tooling(scripts): assertion floors for the class-2 self-tests, one hoisted battery each (#13799 batch 5) by claude[bot] · Pull Request #15217 · objectstack-ai/objectstack

claude · 2026-09-04T06:20:18Z

Batch 5 of the scripts/** assertion-floor work: the class-2-ONLY deferrals from PR #15003's table — files that already carry a block-bodied in-body assertion helper and a body-level verdict site, but fewer than the two named section banners the sectioning criterion needs.

9 floored, 2 deferred with a measured reason. Every candidate was re-verified at the pin 50d6c924b rather than taken from the table; two of the eleven turned out to be class-1 as well, and are deferred rather than forced.

The shape

One battery per file, opened at the top of the self-test body — the hoisted shape PRs #14896 and #15003 landed for files that run cases before any banner — named after the file's self-test, with the floor at the measured case count, SELF_TEST_BATTERIES roster size pinned at 1, and a verdict that refuses below-floor / DID NOT RUN. registerCase() is called from the existing helper's block body; no case is rewritten and no assertion changed.

⛔ No comment is promoted to a section head — that is a judgement per comment this transplant does not make (b2's rule). ⛔ Where a file carries exactly one real banner (check-label-desc-cap), it still gets one battery; a lone banner is not split. Every verdict handshake is kept exactly as landed, in the spelling each file already uses — five return the SELF_TEST_VERDICT sentinel, four set a selfTestReachedVerdict flag because their own exit code stays load-bearing.

Floored (9)

file	batteries	floor (cases)	handshake	sink
`scripts/check-dts-emitted.mjs`	1 (hoisted)	8	flag	`eq`
`scripts/check-logger-receiver-detach.mjs`	1 (hoisted)	47	sentinel	`expect`
`scripts/check-objectql-double-limit.mjs`	1 (hoisted)	47	sentinel	`expect`
`scripts/check-spec-parsed-alias.mjs`	1 (hoisted)	18	sentinel	`check`
`scripts/check-stack-collection-maps.mjs`	1 (hoisted)	18	flag	`eq`
`scripts/check-tenant-chokepoint.mjs`	1 (hoisted)	17	sentinel	`assert`
`scripts/check-whole-set-label-write.mjs`	1 (hoisted)	41	flag	`expect`
`scripts/check-wildcard-fallthrough.mjs`	1 (hoisted)	18	sentinel	`assert`
`scripts/pm/check-label-desc-cap.mjs`	1 (hoisted)	44	sentinel	`t`

258 cases floored across the nine. Two files needed the floor written in their own idiom rather than the landed one, and say so in place: check-label-desc-cap's sink is a counter rather than a list, so a breach prints and folds into failed; check-wildcard-fallthrough refuses in place instead of collecting, so its floor refuses in place too.

⭐ Two floors that are deliberately NOT "every case in the file"

check-whole-set-label-write registers from expect() only. Its two trailing loops walk WHOLE_SET_ACTIONS and ALLOWLIST and push straight to failures, one iteration per row. They are left unregistered on purpose: ALLOWLIST is an allowlist, so it is meant to SHRINK, and a floor that moved with it would red every legitimate removal and train the next author to edit the floor — the one habit these floors exist to prevent (PR Survey: which scripts/** self-tests cannot prove they ran — and the two that now can #13797's ruling, which 155 of 158 scripts/** self-tests have no assertion floor: a battery that never ran is indistinguishable from one that passed #13799's body carries forward). The floor pins the part that does not move with a list. The comment above the roster states this where the next reader will be standing.
check-stack-collection-maps already COUNTED its assertions and printed the number — but nothing compared it, so a case block that stopped running shrank the printed count and the gate stayed green. Evidence, not proof. Its asserted counter and the new ledger now agree at 18, and only the ledger is enforced.

