聚合）：

⚠️ 本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

本 PR 过去 24h 无队列失败记录（首次）。
过去 24h 队列共有 8 个失败构建（不含本次）。

分诊清单：

失败测试在本 PR 改动的包里 → 真回归，修 PR。
失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

Generated by Claude Code · merge-queue-triage workflow (#4859)

github-actions · 2026-10-09T06:51:22Z

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 37894053453 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

分类：failure —— 按下面的日志分诊。

失败的 job（日志抽取，best effort）：

Test Core (5/6) — 失败步骤: Check this shard's timing drift（日志不可读，点进 job 看）

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

⚠️ 断言这一侧有一类例外，判据是断言在测什么，不是它是不是 AssertionError。 断言的对象是产品行为（一个值、一个形状、一次拒收）⇒ 照上面读：真实的行为改变，去查，⛔ 不要重排掉；
断言的对象是这次实验自身的有效性前提（跑完的耗时、负载下的先后、任何只在时间预算内才成立的条件）⇒ 它跟超时是同一类，同样对负载敏感，重排一次是合法的判别手段。
识别是机械的：断言的消息或它比较的值本身点名了一段时长、一个时间戳、一个耗时计数。实测过的一对 —— AssertionError: SecurityPlugin.init() ran: expected false to be true 测的是产品行为（真回归）；
AssertionError: this run took over a second, so second-precision stamps could have differed too: expected 1006 to be less than 1000 测的是实验前提：它守护的那条不变式当时是绿的，同一个 head 原样重排一次即成功。
穿着 AssertionError 外衣的时间测量，仍然是时间测量。（⛔ 这只改「怎么读一次红」，不改「哪些测试可以重排」——后者由别处管。）

跨 PR 相同签名（24h，按失败测试文件聚合）：

⚠️ 本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

⚠️ 本 PR 过去 24h 已在队列失败 1 次（不含本次）。 内容未变而反复失败 ⇒ 高度怀疑 flaky 测试或与同组 PR 的语义冲突，重排不解决。
过去 24h 队列共有 9 个失败构建（不含本次）。

分诊清单：

失败测试在本 PR 改动的包里 → 真回归，修 PR。
失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

Generated by Claude Code · merge-queue-triage workflow (#4859)

github-actions Bot added the size/s label Oct 9, 2026

objectstack-fleet Bot assigned os-elon-musk Oct 9, 2026

github-actions Bot added the tests label Oct 9, 2026

objectstack-fleet Bot mentioned this pull request Oct 9, 2026

Flaky on Test Core: packages/plugins/plugin-sharing share-link-password-webcontainer.test.ts › "a password whose NFKC form differs from its input is one password on both paths" times out at vitest's 5000 ms default #22418

Closed

objectstack-fleet Bot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Oct 9, 2026

objectstack-fleet Bot marked this pull request as ready for review October 9, 2026 06:25

objectstack-fleet Bot enabled auto-merge October 9, 2026 06:25

objectstack-fleet Bot added this pull request to the merge queue Oct 9, 2026

objectstack-fleet Bot mentioned this pull request Oct 9, 2026

fix(auth,services): the services-lane in-process session reads stop renewing a cookie session (#22258) #22396

Merged

objectstack-fleet Bot mentioned this pull request Oct 9, 2026

ci(test-shards): the shard balance is derived on full-run sums while PR and merge_group runs use the affected set — the CLI shard (1/6) measures 34–36 min against 10–20 for the others and sets CI and queue wall time #22075

Closed

Merged via the queue into main with commit b9222dc Oct 9, 2026
39 of 40 checks passed

objectstack-fleet Bot deleted the claude/issue-22418-share-link-scrypt-timeout branch October 9, 2026 07:11

Navigation Menu

Repository navigation

test(plugin-sharing): size the three scrypt cases near vitest's 5000 ms default to their measured work - #22426

test(plugin-sharing): size the three scrypt cases near vitest's 5000 ms default to their measured work#22426
objectstack-fleet[bot] merged 1 commit into
mainfrom
claude/issue-22418-share-link-scrypt-timeout

objectstack-fleet Bot commented Oct 9, 2026

Uh oh!

github-actions Bot commented Oct 9, 2026

Uh oh!

github-actions Bot commented Oct 9, 2026

Uh oh!

github-actions Bot commented Oct 9, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

objectstack-fleet Bot commented Oct 9, 2026

What changes

The rule and the margin

Readings

The enumeration pin: every case in the package that runs a scrypt-class derivation

Mechanism assumptions, checked

Tests and gates

Acceptance notes

Uh oh!

github-actions Bot commented Oct 9, 2026

📓 Docs Drift Check

Uh oh!

github-actions Bot commented Oct 9, 2026

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

Uh oh!

github-actions Bot commented Oct 9, 2026

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants