[objectstack-fleet]

Requested by 田騏 · project thread
Part of #29777 (its second slice; the card closed on the first, PR #29838). Replaces PR #28978.
Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么:带入 2026-11-02 的分片计时刷新数据(与 PR chore(ci): refresh the Test Core shard-timings dataset #28978 逐字节相同),并把 CLI 的文件级切片从 4 改为 2,旧映射 {cli: 4} 移入 PREVIOUS_FILE_SHARDED_PACKAGES。
• 为什么改:新数据里 CLI 降到 1432 秒,自检的切片数推导(pin 3c)只接受 2 片(716 秒 ≤ 736.45 秒);数据与映射必须同一 PR 落地(ci(test-shards): refresh the Test Core dataset and cut the CLI into 4 slices on plain-sum placement, and pin the whole-package leg #26186 形状),ci(test-shards): the density cap carries the declaration closure's own displacement when a split carries the reader slice #29838 已让上限接受带声明片的切法。
• 风险与回滚:只动两个仓根脚本/数据文件,不发布任何包。回滚即 revert 本 PR,回到 4 片与旧数据。pin 3c 余量仅 20 秒,另见 finding ci(test-shards): the slice-count pin has no hysteresis, so a 20 s drift flips the CLI map and turns the weekly refresh red again #29826。
• 先看:scripts/partition-test-shards.mjs 里 FILE_SHARDED_PACKAGES 上方新增的「RE-DERIVED AT 2」推导表与两处映射。
• 两分钟验证:node scripts/partition-test-shards.mjs --self-test,应见 self-test OK 且 file-level slices: @objectstack/cli x2,最密分片 1553.36s,在 1455.81s 上限加 111.86s 闭包分摊之内。
• 需要你做的:无,席位落地后关闭 chore(ci): refresh the Test Core shard-timings dataset #28978。

What changes

Implements the second slice of maintainer ruling record 6596963071 (option B) on #29777.

• scripts/test-shard-timings.json: the generated refresh of 2026-11-02 (72 run summaries), byte-identical to PR chore(ci): refresh the Test Core shard-timings dataset #28978 head b5b4b790 (git diff b5b4b790cb HEAD -- scripts/test-shard-timings.json is empty). Not hand-edited.
• scripts/partition-test-shards.mjs:
  • FILE_SHARDED_PACKAGES changes @objectstack/cli from 4 to 2. Pin 3c derives it: the CLI is 1432.04 s, and the heaviest other serial task is metadata-protocol at 566.50 s, so a slice must be within 1.3 × 566.50 = 736.45 s. At n = 1 the CLI is 1432.04 s (2.53x); at n = 2 each slice is 716.02 s (1.26x), which meets the target. The derivation table sits above the map.
  • PREVIOUS_FILE_SHARDED_PACKAGES = {cli: 4}, the outgoing map, so the refresh lane still decodes the 4-slice summaries it keeps.

Not changed, as ruled: ci.yml's DTS_READER_SLICE and the CLI vitest config's DECLARATION_READER_SLICE.

Verification

• node scripts/partition-test-shards.mjs --self-test exits 0. The bins are 1551.96/1551.44/1553.29/1552.46/1552.43/1553.22/1551.01/1553.36 s, within the 1455.81 s density cap plus its 111.86 s declaration-closure allowance. The whole legs stay within 1868 s, and the CLI is cut x2.
• dispatch-gates --commands: 33 derived, 33 run, all exit 0. --ran reads 0 NOT-MEASURED (a derived zero).

🤖 Generated with Claude Code

[objectstack-fleet]

domain:devx#2 · marchtian · session_017ihL2gLghEqeDKTbDTC7Zd

席位核对 — ACCEPT · domain:devx#2 · head 10ada71679 · 2026-11-03T07:56Z

The seat wrote this PR itself; it is the second slice of ruling record 6596963071 on #29777, so the record states what was checked, not what was claimed.

• Shape: a draft to main carrying Part of #29777 (the card closed on ci(test-shards): the density cap carries the declaration closure's own displacement when a split carries the reader slice #29838), Clause-②: no and Metadata-protocol: none. No assignees, no reviewers, skip-changeset (root script and data only).
• Path surface: check-governed-merges --pr 29856 reads NOT governed (686 changed lines, mostly the generated dataset).
• The ruling: scripts/test-shard-timings.json is byte-identical to PR chore(ci): refresh the Test Core shard-timings dataset #28978 head b5b4b790 (taken with git show, not hand-edited). FILE_SHARDED_PACKAGES cli 4→2, PREVIOUS_FILE_SHARDED_PACKAGES = {cli: 4}, and the derivation table sits above the map. ci.yml's DTS_READER_SLICE and the CLI vitest DECLARATION_READER_SLICE are untouched.
• Gates: --self-test exits 0 with the CLI x2 and the densest shard at 1553.36 s, within 1455.81 + 111.86 s. dispatch-gates --commands derived 33, ran 33, all exit 0; --ran 0 NOT-MEASURED. The seat lands once every check on this head is green.
• Out of scope: pin 3c's 20 s margin is finding ci(test-shards): the slice-count pin has no hysteresis, so a 20 s drift flips the CLI map and turns the weekly refresh red again #29826. Skills-docs: none — CI-internal.
• Landing: non-governed, so the PR is marked ready and auto-merge is enabled. After it merges, PR chore(ci): refresh the Test Core shard-timings dataset #28978 closes as superseded.