Deferred (2) — measured, and a correction to PR #15003's table

Both were recorded there as class-2 only. Measured at the pin they are class-1 as well: neither has a single block-bodied in-body assertion helper. Registering from a for loop body has no landed precedent — the two landed multi-site files (scripts/pr-labels.mjs, scripts/docs-audit/check-audit-scope.mjs) register from two block-bodied helpers, not from loops — so forcing them here would be improvising a new shape on a file whose census class would then read ROSTER while most of the hole stayed open.

file	measured reason
`scripts/check-startup-registry-verdict.mjs`	class-1, not previously recorded. Cases are decided in TWO places: the `for (const c of cases)` loop body, which increments `failures` once per table row (base L1366), and the `expectRoot` helper (base L1380). Registering from `expectRoot` alone would floor only the four dead-root cases and leave the ~20-row analysis table unfloored.
`scripts/js-comment-mask.mjs`	class-1, not previously recorded. The corpus half is a table loop with an inline `failed++` (base L508); the interpolation half uses a concise-arrow sink, `const x = (name, ok, detail) =` followed by an arrow to `extra.push(...)` (base L526), which has no block body to hold `registerCase()`. Concise-arrow sinks are batch 3 / batch 4 territory (the thunk route PR #15156 measured).

⛔ scripts/pm/dispatch-gates.mjs (DEFERRED by the card), scripts/pm/check-half-states.mjs (its own batch) and scripts/partition-test-shards.mjs (batch 4, PR #15198, in flight) are untouched, as are all class-1 and class-3 files.

Evidence

Behaviour is unchanged, per file

--self-test run in a base worktree at 50d6c924b and in this one, exit codes captured by redirect before any pipe, the two tree roots normalised away, stdout and stderr compared in full rather than sampled:

9/9 exit 0 on both sides, with stdout AND stderr byte-identical on every one. The registered-case ledger is a Map write and nothing prints it, so no verdict line moves.

check-dts-emitted is not run by any check: script — it is invoked by about eleven packages' own build scripts — so its normal mode was run too, from packages/client on both trees: output byte-identical, exit 1 on both sides (that package has no dist in either worktree; container state, present identically on the base, and not a verdict about this diff).

Case counts: measured, never transcribed

Each floor was read from the file's own report rather than counted by eye: the roster was pinned to a deliberately unreachable value, the self-test run, and the floor's own breach line (registered N case(s), below its pinned floor of ...) names N. The pin was then set to N. No temporary statement was injected into any file to obtain a count.

Two files carry a hand-written count in their verdict line. check-spec-parsed-alias (18 assertions passed) and check-stack-collection-maps (18 assertions) both agree with the measured count — neither had drifted.

⚠️ One transcribed count HAS drifted, and is filed rather than fixed here

check-wildcard-fallthrough prints ✓ self-test: 17 cases while its body has 18 assert( call sites and registers 18. The literal is left untouched so this change stays a pure no-op on output — correcting it would break the byte-identical proof that is this PR's whole evidence — and it is filed as #15231 (unassigned). A comment records the measurement in place. The floor pinned at the measured 18 is what makes the drift harmless: the count can no longer shrink in silence.

Ablations — three shapes, eleven legs, each from the committed implementation

Every leg registers its restore before the mutation, on absolute paths. The mutation is proved on disk by an exact-substring count and a git hash-object that moved off the HEAD blob's (an empty hash is read as FAILURE, not as "nothing to compare"); the restore is proved by hash equality against the HEAD blob and an empty git diff HEAD — never by an exit code, and never by a bare git checkout -- path, which reads the index the mutation also wrote. Anchors must match exactly one occurrence or the leg aborts. None of these nine files resolves through a dist, so no rebuild leg applies.

The battery opener renamed, all nine files (so the declared battery registers nothing). Exit 0 → 1 on every one, each naming itself: self-test battery "NAME self-test" DID NOT RUN — 0 cases registered, N pinned. The verdict below would have claimed those cases hold. Restored to exit 0 with hash equality on all nine.
A roster row deleted — check-tenant-chokepoint's only entry. Exit 0 → 1, naming both halves: SELF_TEST_BATTERIES declares 0 batteries, below the pinned 1 — a battery deleted from the roster takes its own floor with it and battery "check-tenant-chokepoint self-test" registered 17 case(s) but is not declared in SELF_TEST_BATTERIES. So the roster-size pin is load-bearing at size 1, not decorative.
The introduced registerCase() removed from the sink — check-whole-set-label-write. Exit 0 → 1 with every original case still passing: DID NOT RUN — 0 cases registered, 41 pinned. The file's own green verdict is exactly what the floor refuses.

The census moves by exactly the worklist

node scripts/measure-self-test-floor.mjs --json, computed as set assertions rather than read by eye:

	base `50d6c924b`	this branch `4d7538c4c`
ROSTER	111	120
COUNT	2	2
NONE	55	46

unfloored NONE drops by exactly 9, the number of files floored — True;
the set of files whose class CHANGED equals the worklist exactly, symmetric difference empty — True, and every transition is NONE - to ROSTER;
all nine worklist files read ROSTER after — True;
git diff --name-only against the recorded base equals the worklist exactly (9 files) — True.

Battery names are not path populations

The #15003 hazard: a battery name that reads as a declared path population changes what a gate thinks it watches. Every name here is BASENAME self-test, which contains a space and so cannot pass extractWatchHints' admission test at all. Proved rather than argued: extractWatchHints run over each of the nine files on both trees, hints sorted and compared — 9/9 identical (0, 8, 3, 1, 8, 0, 13, 0 and 1 hints respectively, unchanged). pnpm check:watch-hint-literal exit 0.

Gates

Derived with node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack from this worktree after the edits — no paths passed; the script reads the change set from the merge base. 28 commands derived. Added on top, because a derived family is not a complete account of CI: each edited script's own --self-test, plus the always-runs tail (check:self-test-wired, check-self-test-workflow-commands.mjs and its --self-test, check:declared-population-live, check:ratchet-remedy-authority, check:watch-hint-literal, check:nul-bytes). 39 commands run, every exit code captured by redirect before any pipe.

39/39 RAN-PASS (exit 0). Including check:pm-dispatch-gates — exit 0, dispatch-gates self-test: 1353 cases pass — run in the foreground, plus check:declared-population-live, check:ratchet-remedy-authority, check:watch-hint-literal and check:nul-bytes, and the nine per-file --self-test runs.
Nothing read NOT MEASURED, and no gate needed a container-state re-run.
pnpm lint (whole repo, eslint . --no-inline-config) — exit 0 on 4d7538c4c, 81 s. No narrowing claimed.
Control-byte scan over all nine changed files (grep -naP over the C0 range plus DEL) — no hits.

All of the above ran on 4d7538c4c, which is this branch's head, with git status --porcelain empty.

ℹ️ The derivation printed a STALE TREE notice: origin/main has advanced 5 commits since the pin, and 11 files the derivation reads from changed across that range. The branch is not merged forward, deliberately — none of those commits touches any of the nine files here (git merge-base is still the recorded pin, and the diff against it is exactly the nine), so there is nothing to reconcile and a merge would only invalidate every measurement above. Recorded rather than silently absorbed; CI runs the farm against the real merge base regardless.

scripts/** publishes nothing ⇒ skip-changeset, applied at PR-open time.

⛔ Left as a draft, not armed and not flipped ready — the PM does that.

🤖 Generated with Claude Code

https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

Generated by Claude Code

… hoisted battery (#13799) The self-test decided success by `failures.length === 0` alone, so "every case held" and "the cases never ran" printed the same line. It carries no named section banner, and a comment is not promoted to a section head, so it declares ONE battery opened at the top of the body -- the hoisted shape PRs #14896 and #15003 landed for exactly this case. Floor 8, measured rather than transcribed: the roster was probed with a deliberately unreachable pin and the floor's own report named the registered count. Behaviour is unchanged -- `--self-test` stdout and stderr are byte-identical against origin/main, exit 0 both sides. Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk Co-authored-by: Claude <noreply@anthropic.com>

…e-limit self-tests (#13799) Both decided success by `failures.length === 0` alone. Neither carries the two named section banners the sectioning criterion needs, and a comment is not promoted to a section head, so each declares ONE battery opened at the top of its self-test body -- the hoisted shape PRs #14896 and #15003 landed. Floors 47 and 47, measured by probing the roster with an unreachable pin and reading the count the floor's own report names. `--self-test` stdout and stderr are byte-identical against origin/main for both, exit 0 on both sides. Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk Co-authored-by: Claude <noreply@anthropic.com>

…ps self-tests (#13799) check-spec-parsed-alias decided success by `failures.length === 0` alone. check-stack-collection-maps already COUNTED its assertions and printed the number -- but nothing compared it, so a case block that stopped running shrank the printed count and the gate stayed green: evidence, not proof. Neither carries the two named section banners the sectioning criterion needs, and a comment is not promoted to a section head, so each declares ONE battery opened at the top of its self-test body. Both floors measured at 18. The two hand-written verdict counts agree with the registered count on this tree (`18 assertions passed`, `18 assertions over synthetic sources`), so neither had drifted. `--self-test` stdout and stderr are byte-identical against origin/main for both, exit 0 on both sides. Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk Co-authored-by: Claude <noreply@anthropic.com>

…-tests (#13799) Both decided success by their sink alone (`failures.length === 0`, `failed === 0`), so "every case held" and "the cases never ran" printed the same line. tenant-chokepoint's blocks are headed by unmarked prose and label-desc-cap carries one banner, below the two the sectioning criterion needs; a comment is not promoted to a section head and a lone banner is not split, so each declares ONE battery opened at the top of its self-test body. label-desc-cap's sink is a counter rather than a list, so its floor prints the breach and folds it into `failed` -- the same refusal in that file's own idiom. Floors 17 and 44, measured by probing the roster with an unreachable pin. `--self-test` stdout and stderr are byte-identical against origin/main for both, exit 0 on both sides. Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk Co-authored-by: Claude <noreply@anthropic.com>

…rough self-tests (#13799) Neither carries the two named section banners the sectioning criterion needs, so each declares ONE battery opened at the top of its self-test body. Floors 41 and 18, measured by probing the roster with an unreachable pin. whole-set-label-write registers from `expect()` only. Its two trailing ledger loops walk WHOLE_SET_ACTIONS and ALLOWLIST and push straight to `failures`; they are left unregistered ON PURPOSE, because ALLOWLIST is meant to shrink and a floor that moved with it would red every legitimate removal -- the one habit these floors exist to prevent. The comment above the roster says so in place. wildcard-fallthrough refuses in place rather than collecting, so its floor refuses in place too. ⚠️ Its verdict prints a TRANSCRIBED `17 cases` while the body has 18 assert call sites and registers 18 -- the literal had already drifted one low. It is left untouched so this change stays a pure no-op on output, and filed separately; the floor is what makes the drift harmless. `--self-test` stdout and stderr are byte-identical against origin/main for both, exit 0 on both sides. Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk Co-authored-by: Claude <noreply@anthropic.com>

…y roster `cases` with no failing entry was this self-test's only success condition, so "every case held" and "the cases never ran" printed the same line. Batch 6b of the roster-floor transplant, two shapes at once: - the sink repair PR #15156 landed: the concise arrow `const t = (name, ok, detail) => cases.push(...)` gains a block body and registers the case before recording it. `cases.push` receives exactly the arguments it always did -- no case is rewritten, reordered or re-judged. - the single hoisted battery PR #15217 landed: this file carries no named section banner, so it declares ONE battery opened at the top of the self-test body, floor at the measured count (12), and pins the roster's own size at 1. No comment is promoted to a section head. A breach files into the self-test's own `cases` sink, so the existing verdict reds on it with no verdict line rewritten; the #13798 handshake is untouched. Cases before == after: 12 == 12, measured on runs. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

…-test `cases.filter((c) => !c.cond)` was this self-test's only success condition, so "every case held" and "the cases never ran" printed the same line (#13489). Class-1 sink repair, the PR #15156 shape: the concise-arrow `ok` sink gains a block body that calls `registerCase()` before the unchanged `cases.push`. No assertion condition is inverted or rewritten. Class-2 roster, the PR #15217 shape: ONE battery hoisted to the top of the self-test body, floor at the measured 23, `SELF_TEST_BATTERIES` size pinned at 1. No comment was promoted to a section head. Case count before == after, measured by pinning the roster to an unreachable value and reading the breach line: 23. `--self-test` stdout and stderr are byte-identical to the base tree's, exit 0 on both. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

…attery roster `failed === 0` was this self-test's only success condition, so "every case held" and "the cases never ran" printed the same line. Batch 6b, two shapes at once: - the sink repair PR #15156 landed: the 12 inline `cases.push({...})` sites route through ONE block-bodied in-body helper, `addCase`, which registers the case and then performs the identical push. The case object is passed through untouched and the loop that runs the cases is not touched at all -- no case is rewritten, reordered or re-judged. - the single hoisted battery PR #15217 landed: this file carries no named section banner (its `--- n/m: ... ---` comments label fixtures, not sections), so it declares ONE battery opened at the top of the self-test body, floor at the measured count (12), roster size pinned at 1. No comment is promoted to a section head. A breach files into the self-test's own `failed` counter, so the existing verdict reds on it with no verdict line rewritten; the #13798 handshake is untouched. Normal mode stays green on the tree. Cases before == after: 12 == 12, measured on runs. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

`cases.filter((c) => !c.ok)` was this self-test's only success condition, so "every case held" and "the cases never ran" printed the same line (#13489). Class-1 sink repair, the PR #15156 shape: the concise-arrow `t` sink gains a block body that calls `registerCase()` before the unchanged `cases.push`. No assertion condition is inverted or rewritten. Class-2 roster, the PR #15217 shape: ONE battery hoisted to the top of the self-test body, floor at the measured 20, `SELF_TEST_BATTERIES` size pinned at 1. The file's single named section banner is NOT split on, and no comment was promoted to a section head. Case count before == after, measured by pinning the roster to an unreachable value and reading the breach line: 20 — which agrees with the count the existing verdict line prints. `--self-test` stdout and stderr are byte-identical to the base tree's, exit 0 on both. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

`problems.length` / `failed` were these self-tests' only success condition, so "every case held" and "the cases never ran" printed the same line (#13489). Class-1 sink repair via the THUNK route PR #15198 measured, because each sink write is failure-only: routing failure-only writes through a helper would floor zero cases on a green run, which is no floor at all. `check(() => { ... })` registers the case and then runs the existing site VERBATIM, so registration happens whether or not the site fires and no assertion condition is inverted or rewritten. Per-iteration registration through a block-bodied helper is the landed shape (check-doc-frontmatter, check-test-source-alias both call their helper from inside a `for`). Class-2 roster, the PR #15217 shape: ONE battery hoisted to the top of each self-test body, floor at the measured count, roster size pinned at 1. No comment was promoted to a section head; a single banner was not split on. Floors, each measured by pinning the roster to an unreachable value and reading the breach line -- never transcribed: check-meta-type-normalized 4 (4 thunked sites) check-org-identifier 31 (agrees with its printed `cases.length + 1`) check-error-code-casing 51 (agrees with its printed 46 + 5) `--self-test` stdout and stderr byte-identical to the base tree's on all three, exit 0 both sides; normal mode also byte-identical where the gate has one. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

…ests Same repair as the previous commit: the THUNK route PR #15198 measured, because every sink write in these three is failure-only, plus PR #15217's single hoisted battery with the roster size pinned at 1. Floors, each measured by pinning the roster to an unreachable value and reading the breach line -- never transcribed: check-console-intercept-disarm 11 (its printed 10 cases + the real-tree floor) check-examples-live-imports 36 (agrees with its printed `cases.length`) check-optional-error-sink-contract 66 (its printed 19 + 9x4 spellings + the reject side + 9 derivation + the `run` reference pin) check-examples-live-imports' one sink site is an if/ELSE, so it is wrapped by hand rather than by a brace matcher that would stop at the `if` block's own closing brace; the branches are reindented, not rewritten. ⛔ check-optional-error-sink-contract's two assertions over `baseline.entries` are deliberately left UNREGISTERED and say so in place: they run one-per-row of a shrink-only ledger, and a floor moving with that list would red every legitimate removal (#13797's ruling, carried forward by PR #15217's check-whole-set-label-write). `LOG_CHANNELS` is not that -- it is the contract's own vocabulary, not a list meant to shrink -- so its loops do register. `--self-test` stdout and stderr byte-identical to the base tree's on all three, exit 0 both sides; normal mode byte-identical where the gate has one. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

…tered battery The 22 bare `if (...) throw ...` assertions in `selfTest()` are each wrapped verbatim in the `check(() => { ... })` thunk PR #15198 landed: the condition is carried in untouched, so no `!` can be dropped while the case still registers and still passes. Zero named section banners, so one hoisted battery (the PR #14896 / #15003 / #15217 shape), floored at the measured 22 with the roster's own size pinned at 1. The floor throws, matching this file's own sink, and is evaluated before the verdict so the success line can only be printed by a run that actually ran its cases. Part of #13799 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012zGPuVVX3deAx9LdjK8jCk

The b5 recipe (PR #15217) applied to the remaining eight census rows of #13799 batch 8a. Each file gets ONE battery opened at the top of its self-test body, named `BASENAME self-test`, with the floor read off a run (the roster pinned at 9999 first, the breach line naming N), the roster's own size pinned at 1, `registerCase()` called from the block body of the helper the file already has, and a verdict refusing below-floor / DID NOT RUN / undeclared batteries through the file's own failure sink. Floors measured on a run: check-adr-symbol-anchors 17, symbol-anchors 51, check-i18n-walk-parity 23, check-test-completeness 67, checklist-select 17, release-rehearsal-clone 32, render-release-coverage-anchor 10, run-with-stall-guard 41. Two of them are the census's own warning coming true: symbol-anchors has 40 static `assert(` sites but registers 51 (loops), and render-release-coverage-anchor has 8 static `expect(` sites but registers 10. A floor counted from the source would have been wrong in both. Where the helper is module-level (check-adr-symbol-anchors, symbol-anchors, render-release-coverage-anchor) the case sites call a thin in-body wrapper that registers and then defers to the existing assertion, exactly as PR #15156 landed for that shape; no assertion condition is touched. checklist-select also stops transcribing its case count: the success line's hand-typed `17` is now read off a counter (#15305). It renders the same text today, which is what makes the byte comparison across the change readable. check-test-completeness scopes its floor to the loud run, because `selfTest({ quiet: true })` also runs on every production invocation of that gate, where nothing claims a self-test verdict. Part of #13799 Fixes #15305 Co-authored-by: Claude <noreply@anthropic.com>

claude Bot added the skip-changeset PR has no user-facing published change; bypasses the changeset gate label Sep 4, 2026

github-actions Bot added the size/s label Sep 4, 2026

github-actions Bot added size/m and removed size/s labels Sep 4, 2026

claude added 2 commits September 4, 2026 06:23

github-actions Bot added size/l and removed size/m labels Sep 4, 2026

claude Bot mentioned this pull request Sep 4, 2026

[finding] check-wildcard-fallthrough self-test prints a transcribed 17 cases while the body asserts 18 — a stale count nothing derives or compares #15231

Open

baozhoutao marked this pull request as ready for review September 4, 2026 06:48

claude Bot mentioned this pull request Sep 4, 2026

155 of 158 scripts/** self-tests have no assertion floor: a battery that never ran is indistinguishable from one that passed #13799

Open

baozhoutao enabled auto-merge September 4, 2026 06:48

claude Bot mentioned this pull request Sep 4, 2026

[PM seat] domain:devx @ objectstack — 27 PRs merged 0 rework · lane 70→60 · decision box 2 · 3 devs live + 3 draft PRs open #6023

Open

baozhoutao added this pull request to the merge queue Sep 4, 2026

claude Bot mentioned this pull request Sep 4, 2026

test(scripts): batch 6b — assertion floors for nine scripts/** self-tests (sink repair + hoisted battery) #15248

Merged

claude Bot mentioned this pull request Sep 4, 2026

tooling(scripts): assertion floors for the batch 6a self-tests — sink repair plus one hoisted battery (#13799 batch 6a) #15250

Merged

Merged via the queue into main with commit 54db16d Sep 4, 2026
35 checks passed

baozhoutao deleted the claude/issue-13799-batch5-single-battery branch September 4, 2026 07:16

claude Bot mentioned this pull request Sep 4, 2026

test(scripts): assertion floors for nine self-tests — #13799 batch 8a (b5 hoisted single battery) #15308

Draft

Navigation Menu

Uh oh!

tooling(scripts): assertion floors for the class-2 self-tests, one hoisted battery each (#13799 batch 5) - #15217

tooling(scripts): assertion floors for the class-2 self-tests, one hoisted battery each (#13799 batch 5)#15217
baozhoutao merged 5 commits into
mainfrom
claude/issue-13799-batch5-single-battery

claude Bot commented Sep 4, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

claude Bot commented Sep 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The shape

Floored (9)

⭐ Two floors that are deliberately NOT "every case in the file"

Deferred (2) — measured, and a correction to PR #15003's table

Evidence

Behaviour is unchanged, per file

Case counts: measured, never transcribed

⚠️ One transcribed count HAS drifted, and is filed rather than fixed here

Ablations — three shapes, eleven legs, each from the committed implementation

The census moves by exactly the worklist

Battery names are not path populations

Gates

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

claude Bot commented Sep 4, 2026 •

edited

Loading